working party on agricultural policies and markets drivers of

Organisation for Economic Co-operation and Development

TAD/CA/APM/WP(2020)2/PART2/FINAL

Unclassified English - Or. English

28 May 2020

TRADE AND AGRICULTURE DIRECTORATE

COMMITTEE FOR AGRICULTURE

Working Party on Agricultural Policies and Markets

Drivers of Farm Performance – Part 2

Empirical Country Case Studies

Contact: Catherine Moreddu ([email protected])

JT03462277

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the

delimitation of international frontiers and boundaries and to the name of any territory, city or area.

mailto:[email protected]

2 TAD/CA/APM/WP(2020)2/PART2/FINAL

DRIVERS OF FARM PERFORMANCE Unclassified

Note by the Secretariat

This analysis of drivers of farm performance, mandated under Expected Output

Result 2.1.2.3 of the 2017-18 Programme of Work and Budget of the Committee for

Agriculture aims at empirically identifying the main conditions for, and obstacles to,

productivity and environmental sustainability of different types of farms in selected OECD

countries.

This study has been conducted by Professor Johannes Sauer (Technical University of

Munich) in collaboration with the OECD Farm-Level Analysis Network, and under the

responsibility of Catherine Moreddu.

The findings of the analysis are reported in three documents:

The Part 1 document contains a short description of the methodological framework

and data, an overview of findings by farm type, and conclusions

[TAD/CA/APM/WP(2020)2/PART1/FINAL]. It also includes three annexes.

Annex A contains the detailed methodology framework applied; Annex B contains

a review of the literature, and Annex C contains tables supporting the overview of

findings.

The Part 2 document contains the results of individual farm cases

[TAD/CA/APM/WP(2020)2/PART2/FINAL].

An annex document contains a series of background tables for each case study,

including descriptive statistics and estimation results (Annex D)

[TAD/CA/APM/WP(2020)2/ANN/FINAL].

This document contains Part 2 of the report.

These three documents were declassified by the Working Party on Agricultural Policies

and Markets on 17-18 March 2020.

TAD/CA/APM/WP(2020)2/PART2/FINAL 3


Table of contents

Part 2. Discussion of results by country .............................................................................................. 8

1. Australia ............................................................................................................................................. 8

1.1. Australian dairy farms................................................................................................................... 8 1.2. Australian crop farms .................................................................................................................. 13 1.3. Australian mixed crop and livestock farms................................................................................. 16 1.4. Australian beef and sheep farms ................................................................................................. 20 1.5. Australian beef farms .................................................................................................................. 24 1.6. Australian sheep meat farms ....................................................................................................... 28 1.7. Australian wool farms ................................................................................................................. 32

2. Chile .................................................................................................................................................. 36

2.1. Chilean small-scale fruit farms ................................................................................................... 36

3. Czech Republic ................................................................................................................................ 41

3.1. Czech dairy farms ....................................................................................................................... 41

4. Denmark ........................................................................................................................................... 46

4.1. Dairy farms ................................................................................................................................. 46 4.2. Pig farms ..................................................................................................................................... 50

5. Estonia .............................................................................................................................................. 59

5.1. Estonian dairy farms ................................................................................................................... 59

6. France ............................................................................................................................................... 64

6.1. French dairy farms ...................................................................................................................... 64 6.2. French crop farms ....................................................................................................................... 68

7. Hungary ............................................................................................................................................ 74

7.1. Hungarian crop farms ................................................................................................................. 74

8. Ireland .............................................................................................................................................. 78

8.1. Irish dairy farms .......................................................................................................................... 79 8.2. Irish crop farms ........................................................................................................................... 83 8.3. Irish cattle farms ......................................................................................................................... 87 8.4. Irish sheep farms ......................................................................................................................... 97

9. Italy ................................................................................................................................................. 101

9.1. Italian crop farms ...................................................................................................................... 101

10. Korea ............................................................................................................................................ 107

10.1. Korean rice farms .................................................................................................................... 107

11. Norway ......................................................................................................................................... 112

11.1. Norwegian dairy farms ........................................................................................................... 112 11.2. Norwegian crop farms............................................................................................................. 117



11.3. Norwegian cattle farms ........................................................................................................... 121

12. Sweden .......................................................................................................................................... 126

12.1. Swedish dairy farms ................................................................................................................ 126 12.2. Swedish crop farms ................................................................................................................. 130

13. United Kingdom .......................................................................................................................... 135

13.1. UK dairy farms ....................................................................................................................... 135 13.2. UK cereal farms ...................................................................................................................... 139 13.3. UK mixed crop and livestock farms ....................................................................................... 144 13.4. UK pig farms .......................................................................................................................... 148 13.5. UK poultry farms .................................................................................................................... 153

Tables

Table 1.1. Productivity characteristics of Australian dairy farms, by class .......................................... 10 Table 1.2. Multiple characteristics of Australian dairy farms, by class ................................................ 12 Table 1.3. Productivity characteristics of Australian crop farms, by class ............................................ 14 Table 1.4. Multiple characteristics of Australian crop farms, by class .................................................. 15 Table 1.5. Productivity characteristics of Australian crop and livestock farms, by class ..................... 17 Table 1.6. Multiple characteristics of Australian crop and livestock farms, by class ........................... 19 Table 1.7. Productivity characteristics of Australian sheep and beef farms, by class ........................... 22 Table 1.8. Multiple characteristics of Australian beef and sheep farms, by class ................................. 23 Table 1.9. Productivity characteristics of Australian beef farms, by class ............................................ 25 Table 1.10. Multiple characteristics of Australian beef farms, by class ................................................ 27 Table 1.11. Productivity characteristics of Australian sheep meat farms, by class ............................... 30 Table 1.12. Multiple characteristics of Australian sheep meat farms, by class ..................................... 31 Table 1.13. Productivity characteristics of Australian wool farms, by class ......................................... 34 Table 1.14. Multiple characteristics of Australian wool farms, by class ............................................... 35 Table 2.1. Productivity characteristics of Chilean small-scale fruit farms, by class ............................. 38 Table 2.2. Multiple characteristics of Chilean small-scale fruit farms, by class ................................... 40 Table 3.1. Productivity characteristics of Czech dairy farms, by class ................................................. 43 Table 3.2. Multiple characteristics of Czech dairy farms, by class ....................................................... 45 Table 4.1. Productivity characteristics of Danish dairy farms, by class ................................................ 47 Table 4.2. Multiple characteristics of Danish dairy farms, by class ...................................................... 50 Table 4.3. Productivity characteristics of Danish rearing and fattening pig farms, by type and class .. 52 Table 4.4. Multiple characteristics of Danish rearing and fattening pig farms, by class ....................... 55 Table 4.5. Productivity characteristics of Danish specialised fattening pig farms, by type and class ... 57 Table 4.6. Multiple characteristics of Danish specialised fattening pig farms, by class ....................... 59 Table 5.1. Productivity characteristics of Estonian dairy farms by class .............................................. 61 Table 5.2. Multiple characteristics of Estonian dairy farms, by class ................................................... 63 Table 6.1. Productivity characteristics of French dairy farms, by class ................................................ 65 Table 6.2. Multiple characteristics of French dairy farms, by class ...................................................... 67 Table 6.3. Productivity characteristics of French crop farms, by class ................................................. 70 Table 6.4. Multiple characteristics of French crop farms, by class ....................................................... 72 Table 7.1. Productivity characteristics of Hungarian crop farms, by class ........................................... 75 Table 7.2. Multiple characteristics of Hungarian crop farms, by class ................................................. 77 Table 8.1. Productivity characteristics of Irish dairy farms, by class .................................................... 80 Table 8.2. Multiple characteristics of Irish dairy farms, by class .......................................................... 82



Table 8.3. Productivity characteristics of Irish crop farms, by class ..................................................... 84 Table 8.4. Multiple characteristics of Irish crop farms, by class ........................................................... 86 Table 8.5. Productivity characteristics of Irish cattle rearing farms, by class ....................................... 89 Table 8.6. Multiple characteristics of Irish cattle rearing farms, by class ............................................. 91 Table 8.7. Productivity characteristics of Irish “cattle other” farms, by class ....................................... 94 Table 8.8. Multiple characteristics of Irish “cattle other” farms, by class ............................................. 95 Table 8.9. Productivity characteristics of Irish sheep farms, by class ................................................... 98 Table 8.10. Multiple characteristics of Irish sheep farms, by class ..................................................... 100 Table 9.1. Productivity characteristics of Italian crop farms, by class ................................................ 104 Table 9.2. Multiple characteristics of Italian crop farms, by class ...................................................... 106 Table 10.1. Productivity characteristics of Korean rice farms, by class .............................................. 108 Table 10.2. Multiple characteristics of Korean rice farms, by class.................................................... 111 Table 11.1. Productivity characteristics of Norwegian dairy farms, by class ..................................... 114 Table 11.2. Multiple characteristics of Norwegian dairy farms, by class ........................................... 116 Table 11.3. Productivity characteristics of Norwegian crop farms, by class ...................................... 118 Table 11.4. Multiple characteristics of Norwegian crop farms, by class ............................................ 121 Table 11.5. Productivity characteristics of Norwegian cattle farms, by class ..................................... 123 Table 11.6. Multiple characteristics of Norwegian cattle farms, by class ........................................... 125 Table 12.1. Productivity characteristics of Swedish dairy farms, by class.......................................... 127 Table 12.2. Multiple characteristics of Swedish dairy farms, by class ................................................ 129 Table 12.3. Productivity characteristics of Swedish crop farms, by class ........................................... 132 Table 12.4. Multiple characteristics of Swedish crop farms, by class ................................................. 133 Table 13.1. Productivity characteristics of UK dairy farms, by class ................................................. 136 Table 13.2. Multiple characteristics of UK dairy farms, by class ....................................................... 138 Table 13.3. Productivity characteristics of UK cereal farms, by class ................................................ 141 Table 13.4. Multiple characteristics of UK cereal farms, by class ...................................................... 142 Table 13.5. Productivity characteristics of UK mixed crop and livestock farms, by class ................. 145 Table 13.6. Multiple characteristics of UK mixed crop and livestock farms, by class ....................... 147 Table 13.7. Productivity characteristics of UK pig farms, by class .................................................... 150 Table 13.8. Multiple characteristics of UK pig farms, by class .......................................................... 151 Table 13.9. Productivity characteristics of UK poultry farms, by class .............................................. 154 Table 13.10. Multiple characteristics of UK poultry farms, by class .................................................. 156

Figures

Figure 1.1. Productivity and technical change for Australian dairy farms, by class, 1989 to 2018 ........ 9 Figure 1.2. Multi-dimensional indices for Australian dairy farms ........................................................ 11 Figure 1.3. Productivity and technical change for Australian crop farms, by class, 1989 to 2018 ....... 13 Figure 1.4. Multi-dimensional indices for Australian crop farms ......................................................... 15 Figure 1.5. Productivity and technical change for Australian crop and livestock farms, by class,

1989 to 2018 .................................................................................................................................. 17 Figure 1.6. Multi-dimensional indices for Australian crop and livestock farms ................................... 19 Figure 1.7. Productivity and technical change for Australian beef and sheep farms, by class, 1989

to 2018 ........................................................................................................................................... 21 Figure 1.8. Multi-dimensional indices for Australian sheep and beef farms ......................................... 23 Figure 1.9. Productivity and technical change for Australian beef farms, by class, 1989 to 2018........ 25 Figure 1.10. Multi-dimensional indices for Australian beef farms ....................................................... 27 Figure 1.11. Productivity and technical change for Australian sheep meat farms, by class, 1989 to

2018 ............................................................................................................................................... 29 Figure 1.12. Multi-dimensional indices for Australian sheep meat farms ............................................ 31



Figure 1.13. Productivity and technical change for Australian wool farms, by class, 1989 to 2018 .... 33 Figure 1.14. Multi-dimensional indices for Australian wool farms ...................................................... 35 Figure 2.1. Productivity for Chilean small-scale fruit farms, by class, 2015 ........................................ 37 Figure 2.2. Multi-dimensional indices for Chilean small-scale fruit farms ........................................... 40 Figure 3.1. Productivity and technical change for Czech dairy farms, by class, 2005 to 2015 ............. 42 Figure 3.2. Multi-dimensional indices for Czech dairy farms ............................................................... 44 Figure 4.1. Productivity and technical change for Danish dairy farms, by class, 2010 to 2016 ........... 47 Figure 4.2. Multi-dimensional indices for Danish dairy farms ............................................................. 49 Figure 4.3. Productivity and technical change for Danish rearing and fattening pig farms, by type

and class, 2006 to 2016 ................................................................................................................. 52 Figure 4.4. Multi-dimensional indices for Danish rearing and fattening pig farms, by type ................ 54 Figure 4.5. Productivity and technical change for Danish specialised fattening pig farms, by type

and class, 2006 to 2016 ................................................................................................................. 56 Figure 4.6. Multi-dimensional indices for Danish specialised fattening pig farms, by type ................. 58 Figure 5.1. Productivity and technical change for Estonian dairy farms, by class, 2000 to 2015 ......... 61 Figure 5.2. Multi-dimensional indices for Estonian dairy farms ........................................................... 63 Figure 6.1. Productivity and technical change for French dairy farms, by class, 1990 to 2013 ............ 65 Figure 6.2. Multi-dimensional indices for French dairy farms .............................................................. 66 Figure 6.3. Productivity and technical change for French crop farms, by class, 1989 to 2016 ............. 69 Figure 6.4. Multi-dimensional indices for French crop farms ............................................................... 71 Figure 7.1. Productivity and technical change for Hungarian crop farms, by class, 2001 to 2014 ....... 75 Figure 7.2. Indices for Hungarian crop farms ....................................................................................... 76 Figure 8.1. Productivity and technical change for Irish dairy farms, by class, 2010 to 2018 ............... 80 Figure 8.2. Multi-dimensional indices for Irish dairy farms ................................................................. 81 Figure 8.3. Productivity and technical change for Irish crop farms, by class, 2010 to 2018 ................. 84 Figure 8.4. Multi-dimensional indices for Irish crop farms .................................................................. 86 Figure 8.5. Productivity and technical change for Irish cattle rearing farms, by class, 2010 to 2018 ... 89 Figure 8.6. Multi-dimensional indices for Irish cattle rearing farms ..................................................... 91 Figure 8.7. Productivity and technical change for Irish “cattle other” farms, by class, 2010 to 2018 .. 93 Figure 8.8. Multi-dimensional indices for Irish “cattle other” farms .................................................... 95 Figure 8.9. Productivity and technical change for Irish sheep farms, by class, 2010 to 2018 ............... 98 Figure 8.10. Multi-dimensional indices for Irish sheep farms ............................................................ 100 Figure 9.1. Productivity and technical change for Italian crop farms, by class, 2008 to 2015 ............ 103 Figure 9.2. Multi-dimensional indices for Italian crop farms .............................................................. 105 Figure 10.1. Productivity and technical change for Korean rice farms, by class, 2003 to 2015 ......... 108 Figure 10.2. Multi-dimensional indices for Korean rice farms ........................................................... 110 Figure 11.1. Productivity and technical change for Norwegian dairy farms, by class, 2005 to 2016 . 113 Figure 11.2. Multi-dimensional indices for Norwegian dairy farms ................................................... 115 Figure 11.3. Productivity and technical change for Norwegian crop farms, by class, 2005 to 2016 .. 117 Figure 11.4. Multi-dimensional indices for Norwegian crop farms .................................................... 119 Figure 11.5. Productivity and technical change for Norwegian cattle farms, by class, 2005 to 2016 . 122 Figure 11.6. Multi-dimensional indices for Norwegian cattle farms ................................................... 124 Figure 12.1. Productivity and technical change for Swedish dairy farms, by class, 1997 to 2017 ..... 127 Figure 12.2. Multi-dimensional indices for Swedish dairy farms ....................................................... 128 Figure 12.3. Productivity and technical change for Swedish crop farms, by class, 1997 to 2017 ...... 131 Figure 12.4. Multi-dimensional indices for Swedish crop farms ........................................................ 132 Figure 13.1. Productivity and technical change for UK dairy farms, by class, 1995 to 2017 ............. 136 Figure 13.2. Multi-dimensional indices for UK dairy farms ............................................................... 137 Figure 13.3. Productivity and technical change for UK cereal farms, by class, 1995 to 2017 ............ 140 Figure 13.4. Multi-dimensional indices for UK cereal farms .............................................................. 142



Figure 13.5. Productivity and technical change for UK mixed crop and livestock farms, by class,

1995 to 2017 ................................................................................................................................ 144 Figure 13.6. Multi-dimensional indices for UK mixed crop and livestock farms ............................... 146 Figure 13.7. Productivity and technical change for UK pig farms, by class, 1995 to 2017 ................ 149 Figure 13.8. Multi-dimensional indices for UK pig farms .................................................................. 150 Figure 13.9. Productivity and technical change for UK poultry farms, by class, 1995 to 2017 .......... 154 Figure 13.10. Multi-dimensional indices for UK poultry farms .......................................................... 155



Part 2. Discussion of results by country

For each country case study in Table 1.1 of Part 1

[TAD/CA/APM/WP(2020)2/PART1/FINAL], Part 2 presents the estimation results of a

latent class estimation of a translog production function applied to the various panel

datasets described in Annex D [TAD/CA/APM/WP(2020)2/ANN/FINAL].1 Annex D also

contains additional tables showing the detailed estimated results by country and farm type.

At a later stage, the country sections of Annex D could be directly attached to the country

sections in this Part 2.

1. Australia

1. For Australia, the analysis applies to a sample of broadacre farms (crop, livestock,

sheep meat, wool sheep, mixed crop and livestock, mixed sheep and beef) and to a sample

of dairy farms, all covering the period 1989 to 2018.

2. Australian agriculture accounts for 58% of Australian land use (446 million ha) and

59% of water extractions. It contributes a significant share to the country’s goods and

services exported (about 14%). The mix of Australian agricultural activity is determined

by climate, water availability, soil type and proximity to markets. Livestock grazing is

widespread, occurring in most areas of Australia, while cropping is generally concentrated

in coastal areas (ABARES, 2018).

1.1. Australian dairy farms

3. According to Table A D.1.1, which contains descriptive statistical measures for the

sample, the average milk output per farm is around AUD 1.5 million (total farm output of

about AUD 1.6 million in 2018. The variable cost items increased over the period 1989 to

2018, and the share of hired labour significantly increased for the average Australian dairy

farm. The stocking density has been around 1.8 livestock units (LU) per ha in 2018 with an

average herd size of about 740 cows per farm (an increase from about 1.2 LU per ha and

230 cows in 1989).

4. In the model used to estimate the technology and the Class identification

components for the Australian dairy production the output variable is total output per farm

and year and input variables are cows, land, capital, materials, fodder, fuel and labour. The

Class identification component is based on the indices related to structure, environmental

sustainability, innovation, technology, diversity, individual, location, household and

financial aspects (Table A D.1.2).

5. For Australian dairy farms, three distinct technology classes emerge from the model

estimates (Table A D.1.2). The individual farms are distributed very unevenly across the

three technology classes (Table 1.1), as Class 3 includes about three-quarters of all farms.

6. Farms in Class 1 show a significantly higher productivity performance

— measured as the potential output levels that could be achieved with a given input bundle,

in value per year — than their milk-producing counterparts in classes 2 and 3 (Figure 1.1).

1 The routine offered by the econometric software LIMDEP (version 11) is used for the estimation.



Dairy farms in Class 1 produce about 4 times more per farm and year than farms in Class 2

and about 3 times more per farm and year than dairy farms in Class 3.

7. Dairy farms in Class 1 show a significantly positive technical change (of 1.74% per

year) but it is lower than technical change rate for farms in Class 2 (estimated at 2.86% per

year) and close to that in Class 3 (estimated at 1.79% per year). Hence, although dairy farms

in Class 1 are far more productive, they also seem to increase their productivity at a lower

rate than those in the least productive class, based on a lower technical change increase

ceteris paribus (Figure 1.1).

Figure 1.1. Productivity and technical change for Australian dairy farms,

by class, 1989 to 2018

Source: Table 1.1.

8. If dairy farms in Class 2 produced with the technology of farms in Class 1, they

would be able to more than triple their productivity, and farms in Class 3 would multiply

their productivity by nearly 2.5 (Table 1.1). Overall, if all farms were producing with

Class 1 technology, the productivity of Australian dairy farms would almost double

(Table A C.4 in Part 1).

9. The estimated production functions for each farm class are highly significant and

dairy farms in Class 1 and Class 3 exhibit slightly or significant increasing returns to scale

of respectively 1.0395 and 1.1122, whereas dairy farms in Class 2 exhibit decreasing

returns of around 0.9479 (see Table A D.1.3 for complete class related elasticities).

Switching technologies could — ceteris paribus — result in higher productivity per farm

and also a significantly higher technical change rate for most farms (especially regarding a

switch to Class 1 technology). Various indices reflecting the different dimensions along

which dairy farms can be distinguished are used to robustly identify the reported farm

classes: Farm structure, environmental sustainability of operations, technology

characteristics, degree of diversity, individual characteristics, locational specifics and

financial characteristics. Table A D.1.2 summarises the estimates for the various indices

that were used as components for the Class identifying vector in the latent class estimation.

1.74

2.86

1.79

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0

100

200

300

400

500

600

700

800

Class 1 (19.6%) Class 2 (4.7%) Class 3 (75.7%)

%EUR '000

Productivity level (EUR '000) Technical change (% per year)



Most of the indices considered showed significant estimates. To decide empirically on the

most appropriate number of classes to estimate, various statistical quality tests have been

performed (most prominently the Akaike Information Criteria AIC) which holds for all

country cases considered in this study.

Table 1.1. Productivity characteristics of Australian dairy farms, by class

Latent Class Estimation, Panel 1989 to 2018

Class 1 Class 2 Class 3

Number of observations (% of sample farms) 19.6% 4.7% 75.7%

Prior probability of class membership 0.2886 0.1501 0.5614

Posterior probability of class membership 0.2921 0.1437 0.5641

Productivity level (AUD per year)1

- Class 1 Technology 1 066 148*** 783 558*** 928 848***

- Class 2 Technology 331 439*** 247 704*** 245 438***


Technical Change (% per year)

- Class 1 Technology 1.743*** 1.960*** 1.818***


- Class 3 Technology 6.368*** 1.061* 1.792***

Note: 1. Fitted values at sample means, 2- * significant at 10%, ** significant at 5%, *** significant at 1%.

Source: Own estimations based on latent class model estimates and derivatives.

10. Figure 1.2 summarises the various dairy farm classes in terms of estimated indices,

while Table 1.2 contains the estimates for variables used to construct the indices.

11. Dairy farms in Class 1, which are the most productive and exhibit the lowest

technical change rate per year, have the lowest share of family labour in total labour. They

are significantly larger than the average Australian dairy farm in terms of herd size, but

they still experience increasing returns to scale. Dairy farms in Class 1 score relatively high

on environmental sustainability indicators, as they have the lowest stocking density and

chemicals use per ha. These farms show the highest scores on innovation and

commercialisation with a significantly higher than average investment rate, rented land

share, and use of contract farming. Class 1 farms show a lower than average capital per

labour intensity, while using more capital per cow than the average dairy farm in Australia,

based on high levels of total assets endowment. These farms are less diversified than the

average dairy farm in Australia, and their managers are older and more educated than

average, and more likely to be male. Finally, dairy farms in Class 1 have a higher share of

off-farm income than the national average, receive more subsidies and have a lower debt

ratio (Table 1.2).



Figure 1.2. Multi-dimensional indices for Australian dairy farms

Scaled values at class means, 1989 to 2018

Source: Table A D.1.5.

12. Three-quarters of all farms are in Class 3. They are characterised by intermediary

productivity levels and environmental sustainability scores in terms of stocking density and

chemical application per ha. They are smaller and more specialised than average. They are

less innovative than the average dairy farm in Australia, with lower net investment, lower

share of rented land and use of contract farming. They use more extensive farm practices

than the average dairy farm in Australia, but not to the same extent as the most

environmentally sustainable farms in Class 2. They receive less subsidies than average and

have a lower equity over debt ratio.

13. Class 2 farms are the most environmentally sustainable and the least productive

farms but they are catching up. Only accounting for 5% of all dairy farms in Australia, they

are less innovative, smaller and more family-driven than the average, and their production

is more diversified, and less intensive, with lower stocking density and chemical use per

ha. They receive lower subsidies than the average dairy farm in Australia and have the

highest share of off-farm income.

14. In summary, Australian dairy farms in the three identified technology classes differ

significantly with respect to their economic performance as well as technical development

over time. Dairy farms in Australia show a negative correlation between the productivity

and environmental sustainability of their operations. However, the majority of farms are

medium productive and environmentally sustainable. More productive farms are

significantly less family-driven but more innovative. They employ more non-family labour

and operate with larger herds. These farms also perform relatively well in terms of financial

indicators as, for example, total assets and financial liquidity.

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6Index 1 - Structure

Index 2 - Environmentalsustainability

Index 3 - Innovation/Commercialisation

Index 4 - Technology

Index 5 - DiversityIndex 6 - Individual

Index 7 - Location

Index 8 - Household

Index 9 - Financial

Class 1: Most productive (19.6%)

Class 2: Least productive (4.7%)

Class 3: Medium productive (75.7%)



Table 1.2. Multiple characteristics of Australian dairy farms, by class Deviations from sample means1, 1989 to 2018




Farm structure

Family/hired labour ratio -0.0714 0.2993 0.0001

Herd size (LU) 0.7647 -0.4171 -0.1729

Form of ownership (1=company, 2=partnership/trust, 3=sole trader)

0.0273 0.3165 -0.0265

Environmental sustainability

Stocking density (LU per ha) 0.4044 -0.4010 -0.0803

Chemicals use (AUD per ha) 0.7304 -0.2839 -0.1721

Innovation-commercialisation

Net investment ratio (per total assets) 0.4010 -0.1398 -0.0955

Contract farming (1=yes, 0=no) 0.2434 -0.3093 -0.0442

Share land rented 0.2496 -0.1307 -0.0567

Technology

Capital / labour ratio (AUD per AWU) -0.0202 -0.0629 0.0091

Capital per cow (AUD per LU) 0.0326 0.3640 -0.0308

Fodder per cow (AUD per LU) 0.5861 -0.0725 -0.1477

Labour per cow (AWU per LU) -0.0441 -0.0879 0.0169

Herdtest (1=yes, 0=no) 0.5593 -0.3087 -0.1262

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) 0.0937 -0.3725 -0.0014

Production diversity (yc/ΣY) -0.0563 -0.3339 0.0351

Individual

Age (years)15 0.0796 0.0697 -0.0249

Education (various levels) 0.3553 -0.1373 -0.0838

Gender (1=male, 2=female) 0.3050 0.1654 -0.0893

Location

Region subtropical Dairy NSW -0.0205 -0.0433 0.0080

Region Dairy NSW 0.0947 -0.2229 -0.0109

Region NSW Murray dairy 0.0582 -0.1062 -0.0086

Region West Victoria dairy -0.0764 0.0573 0.0163

Region Vic Murray dairy -0.0497 -0.1329 0.0211

Region Vic Gippsland dairy -0.0171 0.1254 -0.0033

Region subtropical dairy Qld -0.0244 0.0169 0.0053

Region Dairy SA -0.0721 0.0576 0.0152

Region Dairy WA 0.0508 -0.0045 -0.0129

Region Dairy Tas 0.0577 0.2506 -0.0304

(all: 1=yes, 0=no)

Household

Off-farm income share -0.2005 0.6522 0.0120

Age spouse (years) 0.0599 0.0347 -0.0177

Education spouse (various levels) 0.1653 -0.1135 -0.0359

Gender spouse (1=male, 2=female) 0.2229 0.0392 -0.0603

Financial

Total assets (AUD) 0.8336 -0.2611 -0.2003

Total subsidies (AUD) 0.6784 -0.1329 -0.1679

Equity/debt ratio 0.0925 -0.0240 -0.0225

Note: LU: Livestock Unit. AWU: Annual Work Unit. NSW: New-South-Wales; Vic: Victoria; SA: South

Australia; WA: Western Australia; Tas: Tasmania; Qld: Queensland.

1. Deviations from sample means (=0), z-scores based, scaled values.

Source: Estimations.



1.2. Australian crop farms


sample, the average crop output per farm is around AUD 1.77 million (total farm output of

about AUD 2.08 million in 2018). The variable cost items increased over time and the share

of hired labour significantly increases for the average Australian crop farm. The average

cropping farm cultivated about 4 060 ha of land in 2018 (compared to about 1 610 ha in

1989).


components for the Australian crop production the output variable is total output per farm

and year and input variables are land, capital, chemicals, labour, fuel and materials. The




17. For Australian crop farms, two distinct technology classes emerge from the model


three technology classes (Table 1.3), as Class 1 includes over 85% of all farms.



in value per year — than their counterparts in Classes 1 (Figure 1.1). Crop farms in Class 1

produce over 3 times more per farm and year than farms in Class 2. Moreover, the

productivity gap between the two classes is increasing as the least productive farms show

a negative rate of technical change, while for the most productive class, technical change

is increasing, ceteris paribus.

Figure 1.3. Productivity and technical change for Australian crop farms,


Source: Table 1.3.

0.24

-0.72

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0

50

100

150

200

250

300

350

400

450

Class 1 (87.6%) Class 2 (12.4%)

%EUR '000




19. If crop farms in Class 2 produced with the technology of farms in Class 1, they

would be able to double their productivity (Table 1.3). Given the small share of Class 2

farms in the total, the overall productivity of Australian crop farms would increase by about

20% (Table A C5 in Part 1).

Table 1.3. Productivity characteristics of Australian crop farms, by class


Class 1 Class 2

Number of observations (% of sample farms) 87.6 12.4

Prior probability of class membership 0.8938 0.1062

Posterior probability of class membership 0.8228 0.1772


- Class 1 Technology 641 715*** 409 094***



- Class 1 Technology 0.236*** -0.935***

- Class 2 Technology 0.580*** -0.716*

Note: * significant at 10%, ** significant at 5%, *** significant at 1%. 1. Fitted values at sample means.


20. The estimated production functions for each farm class are highly significant. Crop

farms in Class 2 exhibit increasing returns to scale of about 1.0903 whereas farms in

Class 1 show slightly decreasing returns to scale of about 0.9521 (Table A D.1.9).

21. Figure 1.4 summarise the various crop farm classes in terms of estimated indices,


22. Class 1 includes the most productive and least environmentally sustainable farms,

which largely dominate the sector. Their lower environmental sustainability performance

reflects more intensive practices than the average crop farm in Australia. They are larger

operations that are more likely to be partnerships and to be engaged in contracting farm

work. They have less diversified agricultural productions, higher investment in new

technologies and farm more intensively. Their managers are younger and more educated

than the average of all crop farms in Australia.

23. The small share of most environmentally sustainable and least productive farms in

Class 2 are smaller than average in terms of area, more family-driven, and more diversified

operations, with more extensive farming practices (fuel and chemical use per ha). They are

managed by older farmers and are more reliant on off-farm income. They have lower

investment in new technologies and practices, lower debt ratios, but are more capital

intensive than the average of all crop farms in Australia.

24. In summary, Australian crop farms in the two identified technology classes differ


over time. More productive farms are slightly less family-driven and environmentally

sustainable than less productive farms. Innovative farms are again more likely to show a

higher productivity. Farms’ productivity seems not necessarily correlated positively with

land endowment but with the share of hired labour. More productive crop farms operate

with a higher technology intensity than their less productive colleagues, ceteris paribus.



Figure 1.4. Multi-dimensional indices for Australian crop farms



Table 1.4. Multiple characteristics of Australian crop farms, by class

Deviations from sample means1, 1989 to 2018



Farm structure

Family/hired labour ratio -0.0063 0.0446

Land (ha) -0.0020 0.0141


-0.0290 0.2047


Fuel per ha (AUD per ha) 0.0278 -0.1959

Chemicals use (AUD per ha) 0.0656 -0.4632


Net investment ratio (per total assets) 0.0274 -0.1934

Contract farming (1=yes, 0=no) 0.0783 -0.5527

Share land rented 0.0341 -0.2405

Technology

Capital / labour ratio (AUSD per AWU) 0.0021 -0.0149

Capital per ha (AUD per ha) -0.0105 0.0074

Seed per ha (AUD per ha) 0.0160 -0.1132

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) 0.1884 -1.3297

Production diversity (yc/ΣY) 0.1884 -1.3237

Individual

Age (years) -0.0151 0.1065

Education (various levels) 0.0362 -0.2554

Gender (1=male, 2=female) 0.0049 -0.1135

Location

Region pastoral zone -0.0218 0.1537

Region wheat-sheep zone 0.0404 -0.2851

Region high-rainfall zone -0.0271 0.1910

(all: 1=yes, 0=no)

Household

-2

-1

0

1

2Index 1 - Structure





Index 7 - Location

Index 8 - Household

Index 9 - Financial





Off-farm income share -0.0393 0.2776

Age spouse (years) 0.0058 -0.0411

Education spouse (various levels) 0.0354 -0.2500

Gender spouse (1=male, 2=female) 0.0248 -0.1749

Financial

Total assets (AUS$) 0.0430 -0.3039

Total subsidies (AUS$) -0.0105 0.0739

Equity/debt ratio -0.0117 0.0827

Note: AWU: Annual Work Unit. 1. Deviations from sample means (=0), z-scores based, scaled values.


1.3. Australian mixed crop and livestock farms

25. According to Table A D.1.11, which contains descriptive statistical measures for

the sample, the average crop and livestock output per farm in 2018 has been about

AUD 1.4 million. The variable cost items increased over time and the share of hired labour

significantly increases for the average Australian crop and livestock farm. In 2018, the

stocking density was about 1.65 LU per ha, and the average farm had a land endowment of

about 7 180 ha (compared with about 1.74 LU per ha and 2 850 ha in 1989).


components for the Australian crop and livestock production the output variable is total

output per farm and year and input variables are land, capital, labour, livestock units and

other materials. The Class identification component is based on the indices related to

structure, environmental sustainability, innovation, technology, diversity, individual,

location, household and financial aspects (Table A D.1.12).

27. For Australian crop and livestock farms, three distinct technology classes emerge

from the model estimates (Table A D.1.12). The individual farms are distributed very

unevenly across the three technology classes (Table 1.5), as Class 1 includes 86% of all

crop and livestock farms.

28. Farms in Class 1 achieve a higher productivity performance — measured as the

potential output levels that could be achieved with a given input bundle, in value per year

— than their counterparts in classes 2 and 3 (Figure 1.5). Farms in Class 1 produce about

3 times more than farms in Class 3, but the productivity level of farms in Class 2 is only

23% lower than that of the highest performers in Class 1.

29. Crop and livestock farms in Class 2 show a significantly positive technical change,

which is much higher than that for farms in Class 1, indicating that the productivity

difference between the two classes is narrowing. But the performance gap between farms

in Class 3 and other crop and livestock farms in Australia is increasing as the former

experience a negative technical change, ceteris paribus (Figure 1.5).



Figure 1.5. Productivity and technical change for Australian crop and livestock farms,


Source: Table 1.7.

30. If crop and livestock farms in Class 2 produced with the technology of farms in

Class 1, they would be able to increase their productivity by 30%, while the gain for Class 3

farms would be much larger (Table 1.7). Overall, if all farms were producing with Class 1

technology, the productivity of Australian crop and livestock farms would increase by 5%,

reflecting the dominance of Class 1 farms in the total number of farms (Table A C.6 in

Part 1).

Table 1.5. Productivity characteristics of Australian crop and livestock farms, by class











- Class 1 Technology 0.040* 0.924* -2.079***

- Class 2 Technology 0.135** 0.550* -2.325***

- Class 3 Technology 0.455*** -0.681* -0.747*




all dairy farm classes exhibit increasing returns to scale: 1.0331 for farms in Class 1, 1.1912

for farms in Class 2, and 1.1009 for farms in Class 3 (Table A D.1.14).

0.04

0.55

-0.75

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

0

50

100

150

200

250

300


%EUR '000




32. Figure 1.6 summarises the various crop and livestock farm classes in terms of

estimated indices, while Table 1.6 contains the estimates for variables used to construct the

indices.

33. Class 1 groups over 85% of all crop and livestock farms in Australia: The most

productive and least environmentally sustainable. There are more specialised farms, with

an average area, but larger herds, and thus higher stocking density. They are more capital

intensive and likely to be engaged in contract farming. They are operated by younger

farmers, and have lower off-farm income.

34. Class 2 farms (about 5% of all farms) achieve medium productivity and above

average environmental sustainability. They are larger, and more family driven operations,

more likely to be partnerships. They are more diversified and more extensive operations.

They invest more in new technologies, and are less likely to be engaged in contract farming.

They receive more subsidies than average and have lower debt ratios.

35. Class 3 groups the most environmentally sustainable farms, which are the least

productive and account for the remaining 8% of all farms. They are smaller operations with

higher stocking density but lower use of chemicals. They are the most specialised farms.

They are more capital intensive per ha, but invest less in new technologies than average.

Operators are older and more likely to be men. They have the highest share of off-farm

income.

36. In summary, Australian crop and livestock farms in the three identified technology

classes differ with respect to their economic performance as well as technical development

over time. The large majority of these farms (about 86%) produce with a high productivity

and only a slightly lower than average environmental sustainability. These farms are based

on an average family labour share that operate large herd sizes and relatively large land

endowments. The least productive mixed crop-livestock farms in the sector use more than

average hired labour input but produce with significantly less than average land endowment

and livestock units. Those farms are least innovative and score relatively low on financial

stability and liquidity indicators.



Figure 1.6. Multi-dimensional indices for Australian crop and livestock farms



Table 1.6. Multiple characteristics of Australian crop and livestock farms, by class





Farm structure


Herd size (LU) 1.1223 0.1841 -1.3063

Land (ha) 0.0097 0.0564 -0.1360


-0.0187 0.3006 0.0012


Stocking density (LU per ha) 0.1981 -1.3117 1.1136

Chemicals use (AUD per ha) 0.0484 -0.0888 -0.4430

Fuel per ha (AUD per ha) 0.0144 -0.0034 -0.0513


Net investment ratio (per total assets) -0.0141 0.3817 -0.0639

Contract farming (1=yes, 0=no) 0.0476 -0.2218 -0.1869

Share land rented 0.0094 0.0143 -0.0640

Genetics veterinary expenses (AUD per LU) 0.0080 0.0111 -0.0385

Insurance expenses (AUD per LU) -0.0079 0.0046 0.0259

Technology

Capital / labour ratio (AUD per AWU) 0.0067 -0.0003 -0.0691

Veterinary expenses (wo genetics) (AUD per LU) -0.0021 0.0236 0.0260

Share crop land 0.0111 -0.0177 -0.0279

LU per labour (LU per AWU) 0.0631 -0.2860 -0.4687

Capital per LU (AUD per LU) -0.0136 0.2636 -0.0278

Capital per ha (AUD per LU) -0.0338 0.0973 0.2864

Seed per ha (AUD per ha) -0.0218 0.0517 0.0420

-5

-4

-3

-2

-1

0

1

2

3






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Herdtest (1=yes, 0=no) -0.0540 1.0124 -0.0887

Diversity


Production diversity (yc/ΣY) 0.1343 -0.3422 -1.1837

Individual

Age (years) -0.0296 0.0408 0.2789

Education (various levels) 0.0028 0.0482 -0.0595

Gender (1=male, 2=female) -0.0370 0.1720 0.2724

Location

Region pastoral zone -0.0034 0.0479 0.0046

Region wheat-sheep zone 0.0066 -0.2597 0.0975

Region high-rainfall zone (all: 1=yes, 0=no) -0.0152 0.2518 -0.0036

Household


Age spouse (years) -0.0019 -0.2598 0.1853


Gender spouse (1=male, 2=female) -0.0060 -0.2447 0.2180

Financial

Total assets (AUD) 0.0291 0.0724 -0.3460

Total subsidies (AUD) -0.0042 0.1854 -0.0744

Equity/debt ratio 0.0019 0.3627 -0.0145

Note: LU: Livestock Unit. AWU: Annual Work Unit.



1.4. Australian beef and sheep farms


the sample, the average beef and sheep output per farm was about AUD 1.2 million in 2018

(an increase from about AUD 320 000 in 1989). The variable cost items increased over

time and the share of hired labour significantly increases for the average Australian sheep

and beef farm. The average farm had about 13 700 LU in 2018 resulting in an average

stocking density of about 2.4 LU per ha (compared to about 27 000 LU and about 3 LU per

ha in 1989).


components for the Australian beef and sheep meat production the output variable is total

output per farm and year and input variables are land, capital, labour, livestock units, and

fodder. The Class identification component is based on the indices related to structure,

environmental sustainability, innovation, technology, diversity, individual, location,

household and financial aspects (Table A D.1.17).

39. For Australian beef and sheep farms, three distinct technology classes emerge from

the model estimates (Table A D.1.17). The individual farms are distributed very unevenly

across the three technology classes (Table 1.7), as Class 3 includes about three-quarters of

all farms.

40. Beef and sheep farms in Class 2 show a significantly higher productivity

performance — measured as the potential output levels that could be achieved with a given

input bundle, in value per year — than their counterparts in classes 1 and 3 (Figure 1.7).

Dairy farms in Class 1 produce only 15% of the output level obtained by farms in Class 2,

while beef and sheep farms in Class 3 reach 40% of Class 2 achievements.



41. The performance gap between farms in Class 1 and farms in other classes is

increasing as the former show a significantly positive technical change while technical

change is negative for classes 1 and 3 (Figure 1.7). At the same time, productivity levels in

Classes 1 and 3 are likely to converge to lower results at some stage, as the rate of decline

in technical change is larger for the medium performing farms in Class 3.

Figure 1.7. Productivity and technical change for Australian beef and sheep farms, by class,

1989 to 2018

Source: Table 1.7.

42. If beef and sheep farms in Class 3 produced with the technology of farms in Class 2,

they would be able to increase their productivity by 55%, while farms in Class 1 would

triple their productivity with the most performing technology (Table 1.7). Overall, if all

farms were producing with Class 2 technology, the productivity of Australian beef and

sheep farms would increase by close to 40% (Table A C6 in Part 1).

43. The estimated production functions for each farm class are highly significant. Beef

and sheep farms in Class 3 exhibit increasing returns to scale of about 1.1133, farms in

Class 2 show constant returns to scale (1.0106), and farms in Class 1 experience

significantly decreasing returns (Table A D.1.19).

44. Figure 1.8 summarises the various beef and sheep farm classes in terms of


indices.

-0.37

0.20

-1.11-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0

50

100

150

200

250


%EUR '000




Table 1.7. Productivity characteristics of Australian sheep and beef farms, by class






Productivity level (EUR per year)1





- Class 1 Technology -0.367* 4.868*** -1.280**

- Class 2 Technology 1.680* 0.195* 0.695***

- Class 3 Technology -1.043* 3.020*** -1.111***



45. Accounting for less than 10% of all Australian beef and sheep farms, Class 1 farms

are the most environmentally sustainable and the least productive. They are the smallest

operations in terms of herd size and the most family-driven. They use the most extensive

farming practices but are capital intensive. They have a lower asset endowment and lower

debt ratios. They are more likely to be located in pastoral zone, and to be more dependent

on off-farm income.

46. Class 2 groups the most productive farms (17% of all farms), which are least

environmentally sustainable. They are the largest operations, more likely to be organised

as partnerships and to be located in a high-rainfall zone. Their manager is older than

average and better educated. They have the highest levels of investment in new

technologies and are more likely to use contract farming than other farms. They are

endowed with larger assets and are slightly more indebted than average.

47. Close to three-quarters of farms in Class 3 have below average productivity, but

higher environmental sustainability. They are larger than average operations, with lower

levels of investments. They use more extensive farming practices and invest less than the

average farm.

48. In summary, Australian beef and sheep farms in the three identified technology

classes differ significantly with respect to their economic performance as well as technical

development over time. Most productive farms are found to be based on a relatively high

family labour endowment but also large herd sizes. These productive farms produce,

however, with a significantly lower than average environmental sustainability and are only

of medium innovativeness but high technology intensity. Least productive farms, on the

other side, are higher than average environmentally sustainable. The majority of farms in

this sector (around 74%) show a medium productivity level. However, the latter are still

producing with a higher than average environmental sustainability, ceteris paribus.



Figure 1.8. Multi-dimensional indices for Australian sheep and beef farms



Table 1.8. Multiple characteristics of Australian beef and sheep farms, by class





Farm structure

Family/hired labour ratio 0.2384 0.0646 -0.0423

Herd size (LU) -1.0108 0.9888 0.0220

Land (ha) -0.0385 -0.2545 0.0639


-0.0114 -0.1447 0.0351


Stocking density (LU per ha) -0.5083 0.9691 -0.1684

Chemicals use (AUD per ha) -0.3410 0.7735 -0.1418

Fuel per ha (AUD per ha) -0.2468 0.7914 -0.1567



Contract farming (1=yes, 0=no) -0.3491 0.2605 -0.0210


Genetics veterinary expenses per LU (AUD per LU) 0.2812 0.0809 -0.0510

Insurance expenses per LU (AUD per LU) 0.9523 -0.0356 -0.1005

Technology


Veterinary expenses (wo genetics) per LU (AUD per LU)

-0.2042 -0.0353 -0.0510

Share crop land 0.1499 -0.3876 0.0734

LU per labour (LU per AWU) 0.1496 -0.4594 0.0902

Capital per LU (AUD per LU) 0.1562 -0.0382 -0.0089

Herdtest (1=yes, 0=no) -0.0287 0.4090 -0.0923

-1.2-1

-0.8-0.6-0.4-0.2

00.20.40.60.8






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) -0.0014 -0.0404 0.0096


Individual

Age (years)15 -0.0174 0.0690 -0.0141

Education (various levels) -0.1187 0.5102 -0.1056

Gender (1=male, 2=female) -0.0993 0.4626 -0.0968

Location

Region pastoral zone 0.2572 -0.4652 0.0793

Region wheat-sheep zone 0.0953 -0.3358 0.0676

Region high-rainfall zone -0.2367 0.6496 -0.1247

(all: 1=yes, 0=no)

Household

Off-farm income share 0.2804 -0.0540 -0.0194

Age spouse (years) -0.0028 0.2139 -0.0497

Education spouse (various levels) -0.1132 0.4087 -0.0826

Gender spouse (1=male, 2=female) -0.1485 0.4541 -0.0891

Financial

Total assets (AUD) -0.3838 0.5646 -0.0881

Total subsidies (AUD) 0.0871 -0.0503 0.0018





1.5. Australian beef farms


the sample, the average beef output per farm in 2018 was AUD 1.7 million (with a total

output of about AUD 2.3 million). The variable cost items increased and the share of hired

labour significantly increased for the average Australian beef farm over the period

investigated. In 2018, the average herd size was about 73 400 LU with a stocking density

of about 0.59 LU per ha (compared with about 61 500 LU and 0.61 LU per ha in 1989).


components for the Australian beef production the output variable is total output per farm

and year and input variables are land, capital, labour, livestock units and fodder. The Class

identification component is based on the indices related to structure, environmental



51. For Australian beef farms, three distinct technology classes emerge from the model

estimates (Table A D.1.22). The individual farms are distributed unevenly across the three

technology classes (Table 1.9), as Class 1 includes more than two-thirds of all beef farms.

52. Beef farms in Class 2 show a significantly higher productivity performance


in value per year — than their counterparts in classes 1 and 3 (Figure 1.9). Beef farms in

Class 2 produce about 3 times more per farm and year than farms in Class 1 and ten times

more per farm and year than beef farms in Class 3.

53. Beef farms in Class 2 are likely to increase their productivity advantage, ceteris

paribus, as they show a significantly positive technical change (of 0.8% per year) while the



estimated technical change for farms in other classes is negative. The decline is particularly

strong for the least productive farms in Class 3 (Figure 1.9).

Figure 1.9. Productivity and technical change for Australian beef farms,


Source: Table 1.9.

54. If beef farms in Class 1 produced with the technology of farms in Class 2, they

would be able to increase their productivity by 176%, and farms in Class 3 would multiply

their productivity by five (Table 1.9). Overall, if all farms were producing with Class 2

technology, the productivity of Australian beef farms would almost double given that farms

in class 1 account for two-thirds of all beef farms in Australia (Table A C6 in Part 1).

Table 1.9. Productivity characteristics of Australian beef farms, by class











- Class 1 Technology -0.890*** 1.535*** -3.793***

- Class 2 Technology -1.446*** 0.790** -5.017***

- Class 3 Technology -1.086*** 0.397* -3.709***



-0.89

0.79

-3.71-4

-3.5

-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

0

50

100

150

200

250

300

350

400

450

500

class 1 (67.7% of farms) class 2 (21.9% of farms) class 3 (10.4% of farms)

%EUR '000

productivity level (EUR '000) technical change (% per year)




all beef farm classes exhibit nearly constant or slightly increasing returns to scale: 1.0423

for Class 1, 1.0076 for Class 2, and finally 1.0158 for Class 3 (Table A D.1.24).

56. Figure 1.10 summarise the various beef farm classes in terms of estimated indices,


57. Class 1 groups over two-thirds of Australian beef farms, which have average

environmental sustainability, and a below average productivity. They are larger than

average operations in terms of area and herd size. They are close to the average for most

indicators of technology and practices, but their manager is older and more educated than

average and more likely to be a woman. Their asset endowment is below average.

58. Class 2 farms are the most productive and the least environmentally sustainable, as

they have more intensive farming practices. They are smaller, more diversified operations,

but have assets well above average. They investment slightly more in new technologies and

activities than average, and higher debt ratios than average. Their manager is younger and

less educated than average.

59. Class 3 farms are the most environmentally sustainable and the least productive.

They are the largest, most diverse operations. They use the most extensive farming

practices. They are less capital intensive than average and are less likely to use contract

farming. They generate a higher share of off-farm income, and lower debt ratios, but

smaller asset endowment.

60. In summary, Australian beef farms in the three identified technology classes differ


over time. Beef farming in Australia is characterised by a strong positive link between farm

structure (i.e. larger herd size and land endowment as well as less family labour oriented)

and productivity. Least productive farms in this sector operate with a lower than average

technology intensity and financial viability. Most productive beef farms are, however, less

environmentally sustainable than the average beef farm in Australia. These farms are very

innovative but do not necessarily operate with a higher capital and input intensity compared

to medium productive beef farms.



Figure 1.10. Multi-dimensional indices for Australian beef farms



Table 1.10. Multiple characteristics of Australian beef farms, by class





Farm structure

Family/hired labour ratio 0.0067 -0.0222 0.0032

Herd size (LU) 0.4576 -1.1469 0.6893

Land (ha) 0.2352 -0.8602 0.2796


0.1696 -0.5897 0.1368



Chemicals use (AUD per ha) -0.0399 0.1858 -0.1320

Fuel per ha (AUD per ha) -0.0502 0.2051 -0.1051



Contract farming (1=yes, 0=no) 0.0084 0.0896 -0.2445

Share land rented -0.0160 0.0414 0.0169

Genetics veterinary expenses per LU (AUD per LU)

-0.0210 0.0802 -0.0324

Insurance expenses per LU (AUD per LU) -0.0368 0.1242 -0.0216

Technology



0.0043 -0.0123 -0.0023

Share crop land 0.0017 -0.0310 0.0544

-2

-1.5

-1

-0.5

0

0.5

1






Index 7 - Location

Index 8 - Household

Index 9 - Financial






LU per labour (LU per AWU) -0.0127 0.0908 -0.1090

Capital per LU (AUD per LU) -0.0104 -0.0042 0.0766

Herdtest (1=yes, 0=no) -0.0412 0.1648 -0.0788

Diversity



Individual

Age (years) 0.1545 -0.5061 0.0591

Education (various levels) 0.2421 -0.7749 0.0545

Gender (1=male, 2=female) 0.2226 -0.6978 0.0191

Location

Region pastoral zone -0.1569 0.4999 -0.0305

Region wheat-sheep zone 0.0674 -0.2053 -0.0067

Region high-rainfall zone 0.0942 -0.3166 0.0530

(all: 1=yes, 0=no)

Household

Off-farm income share -0.0242 -0.0830 0.3334

Age spouse (years) 0.1928 -0.6027 0.0131


Gender spouse (1=male, 2=female) 0.1818 -0.5204 -0.0893

Financial

Total assets (AUD) -0.1504 0.6294 -0.3460

Total subsidies (AUD) -0.0128 0.0308 0.0188

Equity/debt ratio -0.0053 -0.0830 0.3334




1.6. Australian sheep meat farms


the sample, the average sheep meat output per farm was about AUD 787 500 in 2018 (with

a total output of about AUD 850 000). The variable cost items increased over time and the

share of hired labour significantly increases for the average Australian meat sheep farm.

The average herd size was about 6 200 LU in 2018 and the average land acreage about

12 600 ha which results in a stocking density of about 3.5 LU per ha (compared with about

10 100 LU and about 25 250 ha in 1989).


components for the Australian sheep meat production the output variable is total output per

farm and year and input variables are land, capital, labour, livestock units, and materials.

The Class identification component is based on the indices related to structure,



63. For Australian sheep meat farms, three technology classes with relatively close

productivity performance emerge from the model estimates (Table A D.1.27). The

individual farms are distributed very unevenly across the three technology classes

(Table 1.11), as Class 1 includes nearly three-quarters of all farms.

64. Sheep meat farms in Class 1 show a significantly higher productivity performance


in value per year — than their counterparts in classes 2 and 3 (Figure 1.11). Farms in



classes 2 and 3 produce respectively 81% and 88% of Class 1 output per farm and year,

reflecting very close technologies.

65. All sheep meat farm classes show a significantly positive technical change rate, but

the rate for farms in Class 1 is lower than the technical change rate for farms in other

classes, which shows very high increases of over 3% per year and farm. This suggest that

productivity levels are converging between farms in Class 1 and other farms, ceteris

paribus. The results also indicate an increasing gap between Class 2 and Class 3, as farms

in the latter have both higher productivity levels and higher technical change rate than farms

in the former (Figure 1.11).

66. If sheep meat farms in classes 2 and 3 produced with the technology of farms in

Class 1, their productivity achievements would decrease by respectively -20% and -11%

(Table 1.11). Overall, this would result in a 2% decline of productivity for sheep meat

farms in Australia (Table A C.6 in Part 1). These productivity losses might indicate that

those farms have already adopted a productive technology given their

locational/environmental constraints, which cannot be changed or optimised by the farmer.

Figure 1.11. Productivity and technical change for Australian sheep meat farms,


Source: Table 1.11.


all sheep meat farm classes exhibit decreasing returns to scale: 0.8233 for farms in Class 1,

0.8944 for farms in Class 2, and 0.6081 for farms in Class 3 (Table A D.1.29).

68. Figure 1.12 summarise the various sheep meat farm classes in terms of estimated

indices, while Table 1.12 contains the estimates for variables used to construct the indices.

1.44

3.19

3.85

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

0

20

40

60

80

100

120


%EUR '000




Table 1.11. Productivity characteristics of Australian sheep meat farms, by class
















69. Class 1 includes close to three-quarters of all farms, which are the most productive

and also the most environmentally sustainable of all sheep meat farms in Australia. They

are more diverse operations, with the largest herds, but only above average in terms of land.

As a result, they have the highest stocking density, but otherwise use less chemicals and

fuel per ha. Their capital intensity and investment in new technologies is below average.

70. Class 2 farms are the least productive and have a lower than average environmental

sustainability score, due to lower stocking density. They are smaller in terms of herd size

and land, and most specialised operations, with high capital intensity. They are managed

by older and more educated farmers, more likely to be women. They have the highest share

of off-farm income, receive the highest amount of subsidies and have lower debt ratios.

71. Class 3 groups the least environmentally sustainable farms, which have a

productivity slightly higher the lowest one. They are the largest and most innovative farms

and are more capital intensive than average. They are less diversified than average, their

manager is younger, but better educated than average. They receive the lowest amount of

subsidies and obtain the lowest share of off-farm income.

72. In summary, Australian sheep meat farms in the three identified technology classes

differ slightly with respect to their economic performance, but more significantly with

respect to their technical development over time. Productive sheep meat farms in Australia

are also found more environmentally sustainable than other farms in the sector. They use

less family labour but produce with an average farm size. The least productive farms in the

sector operate with a lower than average environmental sustainability. These farms are,

however, very capital intensive and least diverse. Generally, productivity differences

between farms in the sector are relatively low, ceteris paribus.



Figure 1.12. Multi-dimensional indices for Australian sheep meat farms



Table 1.12. Multiple characteristics of Australian sheep meat farms, by class





Farm structure


Herd size (LU) 0.6664 -1.1499 0.4834

Land (ha) 0.0328 -0.5067 0.1607


0.0003 0.0636 -0.0403



Chemicals use (AUD per ha) -0.1822 0.1837 0.7241

Fuel per ha (AUD per ha) -0.1575 0.0610 0.6860

Innovation- commercialisation

Net investment ratio (per total assets) -0.0163 0.0121 0.0676

Contract farming (1=yes, 0=no) -0.0257 0.0716 0.0741

Share land rented 0.0049 -0.0959 0.0363


0.0208 -0.0920 -0.0392

Insurance expenses per LU (AUD per LU) -0.0424 0.2474 0.0427

Technology

Capital / labour ratio (AUD per AWU) -0.1324 0.6356 0.2175


-0.0229 0.1118 0.0365

Share crop land -0.1999 0.7728 0.4433

LU per labour (LU per AWU) 0.0640 0.5067 -0.1102

Capital per LU (AUD per LU) -0.0504 0.3025 0.0457

-2

-1.5

-1

-0.5

0

0.5

1






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Herdtest (1=yes, 0=no) -0.0065 -0.1021 0.0926

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) -0.3631 1.1963 0.9329

Production diversity (yc/ΣY) -0.0608 0.2005 0.1563

Individual

Age (years) 0.0035 0.2030 -0.1408

Education (various levels) -0.0694 0.2536 0.1630


Location

Region pastoral zone 0.0525 0.0984 -0.3017

Region wheat-sheep zone -0.0551 0.0716 0.2091

Region high-rainfall zone 0.0180 -0.1454 0.0066

(all: 1=yes, 0=no)

Household

Off-farm income share -0.0527 0.6707 -0.1702

Age spouse (years) -0.0165 0.4298 -0.1883

Education spouse (various levels) -0.0504 0.6482 -0.1666

Gender spouse (1=male, 2=female) -0.0408 0.5458 -0.1478

Financial

Total assets (AUD) -0.0570 0.1732 0.1554

Total subsidies (AUD) -0.0371 0.4980 -0.1354

Equity/debt ratio -0.0009 0.0229 -0.0099




1.7. Australian wool farms


the sample, the average sheep related output per farm was about AUD 816 000 in 2018.

The variable cost items increased over time and the share of hired labour significantly

increased for the average Australian sheep wool farm over the period investigated. The

average farm had a herd size of about 6 500 LU and operated a total land area of about

13 700 ha resulting in a stocking density of about 3.49 LU per ha (compared with 7 750 LU,

20 270 ha, and 3.86 LU per ha in 1989).


components for the Australian sheep wool production the output variable is total output per

farm and year and input variables are land, capital, labour, livestock units, and materials.

The Class identification component is based on the indices related to structure,



75. For Australian wool farms, three distinct technology classes emerge from the model

estimates (Table A D.32). The individual farms are distributed very unevenly across the

three technology classes (Table 1.13), as Class 1 includes 80% of all farms.

76. Wool farms in Class 1 show a significantly higher productivity performance


in value per year — than their counterparts in classes 2 and 3 (Figure 1.13). Wool farms in

classes 2 and 3 produce respectively 40% and 68% of the estimated output for farms in

Class 1.



77. The highest rate of technical change per farm (1% per year) is found for farms in

Class 3, which are catching up with the most productive farms in Class 1, ceteris paribus

(Figure 1.1). Conversely, the least productive farms in Class 2 are progressing more slowly,

with a rate of technical change of 0.13% per year.

Figure 1.13. Productivity and technical change for Australian wool farms,


Source: Table 1.13.

78. If wool farms in Class 2 produced with the technology of farms in Class 1, they

would experience a slight decline in productivity performance, according to estimates

(Table 1.13). On the contrary, Class 3 would show a 6% increase in productivity. Overall,

if all wool farms were producing with Class 1 technology, the productivity of Australian

would farms would increase by less than 1% (Table A C.6 in Part 1). As for sheep meat

farms, this suggests that farms in Class 2 and Class 3 have already adopted a productive

technology given their locational/environmental constraints.

79. The estimated production functions for each farm class are moderately significant

and the identified sheep wool farm classes exhibit constant or decreasing returns to scale

(Table A D.1.34).

80. Figure 1.14 summarise the various wool farm classes in terms of estimated indices,


81. Over 80% of all Australian wool farms are most productive and most

environmentally sustainable, as they use more extensive farming practices. They are

smaller and more diverse operations, with lower than average investment in new

technologies and the lowest capital intensity. Given their large number, they are close to

average scores, which they determine to a large extent.

0.56

0.13

0.99

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0

20

40

60

80

100

120


%EUR '000




Table 1.13. Productivity characteristics of Australian wool farms, by class











- Class 1 Technology 0.563*** -1.460** 1.265***

- Class 2 Technology 3.300*** 0.127* 1.439***

- Class 3 Technology 0.543*** -2.846*** 0.988***

Note: * significant at 10%, ** significant at 5%, *** significant at 1%. 1. Fitted values at sample means. Source:

Own estimations based on latent class model estimates and derivatives.

82. This groups the 3% least productive farms, with a below average environmental

sustainability driven by their high stocking density. They are the largest, most diverse

operations, with older and more educated than average managers, more likely to be women.

They have the highest share of off-farm income, receive less subsidies and have lower debt

ratios.

83. Least environmentally sustainable farms achieve medium productivity (16% of all

farms). They are larger than average, more specialised operations, with most intensive

farming practices. They have the highest levels of investment in new technologies and

practices and the highest capital intensity. They are managed by younger farmers and their

households are less dependent on off-farm income. They have the largest asset endowment,

receive the highest subsidies and their debt/equity ratio is slightly below average.

84. In summary, Australian wool farms in the three identified technology classes differ


over time. The most productive sheep farms are again more environmentally sustainable.

Contrary to the meat producing sheep sector, however, more productive wool producing

sheep farms are based on an average family labour endowment that operate significantly

larger than average herds. A small group of least productive farms produce with a

significantly lower than average environmental sustainability using a significantly higher

than average amount of hired labour linked to a relatively high technology intensity.



Figure 1.14. Multi-dimensional indices for Australian wool farms



Table 1.14. Multiple characteristics of Australian wool farms, by class





Farm structure

Family/hired labour ratio -0.0139 -0.3360 0.0688

Herd size (LU) -0.0446 0.5101 -0.0040

Land (ha) -0.0338 0.2742 0.1121


0.0138 0.1283 0.0426


Stocking density (LU per ha) -0.0410 0.8139 0.0420

Chemicals use (AUD per ha) -0.1537 -0.0057 0.7541

Fuel per ha (AUD per ha) -0.1794 1.0557 0.6728

Innovation / Commercialisation

Net investment ratio (per total assets) -0.0384 0.0194 0.1842

Contract farming (1=yes, 0=no) -0.0632 0.0973 0.2904



-0.0336 0.6439 0.0388

Insurance expenses per LU (AUD per LU) -0.0919 0.6345 0.3263

Technology

Capital / labour ratio (AUD per AWU) -0.1487 0.0778 0.7134


-0.0048 -0.1645 0.0557

Share crop land -0.2282 0.1596 1.0868

LU per labour (LU per AWU) 0.0925 -0.6839 -0.3196

Capital per lu (AUS$ per lu) -0.1192 0.1756 0.5498

-2.5

-2

-1.5

-1

-0.5

0

0.5

1






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Herdtest (1=yes, 0=no) -0.0117 0.1503 0.0281

Diversity


Production diversity (yc/ΣY) -0.0397 -0.0660 . 207519

Individual

Age (years) 0.0080 0.2142 -0.0809

Education (various levels) -0.0310 0.5227 0.0499


Location

Region pastoral zone 0.0837 -0.4235 -0.3273

Region wheat-sheep zone -0.0940 0.2414 0.4135

Region high-rainfall zone 0.0286 0.0771 -0.1553

(all: 1=yes, 0=no)

Household

Off-farm income share 0.0156 0.3336 -0.1417

Age spouse (years) -0.0129 0.2808 0.0082

Education spouse (various levels) -0.0277 0.2720 0.0825

Gender spouse (1=male, 2=female) -0.0403 0.5170 0.0967

Financial

Total assets (AUD) -0.0742 -0.0568 0.3744

Total subsidies (AUD) -0.0222 -0.1607 0.1400

Equity/debt ratio -0.0081 0.1305 0.0141




2. Chile

2.1. Chilean small-scale fruit farms

85. For Chile, the analysis applies to a sample of Chilean small-scale fruit farms for

2015. Fruit production in Chile has steadily increased in the last decade covering a total

area of about 318 000 ha. Leading fruit crops are still fresh table grapes, walnuts, apples,

cherries and avocados, whereas the share of hazelnuts, blueberries, cherries and mandarins

are steadily growing. Exports continue to account for more than 60% of Chilean fruit

production for most fruit crops focusing distinctive consumer preferences and markets.

86. According to Table A D.2.1, which shows different statistical measures for the

analysed sample, the average fruit output per farm was about EUR 8 400 per year in 2015

with a land endowment of about 2.9 ha on average. Family labour is very important for

fruit production in Chile accounting for more than twice as much as hired labour. The

average fruit farmer in Chile was of 58 years of age in 2015.


components for Chilean fruit production, the output variable is total output per farm, and

the input variables are land, capital, chemicals, and labour. The Class identification

component is based on the indices related to structure, environmental sustainability,

innovation, technological intensity, diversity, individual characteristics, location and

financial characteristics (Table A D.2.2).

88. For small-scale Chilean fruit farms, three distinct technology classes emerge from

the model estimates (Table A D.2.2). The individual farms are distributed unevenly across



the three technology classes (Table 2.1). Class 1 covers about 45% of all fruit farms,

Class 2 about 36% and Class 3, which is based on only 86 observations, about 19%.

89. Farms in Class 1 show a significantly higher productivity performance — measured

as the potential output levels that could be achieved with a given input bundle, in EUR per

year — than their fruit-producing counterparts in classes 2 and 3 (Figure 2.1). They

produce much more per farm and year than farms in Class 3, and even almost 15 times

more than fruit farms in Class 2.

Figure 2.1. Productivity for Chilean small-scale fruit farms, by class, 2015

Source: Table 2.1

90. If fruit farms in Class 2 produced with the technology of farms in Class 1, they

would be close to tripling their productivity. Hence, a large part of small-scale fruit farms

in Chile was highly productive in 2015. However, more than 50% of all fruit farms could

significantly increase their productivity level by switching to a more productive technology

(Class 1 technology), ceteris paribus (Table 2.1). Overall, if all farms adopted the

technology of the most productive farms in Class 1, the productivity of small-scale fruit

farms in Chile would increase on average by about 45% (Table A C.5 in Part 1).

0

1 000

2 000

3 000

4 000

5 000

6 000

7 000

8 000

9 000


EUR



Table 2.1. Productivity characteristics of Chilean small-scale fruit farms, by class

Latent Class Estimation, cross-section 2015


Number of observations (% of sample farms) 200 (44.6%) 162 (36.2%) 86 (19.2%)




- Class 1 Technology 8 379.8*** 8 379.8*** 8 379.8***





91. The estimated production functions for each farm class are significant at a

satisfactory level given the limited sample size. All farm classes exhibit considerable

increasing returns to scale: 1.1599 for Class 1, 1.1091 for Class 2, and 1.5597 for farms in

Class 3 (see Table A D.2.4 for complete class related elasticities). Switching technologies

could — ceteris paribus — result in higher productivity per farm and in a significantly

higher technical change rate for the majority of farms (especially regarding a switch to

Class 1 technology). Various indices reflecting the different dimensions along which fruit

farms can be distinguished are used to robustly identify the reported classes: Farm structure,

environmental sustainability of operations, technology characteristics, and individual

characteristics, locational and financial specifics. Table A D.2.4 summarises the estimates

for the various indices that were used as components for the class-identifying vector in the

latent class estimation. Many of the indices considered showed significant estimates. To

decide empirically on the most appropriate number of classes to estimate, various statistical

quality tests have been performed (most prominently the Akaike Information Criteria AIC)

which holds for all country cases considered in this study.

92. Figure 2.2 summarises the various small-scale fruit farm classes in terms of


indices.

93. As discussed earlier, fruit farms in Class 1, which are the most productive, have the

lowest share of family labour in total labour (i.e. a significantly lower family per hired

labour ratio than the average fruit farm in the sample) and a significantly above average

acreage size, but they still experience considerable increasing returns to scale. Fruit

producing farms in Class 1 score also very high on environmental sustainability indicators

(such as organic certification probability, chemicals use per ha, soil management and the

use of water treatment technology). These productive and environmentally sustainable

farms show also the highest scores on innovation and commercialisation with a

significantly above average share of rented and irrigated land and the highest probability

of investing in various innovative production and management tools as for example,

fertigation (fertilisation through irrigation water), genetic improvement, or disease control

techniques. Furthermore, these fruit farms comprehensively use technical and accounting

services including from private sources. Class 1 farms show a higher than average capital

per labour intensity, while using more capital per land than the average fruit farm in Chile.

Managers are slightly younger and better educated than the average fruit farmer. Finally,



farms in Class 1 are most likely located in proximity to urban centres and they receive some

public support (Table 2.2).

94. Fruit farms in Class 3 exhibit a significantly lower than average productivity level.

Family labour is of low importance for those farms, which are larger than the average fruit

farm in Chile in terms of land endowment. Given the measured returns to scale

(Table A D.1.4), fruit farms in Class 3 should significantly increase the size of their

production operations to take advantage of economies of scale. However, the

environmental sustainability of these farms is found to be below average based on the

various indicators used. Farms in Class 3 show also a lower than average score on

innovation and commercialisation with a below average share of rented and irrigated land

and a rather low probability of investing in various innovative production and management

tools. Furthermore these fruit farms use technical and accounting services including from

private sources with a lower than average probability compared to the average fruit farm in

Chile in 2015. Their capital intensity is well below average and they employ the least

capital per land of all farm classes. Finally, fruit farms in Class 3 receive low levels of

public support (Table 2.3).

95. Class 2 farms are the least productive in the cross-sectional sample considered.

Family labour is most important for those fruit farms, which are considerably smaller than

the average fruit farm in Chile in terms of land endowment. Fruit farms in Class 2 should

increase the size of their production operations to become more profitable given the

measured positive economies of scale. These farms are found to be among the least

environmentally sustainable fruit producers based on various indicators used (here

especially the lowest probability of using water treatment technology and organic

certification). Farms in this class also have the lowest scores on innovation and

commercialisation with a significantly below average share of rented land and very low

probabilities for investing in various innovative production and management tools as, for

example fertigation, genetic or disease control techniques. Furthermore, these fruit farms

mostly do not use technical and accounting services nor private advisory services. They

have an average capital intensity with a slightly above average capital usage per land unit.

These fruit farms are located further away from urban centres (Table 2.3).

96. In summary, Chilean small-scale fruit farms in the three identified technology

classes differ significantly with respect to their economic performance. Family farms and

comparatively smaller farms show a lower environmental sustainability based on the

measures used in the empirical analysis. Highly environmentally sustainable fruit farms

also have a high probability to be very innovative in terms of investing in state-of-the-art

productive and environmentally sustainable technologies. These farms are located close to

urban centres with a higher than average capital intensity. Farms’ capital intensity is not

necessarily correlated with land endowment, while farms’ productivity shows to be

correlated with acreage size and a higher share of hired labour in total farm labour. Finally,

with more dynamic structural change in the sector, there would be a large scope for

improving productivity and environmental sustainability by having more fruit farms

switching to a more productive and environmentally sustainable fruit production

technology.



Figure 2.2. Multi-dimensional indices for Chilean small-scale fruit farms

Scaled values at class means, 2015


Table 2.2. Multiple characteristics of Chilean small-scale fruit farms, by class

Deviations from sample means1, 2015




Farm structure


Land (ha) 0.1062 -0.2754 0.2717


Organic (probability) 0.1119 -0.0878 -0.0949

Chemicals use (EUR per ha) 0.0581 0.0091 -0.1522

Soil management ratio -0.0397 0.1227 -0.1387

Water treatment technology (prob) 0.1186 -0.1069 -0.0744



Irrigation ratio 0.1983 0.0213 -0.5014

Internet use (prob) 0.1812 -0.1352 -0.1668

Computer use (prob) 0.2169 -0.1778 -0.1696

Fertigation technology (prob) 0.3138 -0.2072 -0.3394

Genetic technology (prob) 0.3321 -0.2737 -0.2565

Input technology (prob) 0.2229 -0.2397 -0.0669

Disease technology (prob) 0.2909 -0.2932 -0.1244

Fair trade (prob) 0.1158 -0.1258 -0.0322

Technical advisory (prob) 0.3309 -0.3125 -0.1811

-0.6

-0.4

-0.2

0

0.2

0.4

0.6





Index 6 - Individual

Index 7 - Location

Index 8 - Household

Index 9 - Financial






Accounting advice (prob) 0.4642 -0.4681 -0.1979

Private advisory (prob) 0.2516 -0.2156 -0.1790

Technology

Capital / labour ratio (EUR per AWU) 0.1295 -0.0901 -0.1313

Capital per land (EUR per ha) 0.0588 0.0095 -0.1548

Individual

Age (years) -0.0762 -0.1491 0.4579

Gender (1-male, 0-female) 0.2052 -0.3724 0.2243

Education (years) 0.0903 0.1114 -0.4198

Experience (years) 0.0682 0.0891 -0.3265

Civil association membership (1-yes, 0-no) -0.0297 -0.0061 0.0804

Location

Distance to urban centre (h) -0.1069 0.1385 -0.0123

Household

Household size (n) 0.0472 -0.2237 0.3117

Ethnic (1-yes, 0-no) -0.1267 0.1771 -0.0391

Financial

Public support (1-yes, 0-no) 0.0713 -0.0212 -0.1258



3. Czech Republic

3.1. Czech dairy farms

97. For the Czech Republic, the analysis applies to a sample of dairy farms covering

the period 2005 to 2015. The Czech dairy sector is highly consolidated having undergone

a severe transformation process that followed the former planned economy approach by the

Socialist regime. The number of dairy farms has declined by about 60% in the last 20 years

with about 1 100 milk producing farms in 2016 delivering nearly 3 000 million litres milk

per year based on more than 373 000 dairy cows (EDF, 2017; Czech Statistical Office,

2017).

98. The total livestock production in the Czech Republic increased over the last

10 years to about 6 836 LUs in total whereas about 72% relate to cattle production. The

density of cattle related livestock production also increased over this period to about

40 units of livestock per 100 ha (Czech Statistical Office, 2017). The average milk yield

per cow significantly increased from about 6 900 litres per cow and year to more than

8 000 litres per cow and year. The average herd size was about 314 cows per farm and the

total milk output in current prices has been reported as about CZK 21 250 million for 2016.

99. The economic situation of milk production in the Czech Republic has been

characterised by low output prices but relatively low production costs compared to other

EU member countries (European Dairy Farmers - EDF, 2017). Specialised dairy farms

account for 9.7% of the total agricultural production and 28.3% for the total milk

production in the Czech Republic. According to the results of the current Farm Structure

Survey (FSS 2016), this type of farming cultivates 8.4% of the total area of agricultural

land in the Czech Republic and accounts for 6.7% of the total number of agricultural

holdings. Nevertheless, the largest volume of milk production from all production areas

monitored is produced by mixed-production type holdings (66.7%) (FADN CZ, 2017).




sample, the average milk output per farm increased over the period 2005 to 2015 due to a

significant increase in herd size. The variable cost items increased over time and the share

of hired labour significantly increases for the average Czech dairy farm. The stocking

density remained more or less at the same level, as did the probability of being engaged in

organic dairy production.


components for the Czech dairy production the output variable is total output per farm and

year and input variables are cows, capital, materials, fodder, fuel and labour. The Class

identification component is based on the indices related to structure, environmental

sustainability, innovation, diversity and location (Table A D.3.2). The individual farms are

distributed very evenly across the three technology classes (Table 3.1). Class 1 covers

about 34% of all dairy farms, Class 2 about 32.5%, and Class 3 about 33.5%.



in EUR per year — than their milk-producing counterparts in Classes 2 and 3 (Figure 3.1).

Dairy farms in Class 1 produce about 6 times more per farm and year than farms in Class 2

and about 12 times more per farm and year than dairy farms in Class 3.

103. Dairy farms in Class 1 show a significantly positive technical change (of 2.06% per

year) compared to a modest positive technical change rate for farms in Class 3 (estimated

at 0.97% per year) and in Class 2 (estimated at 0.37% per year). Hence, although dairy

farms in Class 1 are already far more productive, they also seem to increase their

productivity at a higher rate based on a significant positive technical change development

ceteris paribus.

Figure 3.1. Productivity and technical change for Czech dairy farms, by class, 2005 to 2015

Source: Table 3.1.


would be able to increase their productivity significantly by about 167%, and farms in

Class 3 would increase their productivity 171%. (Table 3.1). If all Czech dairy farms

adopted the technology of most productive farms in Class 1, their average productivity

would increase by 32% (Table A C.4 in Part 1).

2.06

0.37

0.97

0.0

0.5

1.0

1.5

2.0

2.5

0

100

200

300

400

500

600

700

800


%EUR '000

Productivity (EUR '000 per year) Technical change (% per year)



Table 3.1. Productivity characteristics of Czech dairy farms, by class












- Class 2 Technology 1.944*** 0.366*** -2.836***





all dairy farm classes exhibit increasing returns to scale: 1.1211 for Class 1, 1.2020 for

Class 2, and 1.0857 for farms in Class 3 (see Table A D.3.4, for complete class related

elasticities). Switching technologies could — ceteris paribus — result in higher

productivity per farm and also a significantly higher technical change rate for most farms

(especially regarding a switch to Class 1 technology). Various indices reflecting the

different dimensions along which dairy farms can be distinguished are used to robustly

identify the reported farm classes: Farm structure, environmental sustainability of

operations, technology characteristics, degree of diversity, individual characteristics and

locational specifics. Table A D.3.2 summarises the estimates for the various indices that

were used as components for the Class identifying vector in the latent class estimation.

Most of the indices considered showed significant estimates. To decide empirically on the

most appropriate number of classes to estimate, various statistical quality tests have been





107. As discussed earlier, dairy farms in Class 1, which are the most productive and

exhibit a most significant positive technical change rate per year, have the lowest share of

family labour in total labour (i.e. a significantly lower family per hired labour ratio than the

average farm in the sample). They are significantly larger than average, by herd and acreage

size, but they still experience (modest) increasing returns to scale. Dairy farms in Class 1

score relatively high on environmental sustainability indicators (such as stocking density,

chemicals use per ha and probability of producing organic). These farms show the highest

scores on innovation and commercialisation with a significantly higher than average

investment rate, rented land share, and income generated by biofuels production. Class 1

farms show a lower than average capital per labour intensity, while using more capital per

cow than the average dairy farm in the Czech Republic based on high levels of total assets

endowment. These dairy farms are less diversified than the average dairy farmer in the



Czech Republic, and their managers are older than average. Finally, dairy farms in Class 1

are most likely located in less-favoured areas (Table 3.2).

108. Dairy farms in Class 2 show a medium productivity, but a lower than average

technical change per year. They are smaller than the average dairy farm in the Czech

Republic in terms of herd size and land endowment. Hired labour is important and a single

ownership is likely for these farms. Dairy farms in Class 2 should significantly increase the

size of their production operations given the measured economies of scale. They are found

to be the most environmentally sustainable based on various indicators used (such as

stocking density, chemicals use per ha and probability of producing organic). However,

dairy farms in Class 2 invest less than the average dairy farm in the Czech Republic and

generate less income by biofuel production. Their capital intensity is the lowest of all dairy

farms, and they employ the most labour per cow. These dairy farms are the least specialised

of the three classes and their farm managers are of average age. Finally, dairy farms in

Class 2 are likely located in less-favoured areas or areas of higher altitude (Table 3.2).

Figure 3.2. Multi-dimensional indices for Czech dairy farms



109. Class 3 farms are the least productive but show a significant average technical

change rate per year. Family labour is most important for those dairy farms, which are

considerably smaller than the average dairy farm in the Czech Republic in terms of herd

size and land endowment. Single ownership is most probable for these farms. Dairy farms

in Class 3 should increase the size of their production operations to become more profitable

given the identified positive economies of scale. These farms are found to be the least

environmentally sustainable based on various indicators used (such as stocking density,

chemicals use per ha and probability of producing organic). Dairy farms in Class 3 invest

far less than the average dairy farm in the Czech Republic and generate least income from

biofuel production. Their capital intensity is, however, the highest of all dairy farms and

consistent with this notion is the finding that farms in Class 3 employ the lowest quantity

of labour per cow. These farms are least diversified and their farm managers are of lower

than average age. Finally, dairy farms in Class 3 are less likely located in less-favoured

areas or areas of higher altitude (Table 3.3).

-2

-1.5

-1

-0.5

0

0.5

1



Index 3 - Innovation/Cooperation/

Commercialisation


Index 5 - Diversity

Index 7 - Location






110. In summary, Czech dairy farms in the three identified technology classes differ


over time. Innovative dairy farms in the Czech Republic are most likely more productive

than average. Family driven farms and smaller farms do not appear necessarily more

environmentally sustainable based on the measures used in the empirical analysis. Highly

environmentally sustainable dairy farms most likely show a very diverse farm structure and

are most likely located in less-favoured areas with a lower than average capital intensity.

Farms’ capital intensity is positively correlated with herd size, while farms’ productivity is

correlated with herd size and the share of hired labour. Finally, Czech dairy farms

producing with a more environmentally sustainable performance can also exhibit a very

high productivity, ceteris paribus.

Table 3.2. Multiple characteristics of Czech dairy farms, by class





Farm structure


Herd size (LU) 0.9596 -0.2041 -0.7711

Land (ha) 0.8898 -0.0860 -0.8147

Form of ownership (1-self-employment, 2-legal person, 3-cooperative)

0.8975 0.0022 -0.9075


Stocking density (LU per ha) -0.1833 -0.3750 0.5466

Chemicals use (EUR per ha) 0.6963 -0.2225 -0.4878

Organic (probability) -0.2018 0.2034 0.0074

Environmental subsidies (EUR per ha) 0.2933 0.2883 -0.5740




Biofuel Income (EUR) 0.3300 -0.1452 -0.1928

Technology

Capital / labour ratio (EUR per AWU) -0.0768 -0.1241 0.1972

Capital per cow (EUR per LU) 0.0282 -0.1361 0.1029

Labour per cow (AWU per LU) 0.0921 0.2656 -0.3492

Total assets (EUR) 0.8588 -0.2256 -0.6487

Diversity



Individual

Age (years) 0.2623 0.0023 -0.2669

Location

Less Favoured Area payments (EUR) 0.4644 0.2244 -0.6849

Altitude (1- <300m, 2- 300-600m, 3- >600m) -0.0822 0.3633 -0.2676

Note: LU: Livestock Unit. AWU: Annual Work Unit. 1. Deviations from sample means (=0), z-scores based,

scaled values.




4. Denmark

111. For Denmark, the analysis applies to a sample of dairy farms covering the period

2010 to 2016 and to a sample of pig farms covering the period 2006 to 2016. In Denmark,

field crops dominate agricultural land use, but are mostly used for animal feed. Livestock

related production mainly relates to pigs, cattle, chicken and mink. The total food

production in Denmark largely exceeds the food demand of its own population, hence, most

products are exported. The average farm size is about 70 ha, more than 20% of the farms

cultivate more than 100 ha land (Danish Agriculture & Food Council, 2016; Vidø and

Schou, 2017).

4.1. Dairy farms

112. Dairy production in Denmark is characterised by relatively large-scale production

units with an average herd size of about 172 cows per farm and high milk yields per cow

of about 9 500 kg/head in 2016. The total number of dairy farms has been steadily

decreasing to about 3 300 dairy farms in 2016 and an average of about 1 600 tonnes of milk

delivered per farm. The number of farms delivering more than 5 000 tonnes per farm and

year has been significantly increasing in the last ten years (from 6 farms in 2006 to

106 farms in 2016) (Landbrug & Foedevarer, 2017).

113. According to Table A D.4.1, which summarises descriptive statistical measures for

the sample of Danish dairy farms, the average milk output per farm significantly increased

between 2010 and 2016 due to a significant increase in herd size. The variable cost items

also increased over the period, as well as the share of hired labour in the average farm

labour, which outperformed the share of family labour in 2016. The average stocking

density per farm increased slightly over the period investigated, while the probability of

being engaged in organic dairy production slightly declined.


components for the Danish dairy, the output variable is total output per farm and year. The

input variables are cows, land, capital, materials. The Class identification component is

based on the indices related to structure, environmental sustainability, innovation,

technology, diversity, individual and household characteristics, location and financial

aspects (Table A D.4.2).

115. For Danish dairy farms, three distinct technology classes emerge from the model


technology classes (Table 4.1). Class 1 covers about 67% of all dairy farms, Class 2 about

16% and Class 3 about 17%.

116. Dairy farms in Class 1 show a higher productivity performance — measured as the

potential output levels that could be achieved with a given input bundle, in EUR per year —

than their milk-producing counterparts in Classes 2 and 3 (Figure 4.1). They produce about

10% more per farm and year than farms in Class 2 and about 90% more per farm and year

than dairy farms in Class 3 (Figure 4.1).

117. Dairy farms in Class 1 show a positive technical change (of 1.78% per year)

compared to a significant positive technical change rate for farms in Class 2 (estimated at

2.97% per year) and in Class 3 (estimated at 2.02% per year). Hence, although dairy farms

in Class 1 are more productive, productivity seems to grow at a higher rate in slightly less

productive farms in Class 2 and significantly less productive farms in Class 3, due to

significant increases in technical change ceteris paribus (Table 4.1). Switching to Class 2



technology could also lead to a higher positive technical change rate for Class 1 farms (of

up to 3.9% per year compared to the current 1.78% estimated rate).

Figure 4.1. Productivity and technical change for Danish dairy farms, by class, 2010 to 2016

Source: Table 4.1.


would be able to increase their productivity by about 25%. Moreover, the most productive

farms in Class 1 could even further increase their productivity by switching to Class 2

technology (by 6%) implying that Class 1 and Class 2 farms produce with a quite similar

technology (Table 4.1).

Table 4.1. Productivity characteristics of Danish dairy farms, by class



Number of observations (% of sample farms) 11 461 (66.9%) 2 736 (15.9%) 2 924 (17.2%)













1.78

2.97

2.02

0

0.5

1

1.5

2

2.5

3

3.5

0

100

200

300

400

500

600

700

800


% '000 EUR

Productivity ('000 EUR per year) Technical change (% per year)



119. The estimated production functions are highly significant and all dairy farm classes

exhibit (slight to significant) increasing returns to scale: 1.0291 for Class 1, 1.0838 for

Class 2, and 1.1008 for farms in Class 3 (see Table A.B.4.4 for complete class related


productivity per dairy farm and in a significantly higher technical change rate for most

farms (especially regarding a switch to Class 2 technology). Various indices, reflecting the

different dimensions along which dairy farms can be distinguished, are used to identify

robustly the reported farm classes: Farm structure, environmental sustainability of

operations, innovation and technology related characteristics, degree of diversity,

individual and household characteristics as well as locational specifics. Table A.B.4.4

summarises the estimates for the various indices used as components for the class-

identifying vector in the latent class estimation. Most of the indices considered showed

significant estimates. To decide on the empirically most appropriate number of classes to

be estimated, various statistical quality tests have been performed (most prominently the

Akaike Information Criteria AIC) which holds for all country cases considered in this

study.

120. Figure 4.2 summarise the various dairy farm classes in terms of estimated indices,


121. As discussed earlier, dairy farms in Class 1 are slightly more productive and exhibit

a positive technical change rate per year. They have the lowest share of family labour (i.e.

a significantly lower family per hired labour share than the average farm in the sample) and

a significantly above average herd and acreage size but still experience (modest) increasing

returns to scale. Dairy farms in Class 1 score relatively low on environmental sustainability

indicators (such as stocking density, chemicals use per ha and probability of producing

organic). On the other hand, these farms show a slightly higher than average score on

innovation and commercialisation with a higher than average rented land share and higher

than average probability of being engaged in contracting. Class 1 farms show a slightly

lower than average capital per labour intensity and capital per cow intensity than the

average dairy farm in Denmark linked to an average level of assets endowment. They are

less diversified than the average dairy farm, and their managers are slightly younger than

the average Danish dairy farmer. Finally, dairy farms in Class 1 generate lower than

average off-farm income compared to the average Danish dairy farm (Table 4.2).

122. Dairy farms in Class 2 are slightly less productive than their counterparts in Class

1 but show a higher than average technical change per year. Hired labour is more important

for those farms, which are smaller than the average dairy farm in Denmark in terms of herd

size but larger in terms of land endowment. Dairy farms in Class 2 should significantly

increase the size of their production operations given the measured economies of scale (of

about 1.0838). These farms are found to be the most environmentally sustainable based on

the various indicators used (such as stocking density, chemicals use per ha and probability

of producing organic). Furthermore, dairy farms in Class 2 invest significantly more than

the average dairy farm in Denmark and their capital intensity is the highest of all dairy

farms, employing most capital and fodder per cow. Most likely, these dairy farms operate

with innovative milking technologies, e.g. automatic milk systems or milking parlours.

These dairy farms are diversified and their farm managers are of average age. Finally, dairy

farms in Class 2 generate higher than the average off-farm income per farm and operate

with a significantly higher than average level of assets endowment (Table 4.3).



Figure 4.2. Multi-dimensional indices for Danish dairy farms

Scaled Values at Class Means, 2010 to 2016


123. Finally, Class 3 dairy farms are the least productive but show a significant technical

change rate per year. Family labour is most important for those farms, which are

considerably smaller than the average dairy farm in Denmark in terms of herd size and land

endowment. Dairy farms in Class 3 should increase the size of their production operations

to become more profitable given the identified positive economies of scale (of about

1.1008). These farms are found as environmentally sustainable as the average based on the

various indicators used (such as stocking density, chemicals use per ha and probability of

producing organic). Dairy farms in Class 3 invest far less than the average dairy farm in

Denmark and most likely operate with less innovative milking technologies as, for example

pipeline systems or milking carousels. Their capital intensity is the lowest of all dairy farms

and consistently with this notion is the finding that farms in Class 3 employ lowest levels

of capital per cow. These dairy farms are the most diversified and their farm managers are

of higher than average age. Finally, dairy farms in Class 3 generate the lowest off-farm

income of the three classes and show the lowest level of assets endowment (Table 4.3).

124. In summary, Danish dairy farms in the three identified technology classes differ


over time. Innovative farms are most likely productive compared to their peer group and

show the highest technical change rates over time. Family driven farms and comparatively

smaller farms not necessarily show a higher environmental sustainability based on the

measures used in the empirical analysis. Highly environmentally sustainable dairy farms

most likely show a diverse Farm structure and most likely employ an innovative milking

technology. Farms’ capital intensity is not necessarily correlated with herd size, while

farms’ productivity is correlated with the share of hired labour. Finally, more

environmentally sustainable farms exhibit higher than average levels of productivity,

ceteris paribus.

-2.5

-2

-1.5

-1

-0.5

0

0.5

1


Index 2 - Environmental sustainability

Index 3 - Innovation/Cooperation/Commercialisation



Index 7 - Location

Index 8 - Household

Index 9 - Financial

Class 1: Most productive (66.9%) Class 2: Medium productive (15.9%)




Table 4.2. Multiple characteristics of Danish dairy farms, by class

Deviations from sample means1




Farm structure


Herd size (LU) 0.1169 -0.1887 -0.2817

Land (ha) 0.0379 0.2605 -0.3925




Organic (probability) -0.2062 1.0849 -0.2069

Environmental subsidies (EUR per ha) -0.1032 0.5558 -0.1158




Contract farming (prob) 0.0339 -0.0061 -0.1276

Technology

Capital / labour ratio (EUR per AWU) -0.0565 0.4872 -0.2345

Capital per cow (EUR per LU) -0.0481 0.3123 -0.1036

Fodder per cow (EUR per LU) -0.0146 0.0799 -0.0176

Milking system (1-pipes, 2-carousel, 3-AMS, 4-milking parlour, 5-others)

0.0012 0.3592 -0.3409

Diversity


Individual

Age (years) -0.0318 0.0021 0.1228

Farming experience (years) -0.0091 -0.0393 0.0726

Location

Municipality (various) 0.0144 -0.0305 -0.0281

Household

Off-farm income (EUR) -0.0292 0.1931 -0.0663

Financial

Total assets (EUR) 0.0131 0.2892 -0.3217

Total subsidies (EUR) 0.0749 0.3819 -0.6509


Note: LU: Livestock Unit. AWU: Annual Work Unit. 1. Deviations from sample means (=0), z-scores based, scaled

values.


4.2. Pig farms

125. Denmark is one of the world leaders in pig meat production in terms of

specialisation, productivity and share in global production. In total, about 3 250 farms in

2016 were engaged in pig meat production with about 37% of farms pursuing rearing and

fattening, about 41% specialising in fattening pigs and about 15% specialising in rearing

pigs. In total there are nearly 13 million pigs used for production in Denmark and nearly

70% of all Danish pig farms have more than 1 000 pigs compared to a EU28 average of

about 1.6% farms with more than 1 000 pigs (Danish Agriculture & Food Council 2017).

126. According to Table A D.4.5, which shows descriptive statistics for the sample of

Danish pig farms, the average total output per farm significantly increased between 2006



and 2016 due to a significant increase in the livestock herd value. In addition, the variable

cost items of pig farms increased over the period and the share of hired labour outmatched

the share of family labour in 2016. The livestock value per ha increased over the period

investigated and the share of organic production significantly decreased. In 2016 about

22% of all pig farms were engaged in rearing and fattening, 36% were specialised in pig

rearing and about 42% were specialised in fattening.


components for the pig meat production, the output variable is total output per farm and

year. The input variables are labour and livestock, land, capital, fodder, intermediates and

labour. The Class identification component is based on the indices related to structure,

environmental sustainability, innovation, technology, diversity, individual and household

characteristics, location and financial aspects (Table A D.4.6 and Table A D.4.7.).

128. With respect to specialised rearing pig farms, the empirical analysis did not allow

the identification of significantly different technologies for separate classes of farms. This

suggests that pig rearing production technologies in Denmark are highly mature and merely

differ between farms, hence, the subsequent empirical analyses focuses on the other types

of pig farms in Denmark for which significantly heterogeneous technologies could be

identified.

4.2.1. Danish rearing and fattening pig farms

129. For rearing and fattening pig farms in Denmark, three distinct technology classes

emerge from the model estimates (Table A D.4.6). The individual farms are distributed

unevenly across the three technology classes with Class 1 and Class 3 of almost equal size

including the majority of farms (Table 4.4). Class 1 covers about 48% of all dairy farms,

Class 2 about 9%, and Class 3 about 43%.



in EUR per year — than their pig rearing and fattening counterparts in classes 1 and 2

(Figure 4.4). The average output value per farm and per year of pig farms in Class 3 is

twice as high as that of farms in Class 1, and nearly 4 times higher than that of pig farms

in Class 2. Nevertheless, switching technologies could result in higher productivity growth

over time: pig farms in Class 3 show a significantly positive technical change (estimated at

1.86% per year) compared to a still considerable positive technical change rate for farms

in Class 1 (estimated at 1.33% per year) and a modest rate in Class 2 (estimated at 0.63%

per year). Hence, although pig farms in Class 3 are already far more productive, they also

seem to increase their productivity at a higher rate based on a significant positive technical

change development ceteris paribus (Table 4.4).

131. If rearing and fattening farms in Class 2 produced with the technology of farms in

Class 1, they would be able to slightly increase their productivity. However, the results

suggest that the different class technologies are very close to the optimal technological

choice for the farms in each individual class. By switching technologies, only slight

performance improvements seem possible at the individual farm level.



Figure 4.3. Productivity and technical change for Danish rearing and fattening pig farms,

by type and class, 2006 to 2016

Source: Table 4.3.

Table 4.3. Productivity characteristics of Danish rearing and fattening pig farms, by type

and class

Latent Class Estimations, Panel 2006 to 2016


Number of observations (% of sample farms) 2 369 (48.0%) 433 (8.8%) 2 132 (43.2%)




- Class 1 Technology 626 704*** 355 908*** 1 265 6910***









132. The estimated production functions are highly significant and all rearing and

fattening pig farm classes exhibit increasing returns to scale: 1.0473 for Class 1, 1.0703 for

Class 2, and 1.0729 for farms in Class 3 (see Table A D.4.11 for complete class related

elasticities). Switching technologies could — ceteris paribus — result in slightly higher

productivity per farm for a minority of farms, and a significantly higher technical change

rate for most farms (especially for a switch to Class 3 technology). Various indices,

reflecting the different dimensions along which rearing and fattening pig farms can be

distinguished, are used to robustly identify the reported farm classes: farm structure,

environmental sustainability of operations, technology characteristics, degree of diversity,

1.33

0.63

1.87

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

0

200

400

600

800

1 000

1 200

1 400

1 600


%'000 EUR




individual and household characteristics and locational as well as financial specifics.

Table A D.4.7 summarises the estimates for the various indices that were used as

components for the class-identifying vector in the latent class estimation. Most of the

indices considered showed significant estimates. To decide on the empirically most

appropriate number of classes to be estimated, various statistical quality tests have been



133. Figure 4.4 summarises the various rearing and fattening pig farm classes in terms

of estimated indices, while Table 4.4 contains the estimates for variables used to construct

the indices.

134. As discussed earlier, pig farms in Class 3 are the most productive and exhibit the

most significant positive technical change rate per year. They have the lowest share of


average farm in the sample) and an average herd and acreage size significantly above the

average of all farms in the sample, but they still experience (considerable) increasing

returns to scale. Pig farms in Class 3 score relatively low on environmental sustainability

indicators (such as stocking density, chemicals use per ha and probability of producing

organic). These farms, however, show the highest scores on innovation and

commercialisation with a significantly higher than average investment rate, rented land

share, and income generated by biofuels production. Class 3 farms show a higher than

average capital per labour intensity, while using less capital per livestock unit than the

average rearing and fattening pig farm in Denmark, based on high levels of total assets

endowment. These pig farms are less diversified and their managers are younger than the

average pig farmer in Denmark (Table 4.6).

135. The productivity of pig farms in Class 1 is medium and their technical change per

year is average. Family labour is more important than hired labour for those farms, which

are smaller than the average rearing and fattening pig farm in Denmark in terms of herd

size and land endowment. A single ownership is very likely for these pig farms. Rearing

and fattening pig farms in Class 1 should significantly increase the size of their production

operations given the measured economies of scale. The environmental sustainability of

these farms is found to be medium based on various indicators used (such as stocking

density, chemicals use per ha and probability of producing organic). Pig farms in Class 1

invest slightly less than the average pig farm in Denmark, but generate more than average

income by biofuel production. These pig farms are least specialised and their farm

managers are older than the average rearing and fattening pig farmer. Finally, pig farms in

Class 1 generate about average off-farm income and operate with a lower than average

assets endowment (Table 4.4).



Figure 4.4. Multi-dimensional indices for Danish rearing and fattening pig farms, by type




change rate per year. Family labour is most important for those rearing and fattening pig

farms, which are considerably smaller than the average pig farm in Denmark in terms of

herd size and land endowment. Pig farms in Class 2 should increase the size of their

production operations to become more profitable given the estimated positive economies

of scale. These farms, however, are found to be the most environmentally sustainable based

on various indicators used (mainly due to their significantly lower chemicals use per ha and

the high probability of producing organic). Pig farms in Class 2 are least innovative and

invest far less than the average rearing and fattening pig farm in Denmark. They generate

the lowest income from biofuel production. Their capital intensity is rather low and they

have the lowest share of hired labour in total labour among the three classes. These pig

farms are the least diversified and their farm managers are of higher than average age.

Finally, rearing and fattening pig farms in Class 2 operate with a significantly lower than

average asset endowment (Table 4.4).

137. In summary, Danish rearing and fattening pig farms in the three identified

technology classes significantly vary with respect to their economic performance as well

as technical development over time. Innovative pig farms in Denmark are most likely more

productive than their peer group. Family driven farms and comparatively smaller farms

show a higher environmental sustainability based on the measures used in the empirical

analysis. Highly environmentally sustainable pig farms not necessarily show a diverse farm

structure and are not per se less capital intensive. Farms’ productivity is, however,

positively correlated with herd size and the share of hired labour, ceteris paribus.

-1.5

-1

-0.5

0

0.5

1

1.5Index 1 – Structure


Index 3- Innovation/Cooperation/Commercialisation

Index 4 – Technology

Index 5 – DiversityIndex 6 – Individual

Index 7 – Location

Index 8 – Household

Index 9 – Financial






Table 4.4. Multiple characteristics of Danish rearing and fattening pig farms, by class





Farm structure


Livestock value (EUR) -0.4617 -0.5201 0.6187

Land (ha) -0.2553 -0.8797 0.4623

Legal form (1- sole owner, 2- partnership, assets owned, 3- partnership, assets rented, 4- legal corporation, 5- other special arrangement)

-0.1869 -0.0171 0.2112

Family succession (0- no, 1- yes) 0.0685 -0.2896 -0.0173


Livestock value per ha (EUR per ha) -0.0169 0.1463 -0.0109

Chemicals use (EUR per ha) -0.1989 -0.8738 0.3986




Net investment ratio (per total assets) -0.0911 -0.2445 0.1509

Share land rented -0.0281 -0.5591 0.1447

Contract farming (prob) 0.0869 -0.4789 0.0007

Biofuel income (EUR) 0.0115 -0.1266 0.0129

Technology


Capital per livestock value 0.0406 -0.0329 -0.0385

Fodder per livestock value 0.0815 -0.0402 -0.0824

Diversity



Individual

Age (years) 0.1414 0.1054 -0.1786

Location

Municipality (various) 0.0362 0.2415 -0.0893

Household

Off-farm income (EUR) 0.0135 -0.0008 -0.0149

Financial

Total assets (EUR) -0.2376 -0.7444 0.4152

Total subsidies (EUR) -0.2063 -0.7541 0.3823




4.2.2. Danish specialised fattening pig farms

138. With respect to specialised fattening type pig farming, three distinct technology

classes emerge from the model estimates (Table A D.4.8). The individual farms are

distributed unevenly across the three technology classes (Table 4.5). Class 1 covers about

19% of all pig fattening farms, Class 2 about 60%, and Class 3 about 21%.



in EUR per year — than their fattening pig producing counterparts in Classes 2 and 3



(Figure 4.5). Pig farms in Class 1 produce about 2 times more per farm and year than farms

in Class 2 and nearly 3.5 times more per farm and year than pig farms in Class 3.

140. Pig farms in Class 1 also show a significantly positive technical change (of about

2.08% per year) compared to a modest positive technical change rate for farms in Class 2

(of about 1.84% per year) and in Class 3 (of about 1.61% per year). Hence, although pig

farms in Class 1 are already far more productive, they also seem to increase their

productivity at a higher rate given the significantly positive technical change development


141. If pig farms in Class 2 produced with the technology of farms in Class 1, they would

be able to increase their productivity by 17%, and farms in Class 3 would increase their

productivity by about 31%, ceteris paribus (Table 4.5). Overall, if all farms adopted the

technology of most productive farms in Class 1, their productivity would increase on

average by 13% (Table A C.6 in Part 1).

Figure 4.5. Productivity and technical change for Danish specialised fattening pig farms,

by type and class, 2006 to 2016

Source: Table 4.5.

142. The estimated production functions are highly significant and all fattening pig farm

classes exhibit modest to significant increasing returns to scale: 1.0747 for Class 1, 1.2494

for Class 2, and 1.0199 for farms in Class 3 (see Table A D.4.12 for complete class related


productivity per farm and a significantly higher technical change rate for most farms

(especially regarding a switch to Class 1 technology). Various indices, reflecting the

different dimensions along which dairy farms can be distinguished, are used to robustly

identify the reported farm classes: Farm structure, environmental sustainability of

operations, technology characteristics, degree of diversity, individual characteristics and

locational specifics. Table A D.4.8 summarises the estimates for the various indices that

were used as components for the class-identifying vector in the latent class estimation. Most

of the indices considered showed significant estimates. To decide on the empirically most

appropriate number of classes to be estimated, various statistical quality tests have been



2.08

1.84

1.61

0.00

0.50

1.00

1.50

2.00

2.50

0

200

400

600

800

1000

1200


% '000 EUR

Type specialist fattening




Table 4.5. Productivity characteristics of Danish specialised fattening pig farms, by type and

class

Latent Class Estimations, Panel 2006 to 2016















143. Figure 4.6 summarises the various specialist fattening pig farm classes in terms of


indices.

144. As discussed earlier, pig farms in Class 1 are the most productive and exhibit the

most significantly positive technical change rate per year. They have the lowest share of


average farm in the sample) and a significantly above average herd and acreage size, but

they still experience (considerable) increasing returns to scale. Pig farms in Class 1 score

relatively low on environmental sustainability indicators (especially chemicals use per ha

and probability of producing organic). These farms, however, show the highest scores on

innovation and commercialisation with a significantly higher than average investment rate,

rented land share, and income generated by biofuels production.

145. Class 1 farms show a higher than average capital per labour intensity, while using

more capital per livestock unit than the average fattening pig farm in Denmark based on

high levels of total assets endowment. These pig farms are diversified, they operate with a

larger than average asset endowment and their managers are younger than the average

fattening pig farmer in Denmark (Table 4.6).

146. Specialised fattening pig farms in Class 2 are of medium productivity, but still show

a considerably positive technical change per year. Hired labour is less important for those

farms, which are smaller than the average pig farm in Denmark in terms of herd size and

land endowment. A single ownership is very likely for these farms. Pig farms in Class 2

should significantly increase the size of their production operations given the measured

economies of scale (of about 1.2494). These farms are found to be of average

environmental sustainability based on various indicators (such as stocking density,

chemicals use per ha and probability of producing organic). However, pig farms in Class 2

invest slightly less than the average fattening pig farm in Denmark and generate less income

by biofuel production. Their capital intensity is about average, and they employ more



labour per capital than their counterparts in Class 1. Diversification of activities in these

pig farms is medium and their farm managers are of average age (Table 4.6).

Figure 4.6. Multi-dimensional indices for Danish specialised fattening pig farms, by type


Source: Table A D 4.14.


change rate per year. Family labour is most important for those fattening pig farms, which

are significantly smaller than the average specialised fattening pig farm in Denmark in

terms of herd size and land endowment. Single ownership is most probable for these farms.

Pig farms in Class 3 operate close to the optimum scale and only experience slightly

positive economies of scale. These farms are found to be the most environmentally

sustainable based on various indicators used (such as stocking density, chemicals use per

ha and probability of producing organic). Pig farms in Class 3 invest far less than the

average pig farm in Denmark, have the lowest share of rented land and generate least

income from biofuel production. Their capital intensity is the lowest of all fattening pig

farms and consistently with this notion is the finding that farms in Class 3 employ the

highest amount of labour per capital unit. These pig farms are the least diversified and their

farm managers are of higher than average age. Finally, pig farms in Class 3 operate with

the lowest level of asset endowment (Table 4.6).

148. In summary, Danish specialised fattening pig farms in the individual technology

classes significantly vary with respect to their economic performance as well as technical

development over time. Innovative fattening pig farms in Denmark are most likely more

productive compared to their peer group. Family driven farms and comparatively smaller

farms show a higher environmental sustainability performance based on the measures used

in the empirical analysis. Highly environmentally sustainable pig farms are not necessarily

more diversified but operate with a lower than average capital intensity. Farms’ capital

intensity is positively correlated with herd size, while farms’ productivity is correlated with

herd size and the share of hired labour. Finally, Danish fattening pig farms producing more

environmentally sustainably can also exhibit a relatively high productivity, ceteris paribus.

-1.5

-1

-0.5

0

0.5

1Index 1 – Structure



Index 4 – Technology

Index 5 – DiversityIndex 6 – Individual









Table 4.6. Multiple characteristics of Danish specialised fattening pig farms, by class





Farm structure


Livestock value (EUR) 0.5793 -0.1177 -0.2025

Land (ha) 1.1302 -0.0917 -0.7858

Legal form (1- sole owner, 2- partnership, assets owned, 3- partnership, assets rented, 4- legal corporation, 5- other special arrangement)

0.1106 -0.0803 0.1251

Family succession (0- no, 1- yes) -0.0059 0.0381 -0.1023


Livestock value per ha (EUR per ha) -0.0169 -0.0274 0.0932


Organic (probability) 0.0568 -0.0407 0.0629

Environmental subsidies (EUR per ha) 0.0818 -0.0092 -0.0701




Contract farming (prob) 0.1795 0.1227 -0.5136

Biofuel income (EUR) 0.1743 -0.0268 -0.0854

Technology


Capital per livestock value 0.0985 -0.0111 -0.0598

Fodder per livestock value 0.1106 -0.0803 0.1251

Diversity



Individual

Age (years) -0.1007 0.0132 0.0557

Location

Municipality (various) -0.4847 0.0308 0.3611

Household


Financial

Total assets (EUR) 1.0406 -0.0964 -0.6897

Total subsidies (EUR) 1.1578 -0.1116 -0.7549




5. Estonia

5.1. Estonian dairy farms

149. In Estonia, the analysis applies to a sample of dairy farms covering the period 2000

to 2015. Dairy farming in Estonia accounts for a significant portion of the agricultural

sector, with milk production close to 30% of the total value of agricultural production in

recent years. The production of milk in 2017 amounted to about 792 000 tonnes, and



increased by 1% compared to 2016. This was due to both a slight increase in the number

of dairy cows (to 88 400 cows) and higher annual milk produced per cow (Statistics

Estonia, 2018).

150. According to Table A D.5.1, which summarises the descriptive statistical measures

for the Estonian sample, Estonian dairy farms significantly increased milk output between

2000 and 2015, following a significant increase in herd size. As can be expected of a

transition economy, all cost items increased over the period as did the labour use by the

average dairy farm whereas the share of family labour relatively decreased. Stocking

density slightly increased, while the probability of being engaged in organic dairy

production more or less remained at the same level.


components for the Estonian dairy production, the output variable is total output per farm

and year and input variables are dairy cows, capital, materials, fodder, fuel and other costs

(Box 1). The Class identification component is based on the indices related to structure,

environmental sustainability, innovation, diversity, individual and location

(Table A D.5.2).

152. For Estonian dairy farms, two distinct technology classes emerge from the model

estimates (Table A.B.5.2). Class 1 covers more than 80% of all dairy farms in the sample

and Class 2 the remaining 17%.

153. Farms in Class 1 show a significantly lower productivity performance — measured

as the potential output levels that could be achieved with a given input bundle, in EUR per

year — than their counterparts in Class 2 (Figure 5.1). The latter produce about thirteen

times more per farm and year than the former.

154. Farms in Class 1 show a significantly positive technical change (estimated at 2.16%

per year) compared to a modest positive technical change rate for farms in Class 2

(estimated at 0.346% per year) Hence, although dairy farms in Class 1 are far less

productive, they seem to catch up given the significant positive technical change

development (Figure 5.1).

155. If dairy farms in Class 1 would produce with the technology of farms in Class 2,

they would be able to significantly increase their productivity by about 4 times (Table 5.1).

Overall, this would result in an average productivity increase of about 85% for all dairy

farms in Estonia (Table A C.4).



Figure 5.1. Productivity and technical change for Estonian dairy farms,


Source: Table 5.1.

Table 5.1. Productivity characteristics of Estonian dairy farms by class


Class 1 Class 2

Number of observations (% of sample farms) 2 431 (83%) 504 (17%)



Productivity level (EUR per year) 1




- Class 1 Technology 2.16*** 0.699***


Note: * significant at 10%, ** significant at 5%, *** significant at 1%.1. Fitted values at sample means.


156. The estimated production functions are highly significant and both dairy farm

classes exhibit increasing returns to scale, which are slightly lower for farms in Class 2

(1.1799) than for farms in Class 1 (1.1896) (see Table A D.5.4 for complete class related

elasticities). Switching technologies between classes could — ceteris paribus — result in

higher productivity per farm but lower technical change for most farms (and vice versa).

Various indices, reflecting the different dimensions along which farms can be

distinguished, are used to robustly identify the reported farm classes: Farm structure,


individual characteristics and locational specifics. Table A D.5.4 summarises the estimates

for the various indices that were used as elements of the Class identifying vector in the

2.16

0.346

0

0.5

1

1.5

2

2.5

0

50

100

150

200

250

300

350

400

450

Class 1 (83%) Class 2 (17%)

%EUR '000




latent class estimation. Beside the index for locational specifics, all other indices showed

significant estimates.

157. Figure 5.2 summarise the two dairy farm classes in terms of estimated indices,


158. Dairy farms in Class 1 are less productive but show most significant positive

technical change per year. They have a relatively high share of family labour (i.e. a

significantly lower share of hired versus family labour than the average farm in the sample)

and a below average herd and area size, which is consistent with the estimated increasing

returns to scale for those farms. Dairy farms in Class 1 score high on environmental

sustainability characteristics (for example stocking density, chemicals use per ha and

probability of producing organic) but show a lower than average investment rate and

willingness to co-operate. Farms in Class 1 show a low capital intensity, but produce with

more fodder per cow. These dairy farms are more diversified, and their managers are a bit

older than the average Estonian dairy farmer. Finally, dairy farms in Class 1 are more likely

located in less-favoured areas and Natura 2000 regions compared to the average dairy farm

in Estonia (Table 5.2).

159. Dairy farms in Class 2 are highly productive, but show a lower than average

technical change per year. Hired labour is essential for those farms, the hired to family

labour ratio being significantly larger than in Class 1. Farms in Class 2 have a significantly

above average herd and area size, but should increase further their production to take

advantage of economies of scale. These farms score lower than average on various

indicators used (stocking density, chemicals use per ha and probability of producing

organic) but invest significantly more than the average dairy farm in Estonia. They are also

highly likely to cooperate with other farms. Their capital intensity is significantly higher

than the average sector capital intensity, however, farms in Class 2 employ more labour per

cow. These dairy farms are more specialised and their farm managers are younger than the

average dairy farmer in Estonia. Finally, dairy farms in Class 2 are less likely located in

less-favoured areas and Natura 2000 regions compared to farms in Class 1 and the average

dairy farm in Estonia (Table 5.2).

160. In summary, Estonian dairy farms in the two identified technology classes differ

significantly with respect to their economic performance and technical development over

time. Innovative dairy farms are likely to be more productive. Family driven farms and

comparatively smaller farms show a higher environmental sustainability based on the

indicators used. Highly environmentally sustainable dairy farms are most likely located in

less-favoured areas. Farms’ capital intensity and the degree of specialisation are positively

correlated with herd size, farms’ productivity is correlated with herd size and the share of

hired labour. Finally, dairy farms producing more environmentally sustainably in Estonia

appear to be less productive ceteris paribus.



Figure 5.2. Multi-dimensional indices for Estonian dairy farms



Table 5.2. Multiple characteristics of Estonian dairy farms, by class


Class 1: Least productive (83%)

Class 2: Most productive (17%)

Farm structure

Hired / family labour ratio -0.2177 1.0501

Herd size (livestock units) -64.2584 309.9446

Land (ha) -201.4147 971.506


Stocking density (livestock units per ha) -0.0514 0.2478

Chemicals use (EUR per ha) -302.4744 33.2247

Organic (probability) 0.0224 -0.1082


Net investment ratio (per total assets) -0.7852 0.1603

Cooperation (probability) -0.1332 0.6424

Technology

Capital / labour ratio (EUR per hour) -1.9849 9.5742

Capital / cow ratio (EUR per cow) -169.043 815.3643

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) -0.0017 0.0084

Fodder / cow ratio (EUR per cow) 0.0292 -0.1406

Labour per cow ratio (hour per livestock unit) -0.0533 0.2569

Individual

Age (years) 0.3362 -1.6217

Location

Less Favoured Area (probability) 0.0074 -0.0356

Natura 2000 (probability) 0.0001 -0.0006

Note: 1. Deviations from sample means (= 0), scaled values.


-20

-15

-10

-5

0

5

10



Index 3 -Innovation/Cooperation/

Commercialisation

Index 4 - TechnologyIndex 5 - Diversity


Index 7 - Location





6. France

161. In France, the analysis applies to a sample of crop farms covering the period 1989

to 2016 and to a sample of dairy farms covering the period 1990 to 2013.

6.1. French dairy farms


sample, the average milk output per farm was EUR 180 000 in 2018 (with a total output of

about EUR 261 000). The variable cost items increased over the time period investigated.

The average French dairy farm has a herd size of about 59 cows, which results in a stocking

density of about 1.48 LU per ha in 2018 (35 dairy cows and 1.62 LU per ha in 1990).


components for the French dairy production the output variable is total output per farm and

year and input variables are land, dairy cows, capital, labour and other materials. The




164. For French dairy farms, three distinct technology classes emerge from the model


technology classes (Table 6.1), as Class 2 includes over half of all farms, while Class 1 and

Class 3 share more equally the remaining farms.


potential output levels that could be achieved with a given input bundle, in value per year —

than their milk-producing counterparts in classes 2 and 3 (Figure 6.1). Dairy farms in

Class 1 produce about 20% more per farm and year than farms in Class 2 and about twice

as much per farm and year than dairy farms in Class 3.

166. Dairy farms in all classes show a negative technical change, which is larger for

Class 1 farms and close to zero for Class 2 and Class 3 farms. Hence, although dairy farms

in Class 1 are more productive than others, they also seem to decrease their productivity

faster than those in the least productive classes, ceteris paribus (Figure 6.1).

167. If dairy farms in Class 2 and Class 3 produced with the technology of farms in

Class 1, they would be able to increase their productivity by 10% and 39% respectively,

indicating that technologies in Class 1 and Class 2 are not much different (Table 4.1).

Overall, this would result in an average 13% increase in productivity for the French dairy

sector (Table A C.4 in Part 1).



Figure 6.1. Productivity and technical change for French dairy farms, by class, 1990 to 2013

Source: Table 6.1.

Table 6.1. Productivity characteristics of French dairy farms, by class



Number of observations (% of sample farms) 21% 51.6% 27.4%








- Class 1 Technology -0.560*** -0.932*** -0.788***

- Class 2 Technology -0.133*** -0.051* -0.041*

- Class 3 Technology -0.549*** -0.193*** -0.091**




all dairy farm classes exhibit increasing returns to scale: 1.0367 for farms in Class 1, 1.0646

for farms in Class 2 and 1.1484 for farms in Class 3 (Table A D.6.4). This suggests that an

increase in herd size would result in additional revenue for the dairy farms (ceteris paribus).



-0.56

-0.051

-0.091

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0

20

40

60

80

100

120

140

160

180


%EUR '000 EUR




170. Class 1 farms are the most productive farms with slightly below average

environmental sustainability account for over a quarter of all farms. They are more

specialised, larger operations, with higher assets, more likely to be partnerships and include

older farmers. Investment in innovative technologies and activities is above average. They

are more likely to use more extensive practices, including organic production.

171. Class 2 groups over half of all farms that are in the least environmentally

sustainable and achieve productivity levels close to the highest. They are more diversified

than others, use more intensive practices and are more likely to be in plains. They have

scores close to the average of all farms for most other indices.

172. Class 3 farms are the most environmentally sustainable, least productive farms.

They are smaller than average in terms of herd size, and more specialised. They use more

extensive technologies and practices and have lower capital intensity. They are more likely

to be located in mountainous areas and have lower debts than average.

173. In summary, French dairy farms in the three identified technology classes differ


over time. Most productive farms are less family driven and comparatively larger in terms

of herd size and land endowment. These farms show an average environmental

sustainability based on the measures used in the empirical analysis. Highly environmentally

sustainable dairy farms are nevertheless most likely family driven with a lower than

average herd size and lower than average productivity level. The capital intensity of farms

seems positively correlated with herd size, while their productivity seems correlated with

herd size and the share of hired labour, ceteris paribus.

Figure 6.2. Multi-dimensional indices for French dairy farms



-1.5

-1

-0.5

0

0.5

1



Index 3 - Innovation/commercialisation



Index 7 - Location

Index 8 - Household

Index 9 - Financial






Table 6.2. Multiple characteristics of French dairy farms, by class





Farm structure


Herd size (LU) 0.2318 -0.0122 -0.1549

Form of ownership (1-family farms, 2-partnerships, 3-other)

0.5141 -0.1553 -0.1016



Chemicals use (EUR per ha) -0.0419 0.2100 -0.3639

Organic production (1=yes, 0=no) 0.1688 -0.0531 -0.0294

Fuel per ha (EUR per ha) 0.5062 -0.0332 -0.3259

Environmental subsidies per ha (EUR per ha) 0.6383 -0.2416 -0.0342

Tillage area (ha) 0.3351 0.0778 -0.4039



Share contract farming 0.3086 0.0611 -0.3519


Biofuel income (EUR) 0.1721 -0.0255 -0.0839

Miscellaneous income (EUR) 0.0180 0.0119 -0.0362

Insurance expenses (EUR) 0.4782 -0.0850 -0.2066

Technology


Capital per cow (EUR per LU) 0.4331 -0.0766 -0.1879

Cow per labour (LU per AWU) -0.0914 0.0195 0.0333

Fodder per cow (EUR per LU) 0.1691 -0.0515 -0.0326

Materials per cow (EUR per LU) 0.0454 0.0405 -0.1111

Diversity



Forest area (ha) -0.0496 -0.0692 0.1686

Individual

Age (years) 0.3333 -0.1518 0.0305



Location

Subregion 1 nuts 2 FR10 0.0230 -0.0046 -0.0091

Subregion 2 nuts 2 FR21 0.0909 0.0178 -0.1033



Subregion 5 nuts 2 FR24 -0.0435 -0.0001 0.0336

Subregion 6 nuts 2 FR25 -0.1120 0.1088 -0.1192





Subregion 11 nuts 2 FR43 0.1361 -0.0761 0.0390





Subregion 16 nuts 2 FR62 -0.0963 0.2041 -0.3108







(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) 0.0421 -0.3461 0.6202

Less favoured area payments (EUR) 0.0391 -0.3772 0.6810

Household

Off-farm income share 0.1217 -0.0299 0.0638

Rural support (EUR) 0.8113 -0.2736 -0.1066

Financial

Total assets (EUR) 0.4840 -0.0489 -0.2791



Notes: LU: Livestock Unit. AWU: Annual Work Unit.



6.2. French crop farms


sample, the average crop output per farm was about EUR 167 000 in 2016 (with a total

output of about EUR 193 000). The variable cost items significantly increased over the

period analysed. The average crop farm cultivated about 136 ha of land in 2016 (a

significant increase considering the average area operated was about 77 ha in 1989).


components for the French crop production the output variable is total output per farm and

year and input variables are land, capital, labour and materials. The Class identification


innovation, technology, diversity, individual, location, household and financial aspects

(Table A D.6.7).



176. For French crop farms, four technology classes emerge from the model estimates

(Table A D.6.7). The individual farms are distributed unevenly across the three technology

classes (Table 6.3), as Class 1 includes over half of all farms, while Class 3 and Class 4

group less around 10% of all farms.

177. Farms in Class 1 show a productivity performance — measured as the potential

output levels that could be achieved with a given input bundle, in value per year — around

10% higher than their counterparts in classes 1 and 3 (Figure 6.3). Crop farms in Class 2

are significantly less productive than other farms, and produce about a third of farms in

Class 4.

178. The highest rate of technical change per year and per farm is found in for farms in

Class 1, suggesting they will quickly catch up with farms in Class 4, which show a fast

decline rate of technical change, ceteris paribus (Figure 6.3). Farms in Class 3 will also

achieve higher productivity as they have a positive rate of technical change, but more

slowly than farms in Class 1. Farms in Class 2, which account for a quarter of all farms,

are likely to see their productivity deteriorate as they show a strong negative rate of

technical change, and their productivity gap with farms in classes 1 and 3 increase.

Figure 6.3. Productivity and technical change for French crop farms, by class, 1989 to 2016

Source: Table 6.3.


would be able to increase their productivity by 40% (Table 6.3). For farms in classes 1

and 3, this would result in a deterioration of their productivity performance of -14% and

58% respectively (Table A C.5 in Part 1). The highest productivity gains for all farms

would occur if they adopted the technology of farms in Class 3, but that would require the

adoption of most intensive and less environmentally sustainable farming practices

(Table 6.4).

1.24

-2.39

0.77

-2.43

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

0

20

40

60

80

100

120

140

160

180

Class 1 (55.4%) Class 2 (24.6%) Class 3 (8.5%) Class 4 (11.6%)

%EUR '000




Table 6.3. Productivity characteristics of French crop farms, by class


Class 1 Class 2 Class 3 Class 4

Number of observations (% of sample farms) 55.4 24.6 8.5 11.6

Prior probability of class membership 0.8729 0.0843 0.0037 0.0391

Posterior probability of class membership 0.5176 0.2579 0.0925 0.1319


- Class 1 Technology 147 931*** 82 388*** 103 736*** 234 990***





- Class 1 Technology 1.242*** -0.322*** 0.073* 6.364***

- Class 2 Technology -0.824*** -2.394*** -1.624*** 3.513***

- Class 3 Technology -0.458*** 0.350* 0.769*** 4.146***

- Class 4 Technology -0.838*** -3.899*** -0.034* -2.431***



180. The estimated production functions for each crop farm class are highly significant

and x farms in Class 1 and Class 2 exhibit increasing returns to scale of about 1.0241 or

1.1479, respectively (Table A D.6.9). Those farms in Class 3 and Class 4, however, show

slightly decreasing returns to scale of about 0.9724 and 0.9639, respectively.

181. Figure 6.4 summarises the various crop farm classes in terms of estimated indices,


182. Class 1 farms group 55% of all farms have a close to highest productivity, and

above average environmental sustainability. They are larger operations (ha) and are more

likely to be partnerships. They invest more in new technologies and are more capital

intensive. They receive the highest subsidies. Class 2 farms are the least productive farms

and achieve lower than average environmental sustainability. They are smaller, intensive

and specialised operations, with lower investment and capital intensity than average. They

are more likely to be located in remote areas, and to have off-farm income.

183. Class 3 farms are about 8% of all farms and have close to highest productivity, but

are the least environmentally sustainable. They are the smallest operations, most diversified

and reliant on family labour. They are operated by younger farmers and have high

investment rates, and high capital per ha.

184. Class 4 farms are the most productive and environmentally sustainable. As in

Class 1, they are larger operations and are more likely to be partnerships, but rely more on

family labour. They are operated by older farmers, which invest less but are more likely to

have biofuel production. They are also more likely to be in more rural areas.

185. In summary, French crop farms in the four identified technology classes differ


over time. Innovative farms are more likely to show a higher productivity. Family driven

farms and smaller farms in terms of acreage are less environmentally sustainable and less



productive, ceteris paribus. Capital intensiveness of production is negatively correlated

with the environmental sustainability of operations, whereas diverse farms are not

operating with at a higher environmental sustainability level. The productivity of crop

farms is linked to the farms’ structure. Finally, the empirical results at unweighted class

level suggest that environmentally sustainable crop farming is positively correlated with a

higher economic productivity, ceteris paribus.

Figure 6.4. Multi-dimensional indices for French crop farms



-2

-1.5

-1

-0.5

0

0.5

1

1.5






Index 7 - Location

Index 8 - Household

Index 9 - Financial

Class 1: Medium productive (55.4%) Class 2: Least productive (24.6%)

Class 3: Medium productive (8.5%) Class 4: Most productive (11.6%)



Table 6.4. Multiple characteristics of French crop farms, by class






Farm structure

Family/hired labour ratio -0.0047 -0.0827 0.0383 0.1699

Land endowment (ha) 0.1826 -0.5038 -0.7904 0.7770


0.1103 -0.2677 -0.1640 0.1612


Chemicals use (EUR per ha) -0.0228 -0.2038 0.5971 0.1028

Organic production (1=yes, 0=no) 0.0268 -0.0446 -0.1012 0.0410

Fuel per ha (EUR per ha) -0.0125 -0.0123 0.0662 0.0375

Environmental subsidies per ha (EUR per ha)

0.0247 -0.0605 -0.1450 0.1167

Tillage area (ha) 0.1875 -0.5019 -0.7912 0.7503


Net investment ratio (per total assets) 0.2318 -0.4760 0.1959 -0.2424

Share contract farming -0.0079 -0.0237 0.1339 -0.0103

Share land rented 0.1712 -0.5292 -0.0575 0.3467

Biofuel income (EUR) 0.0443 -0.1490 -0.1462 0.2119

Miscellaneous income (EUR) 0.0986 -0.3756 0.1861 0.1890

Insurance expenses (EUR) 0.1493 -0.6287 -0.1779 0.7515

Technology

Capital / labour ratio (EUR per AWU) 0.2974 -0.3969 -0.2432 -0.4015

Labour per ha (AWU per ha) -0.1336 -0.0912 1.3781 -0.1803

Capital per ha (EUR per ha) -0.0016 -0.2067 1.0677 -0.3383

Materials per ha (EUR per ha) -0.0475 -0.0551 0.5476 -0.0585

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) -0.0307 0.2773 -0.5353 -0.0483

Production diversity (yc/ΣY) -0.0605 -0.1606 1.0496 -0.1411

Forest area (ha) -0.0368 0.1136 -0.0241 -0.0474

Individual

Age (years) -0.1584 -0.2213 -0.3049 1.4519



Location

Subregion 1 nuts 2 FR10 0.0932 -0.1557 -0.2150 0.0427

Subregion 2 nuts 2 FR21 0.0643 -0.0372 -0.2874 -0.0172


Subregion 4 nuts 2 FR23 0.0466 -0.0884 0.0324 -0.0589

Subregion 5 nuts 2 FR24 -0.0184 0.0939 -0.2630 0.0821




Subregion 9 nuts 2 FR41 0.0168 -0.0607 -0.1689 0.1726



Subregion 12 nuts 2 FR51 -0.0433 0.0486 0.0839 0.0422

Subregion 13 nuts 2 FR52 -0.0085 -0.0270 0.1703 -0.0270


Subregion 15 nuts 2 FR61 -0.0317 0.0447 0.1251 -0.0353











Subregion 26 nuts 2 FRA1 -0.0278 -0.0067 0.0160 0.1356



(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) -0.1004 0.1846 -0.0999 0.1620

Less favoured area payments (EUR) -0.0531 0.3697 -0.0637 -0.4843

Household

Off-farm income share -0.0673 0.1919 -0.0687 -0.0352

Rural support (EUR) -0.0154 -0.1141 -0.1432 0.4211

Financial

Total assets (EUR) 0.2226 -0.5737 -0.0665 0.2023

Total subsidies (EUR) 0.2120 -0.1623 -0.0942 -0.6007

Equity/debt ratio -0.0082 -0.0048 -0.0083 0.0556

Notes: AWU: Annual Work Unit. 1. Deviations from sample means (=0), z-scores based, scaled values.




7. Hungary

7.1. Hungarian crop farms

186. In Hungary, the analysis applies to a sample of crop farms covering the period

2001 to 2014. Crop production in Hungary is very diverse, and dominated by cereals,

oilseeds and industrial crops. The major share of agricultural area is used for wheat and

maize production, followed by sunflower, barley and rapeseed production. Alfalfa, oat,

potatoes and sugar beet are also produced in Hungary.

187. The average utilised agricultural area per farm has been about 44 ha in 2016 with

an average employment rate of about 1.5 persons per farm (Hungarian Chamber of

Commerce, 2017). Private farms cultivated about 54% of the agricultural area, while about

36% belonged to enterprises and about 10% to holdings. Enterprises had the highest share

of arable land and grassland in total area. Structural change in Hungarian agriculture

remains high: whereas the number of farms with less than 4 ha of land decreased in the

period 2013 to 2016, those cultivating more than 4 ha increased. The number of farms

holding between 50 and 200 ha most significantly expanded, but those cultivating between

300 and 1 000 ha also increased. The average size of agricultural enterprises fell from 310

to 253 ha between 2013 and 2016, the average size of private farms increased to about

7.6 ha.


Hungarian crop farm sample, the crop output of the average Hungarian crop farm increased

between 2001 and 2014, mainly due to productivity improvements as for example, the

acreage per farm more or less remained the same over the 15 years considered. The variable

input costs increased as expected, the labour structure changed only slightly and finally the

probability of being engaged in organic production did not increase dramatically.


components for the Hungarian crop production the output variable is the total output per

farm and year and the inputs variable are land, capital, chemicals, materials and labour. The

Class identification vector consists of all seven indices as outlined in Table A D.7.2.

190. For Hungarian crop farms, three distinct technology classes emerge from the model

estimates (Table A D.7.2). Class 3 comprises half of the crop farms in the sample, while

the other half is distributed relatively evenly across the first two classes (about 28% in

Class 1 and about 22% in Class 2) (Table 7.1).

191. Farms in Class 1 show the highest productivity performance — measured as the


followed by crop farms in Class 3 and Class 2. The first class produces on average about

two times more per farm and year than farms in Class 2 (Figure 7.1).

192. Farms in Class 1 show a significantly positive technical change per year (of about

5% per year). However, crop farms in the other two classes also experience considerable

technical change (of about 4% per year for Class 3 farms and about 2% for Class 2 farms).

Hence, despite crops farms in Class 2 being far less productive than their counterparts in

the other two classes, they also seem to catch up in their performance based on a significant

positive technical change development (Figure 7.1). Finally, significant scale effects are

found for all classes: about 1.1366 for farms in Class 1, about 1.1758 for farms in Class 3,

and about 1.1212 for farms in Class 3 (Table A D.7.4).




would be able to significantly increase their productivity by about 1.5 times, while those in

Class 3 would increase their productivity by about 1.3 times (Table 7.1). Overall, this

would result in a 20% average increase in productivity for crop farms in Hungary

(Table A C.5).

Figure 7.1. Productivity and technical change for Hungarian crop farms,


Source: Table 7.1.

Table 7.1. Productivity characteristics of Hungarian crop farms, by class



Number of observations (% of sample farms) 3 933 (28%) 3 057 (22%) 7 138 (50%)



Productivity level (EUR per year) 1

- Class I Technology 123 355*** 86 131*** 145 702***

- Class II Technology 83 118*** 57 234*** 101 146***

- Class III Technology 94 459*** 64 814*** 116 021***


- Class I Technology 5.003*** 4.987*** 5.157***

- Class II Technology 2.246*** 2.007*** 2.284***

- Class III Technology 4.212*** 4.048*** 3.994***



194. Figure 7.2 summarises the three crop farm classes with respect to the various

indices used to identify the class membership of individual crop farms, while Table 4.7

contains the estimates for variables used to construct the indices. All dimensions considered

to distinguish crop farms are found statistically significant (Table A D.7.4).

5.00

2.01

3.99

0

1

2

3

4

5

6

0

20

40

60

80

100

120

140

Class 1 (28%) Class 2 (22%) Class 3 (50%)

%EUR '000




195. Farms in Class 1 are the most productive operations with the highest annual

technical change rate. These crop farms have the highest share of family labour in the

sample but operate with a below average farms size with increasing returns to scale. Farms

in Class 1 have most probably a single owner. They score relatively low on sustainability

(depending on the selected indicators: chemicals used, organic production, and

environmental subsidies received). These farms show the lowest share of rented land but a

higher than average probability of being engaged in biofuel production. They have a

medium to high input intensity, are rather specialised and generated only a small amount

of income from non-crop agricultural production. The farm managers are younger than

average, and off-farm income is of average importance. Finally, these crop farms are least

likely located in less-favoured areas but likely in nitrate vulnerable zones (Table 7.2).

Figure 7.2. Indices for Hungarian crop farms



196. Farms in Class 2 are about 24% of the total sample and are the least productive with

the lowest but still positive technical change per year (of about 2%). They have a below

average share of family labour, but are also relatively small exhibiting significantly

increasing returns to scale (of about 1.1757). Crop farms in this class have most likely a

single owner and score high on sustainability (depending on the selected indicators:

chemicals used, organic production, and environmental subsidies received). Those farms

show a very high input intensity and a high probability of biofuel production. They have

an average degree of diversity but show a significant probability of additional off-farm

income. The farm manager is slightly older than average and the farm is very likely located

in less-favoured areas of Hungary (Table 7.2).

197. Class 3 farms are nearly as productive as farms in Class 1 and show a positive rate

of technical change of about 4% per year. For those crop farms family labour is of least

importance and most likely more than one owner is registered for those farms. They are

significantly larger than the average crop farm size in Hungary, however, still exhibit

-3

-2

-1

0

1






Index 7 - Location

Class 1: Most productive (28%) Class 2: Least productive (22%)

Class 3: Medium productive (50%)



positive scale effects (of about 1.1212). Those farms also operate with a relatively high

sustainability (depending on the selected indicators used: chemicals used, organic

production, and environmental subsidies received). They have the highest share of rented

land and are most capital intensive and least labour intensive. However, farms in Class 3

are more diversified than the average crop farm in Hungary and are least relying on off-

farm income, and invest more than the average crop farm.

198. In summary, Hungarian crop farms in the three identified technology classes differ

significantly with respect to economic performance and technical development over time.

Innovative farms are more likely to be productive, while family driven farms and smaller

farms in terms of acreage can be innovative. Highly sustainable crop farms are most likely

located in less-favoured areas. Capital intensity and a low level of labour use are positively

correlated with farm size. The productivity of crop farms is not linked to economic size, to

family labour share in total labour or the type of ownership. Finally, the results suggest that

more sustainable crop farms in Hungary are most likely less productive.

Table 7.2. Multiple characteristics of Hungarian crop farms, by class





Farm structure

Family labour (hour per year) 282.906 -1.678 -155.161

Land (ha) -48.471 -66.682 55.265

Sole ownership (probability) 0.023 0.063 -0.039

Sustainability


Organic (probability) -3.14E-03 2.26E-03 7.63E-04

Environmental subsidies (EUR per ha) -67.563 -96.512 60.561


Net investment ratio (per total assets) 9.06E-03 1.36E-03 7.31E-03

Share Land Rented -0.012 -7.16E-03 9.46E-03

Biofuel production (probability) 3.26E-03 4.12E-03 -3.56E-03

Technology

Labour / capital ratio (hour per EUR) 2.183 2.408 -2.234

Materials per ha (EUR per ha) 10.297 30.685 -18.815

Labour per ha (hour per ha) 2.218 23.773 -11.403

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) 0.013 1.43E-03 -7.43E-03

Livestock production (probability) -0.043 -0.028 0.036

Other output (probability) -0.035 -0.053 0.042

Individual

Age (years) -1.101 2.179 -0.327

Off-Farm work (probability) -0.01 0.015 -6.82E-04

Location

Less Favoured Area (probability) -0.011 5.18E-03 3.91E-03

Nitrate Vulnerable Zone (probability) 0.022 -0.021 -3.21E-03

Note: 1. Deviations from sample means (=0), scaled values.




8. Ireland

199. In Ireland, the analysis applies to the following farm samples: dairy farms, crop

farms, cattle rearing farms, cattle other farms and sheep farms, all covering the period 2010

to 2018.

200. The analysis presented in this section is based on data from the Teagasc, National

Farm Survey, and provider of data to the Farm Accountancy Data Network (FADN). The

same dataset is used on an annual basis by Teagasc researchers to examine the economic,

environmental and social sustainability of for a range of farm types in Ireland. Given that

the sustainability metrics used in the Teagasc report differ from those employed in the

current study (and reported on below) are based on the same dataset it is important to

1) outline how the metrics differ and 2) how the results differ. Box 8.1 below outlines the

metrics and results from the most recent Teagasc report (Buckley et al., 2019).

Box 8.1. Main findings on farm performance in the Teagasc Sustainability Report

Using data compiled through the Teagasc National Farm Survey (NFS), Buckley et al.

(2019) track the performance of Irish farms in terms of economic, environmental and social

sustainability. While the primary focus of the Teagasc NFS is measurement of economic

performance, in recent years the scope of the survey has been expanded to include a

growing range of environmental metrics, including farm level greenhouse gas emissions.

The 2019 edition of the report also includes metrics for farm level ammonia emissions for

the first time.

Economic sustainability is measured by the following variables: economic return to land,

profitability, productivity of labour, economic viability, market orientation.

Environmental sustainability is measured by the following variables: agriculture

Greenhouse Gas (GHG) emissions per farm, per hectare, and per kg of output, energy GHG

emissions per farm and per kg of output, ammonia emissions per farm, per hectare and per

kg of output, Nitrogen (N) balance, N use efficiency, N surplus per kg of output,

Phosphorous (P) balance and P use efficiency.

Social sustainability is measured by the following variables: household vulnerability,

agricultural education, isolation risk, age profile and work life balance.

Innovation indictors: A range of sector specific innovation indicators are also included in

the report.

Main findings, relevant for comparison purposes

The report illustrates the income gap between dairy farms and other farm types in Ireland,

which has accelerated with the growth in Irish milk production in recent years. However,

the report also highlights that within each farm type there is a wide range in income levels

across the farm population. This fact tends to be overlooked when attention is focused only

on the average level of income for particular farm types.

The report illustrates that farm level emissions efficiency is improving, with a trend towards

fewer emissions of greenhouse gases and ammonia per unit of product produced. However,

the report also shows that emissions of GHG and ammonia over time are increasing on

farms that are growing in size.



The report also demonstrates the positive relationship between economic profitability and

emissions efficiency, with the highest levels of emissions efficiency tending to be found on

the most profitable farms.

This suggests that improvement in economic sustainability can be achieved side by side

with improvements in emissions efficiency.

Source: Buckley, C. T. Donnellan, E. Dillon, K. Hanrahan, B. Moran and M. Ryan (2019), Teagasc National

Farm Survey 2017 Sustainability Report, Agricultural Economics and Farm Surveys Department,

Rural Economy and Development Programme, Teagasc, Athenry, Co. Galway, 26 March,

www.teagasc.ie/media/website/publications/2019/2017-sustainability-report-250319.pdf

8.1. Irish dairy farms


sample, the average milk output per farm in 2018 was about EUR 231 000 (with a total

output of about EUR 251 000). The variable cost items increased over time and the share

of hired labour significantly increases for the average Irish dairy farm in the period

investigated. The average dairy farm in the sample operated in 2018 with a herd size of

about 84 cows (about 65 cows in 2010) and a related stocking density of about 1.96 LU per

ha (about 1.72 LU per ha in 2010).


components for Irish milk production, the output variable is total output per farm and year

and input variables are land, capital, dairy cows, labour and materials. The




203. For Irish dairy farms, three distinct technology classes emerge from the model


three technology classes (Table 8.1), as Class 1 includes more than half of all farms and

Class 2 less than 8%.




Class 1 produce per farm and year about twice the output of farms in class 2 and 3, which

show similar levels of productivity. However, while farms in Class 2 show a significant

positive technical change, which is larger for Class 1 farms, technical change is negative

for farms in Class 3. As a result, farms in Class 2 are reducing their productivity gap with

farms in Class 1, while farms in Class 3 are increasing their productivity gap with both

Class 1 and Class 2 farms, ceteris paribus (Figure 8.1).


Class 1, they would be able to increase their productivity by 36% and 19% respectively,

suggesting that technologies in Class 1 and Class 3 are relatively close (Table 8.1). Overall,

the adoption of Class 1 technologies by all farms would result in an average 7% increase

in productivity for Irish dairy farms (Table A C.4 in Part 1). In practice, however, natural

and human resource constraints may limit the adoption of the best technology by other

classes.

http://www.teagasc.ie/media/website/publications/2019/2017-sustainability-report-250319.pdf



Figure 8.1. Productivity and technical change for Irish dairy farms,


Source: Table 8.1.

Table 8.1. Productivity characteristics of Irish dairy farms, by class



Number of observations (% of sample farms) 51.2 7.9 40.9








- Class 1 Technology 0.277* 0.135* 0.302*


- Class 3 Technology -0.731** -0.343* -0.385*




and all dairy farm classes exhibit slightly significantly increasing returns to scale: 1.0244

for farms in Class 1, 1.1393 for farms in Class 2, and 1.0591 for farms in Class 3

(Table A D.8.4).



0.28

0.56

-0.39

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

0

50

100

150

200

250

300


%'000 EUR




208. Class 1 farms are the most productive and least environmentally sustainable and

achieve small but positive technical change. They are much larger than average in terms of

herd size and have more diversified operations. They invest much more in new technologies

and activities than average and are more capital intensive, based on larger assets

endowment than average. Their operators are younger and more likely to be women.

209. Class 2 farms are the least productive, with negative technical change and below

average environmental sustainability. They are have fewer animals than average, but are

more diversified. They have lower than average levels of investment and capital intensity.

Their operators are more likely to be older and male. They are more likely to be in less

favoured areas, have lower assets and receive lower levels of subsidies.

210. Class 3 farms are the most sustainable in environmental terms, with close to lowest

productivity and negative technical change accounts for about 40% of all farms. They are

smaller operations, relying mainly on family labour, and their operators are older than

average. They have lower than average levels of investment. They are more likely to be in

less favoured areas, and have lower assets.

211. In summary, Irish dairy farms in the three identified technology classes differ


over time. Innovative farms are most likely more productive. Family driven and smaller

farms are operating with a slightly higher environmental sustainability relative to the

average sector level. Dairy farms with lower environmental sustainability most likely show

a diverse structure with an average capital intensity. The productivity of farms is correlated

with herd size, ceteris paribus. Around half of the Irish dairy sector exhibits high

productivity and innovativeness, however, less than average environmental sustainability.

Figure 8.2. Multi-dimensional indices for Irish dairy farms



-1-0.8-0.6-0.4-0.2

00.20.40.60.8






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Table 8.2. Multiple characteristics of Irish dairy farms, by class





Farm structure


Herd size (LU) 0.4647 -0.1100 -0.5598


Stocking density (LU PER ha) 0.1929 -0.1377 -0.2147


Organic production (1=yes, 0=no) -0.0656 0.0783 0.0670

Fuel per ha (EUR per ha) -0.0586 -0.0302 0.0791

Environmental subsidies per ha (EUR per ha) 0.0971 -0.1653 -0.0896

Tillage area (ha) 0.1925 -0.1352 -0.2146

REPS participation (1=yes, 0=no) 0.1221 -0.1803 -0.1179

Water charges (EUR) -0.0676 0.2277 0.0407

Nitrate derogation (1=yes, 0=no) 0.1427 -0.0200 -0.1745

Innovation -commercialisation


Share contract farming 0.4232 -0.5152 -0.4300


Energy crops area (ha) 0.0199 -0.0567 -0.0140



E-Profit monitoring (1=yes, 0=no) 0.3794 -0.5964 -0.3595

Professional/Advisory expenses (EUR) 0.4027 -0.4411 -0.4185

Technology


Capital per cow (EUR per LU) 0.0091 -0.3645 0.0588

Cow per labour (LU PER AW) 0.3487 -0.0900 -0.4186

Fodder per cow (EUR per LU) -0.0354 -0.3331 0.1084

Materials per cow (EUR per LU) 0.0067 -0.3275 0.0547

Diversity



Forest area (ha) 0.0549 0.0250 -0.0734

Individual

Age (years) -0.1709 0.1798 0.1791

Agricultural education operator (1 = Full-Time 3rd Level Agri Course, 2 = Farm Apprenticeship Scheme, 3= Certificate in Farming, 4= 1 yr Ag College, 5 = Course > 60 hours, 6= Course < 60 hours, 7 = Other)

0.1939 -0.4263 -0.1603

Gender operator (1-male, 0-female) -0.1091 0.2320 0.0918

Marital status (1-married, 0-other) -0.0209 0.1287 0.0014



Location

Subregion 1 nuts 2 -0.3088 0.4182 0.3056

Subregion 2 nuts 2 -0.0170 0.0651 0.0087

Subregion 3 nuts 2 0.1590 -0.2757 -0.1457

Subregion 4 nuts 2 0.0113 -0.0923 0.0036

Subregion 5 nuts 2 -0.0987 -0.1256 0.1476

Subregion 6 nuts 2 0.2281 -0.3226 -0.2231

Subregion 7 nuts 2 0.0263 0.2150 -0.0743

Subregion 8 nuts 2 -0.1793 0.2392 0.1781

(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) 0.0526 0.0584 -0.0769

Less favoured area (1 = Not-disadvantaged, 2 = Less severely disadvantaged,

3 = Severely disadvantaged)

-0.4104 0.5080 0.4153

Household

Off-farm income (EUR) 0.0560 0.0266 -0.0751

Rural support (EUR) -0.2876 0.4293 0.2770

Part-time farming (1-yes, 0-no) -0.1967 0.2935 0.1894

Household size (n) 0.0753 -0.1070 -0.0736

Financial

Total assets (EUR) 0.4322 -0.4227 -0.4590





8.2. Irish crop farms


sample, the average crop output per farm in 2018 was about EUR 84 000 (with a total


of hired labour significantly increased for the average Irish crop farm over the period

investigated. The average Irish crop farm in the sample operated about 87 ha in 2018

(increased from about 72 ha in 2010).


components for the Irish crop production, the output variable is total output per farm and

year and input variables are land, capital, labour chemicals and materials. The




214. For Irish crop farms, three distinct technology classes emerge from the model


three technology classes (Table 8.3), as Class 3 includes about three-quarters of all farms.

215. Crop farms in Class 1 show a higher productivity performance — measured as the


than their counterparts in classes 2 and 3 (Figure 8.3). Crop farms in Class 1 produce three

times more output per farm and year than farms in Class 2 and twice the output of farms in

Class 3. While technical change growth at a slow rate for farms in Class 1, they are likely

to increase their productivity advantage over other farms, ceteris paribus, as both farms in

classes 2 and 3 show negative rates of technical change (Figure 8.3).



Figure 8.3. Productivity and technical change for Irish crop farms,


Source: Table 8.3.


would be able to more than triple their productivity, while the gain would be 22% for farms

in Class 3 (Table 8.3). Overall, the productivity of Irish crop farms would increase on

average by 45% (Table A C.5 in Part 1). In practice, however, natural and human resource

constraints may limit the adoption of the best technology by other classes.

Table 8.3. Productivity characteristics of Irish crop farms, by class












- Class 2 Technology -1.229* -1.222** -1.574**

- Class 3 Technology -1.947*** 1.217* -0.915**



0.10

-1.22

-0.92

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0

50

100

150

200

250


%'000 EUR





and crop farms in Class 1 and Class 3 exhibit significantly increasing returns to scale of

about 1.6294 and 1.4023 (Table A D.8.9). However, crop farms in Class 2 operate with

nearly constant returns to scale (1.0056).



219. Class 1 groups the most productive farms, with lower than average environmental

sustainability and accounts for a third of all farms. They are larger (ha) family-driven

operations. They use variable inputs and capital more intensively than average. They invest

much more than other farms in new technologies. They receive more subsidies per ha and

have much higher assets than average. Their operator is more likely to be a woman, to have

lower agricultural training and to be younger than average.

220. Class 2 farms are the least productive farms and also the least environmentally

sustainable of all crop farms in Ireland. They are smaller, more specialised operations, with

lower investment in technologies. They are more likely to be engaged in contract farming

and to cultivate energy crops. Their operators are older and hold more agricultural training.

They are more likely to be part-time and have a higher share of off-farm income.

221. Class 3 groups the most environmentally sustainable farms achieving productivity

levels close to the highest. They are smaller, more diversified operations than average, and

are more likely to be engaged in agri-environmental schemes. They have low capital

intensity and are more likely to be in less-favoured areas. Overall, they receive less

subsidies per ha than average.

222. In summary, Irish crop farms in the three identified technology classes differ


over time. For crop farming in Ireland, a strong positive correlation between the level of

innovativeness and the productivity of farming operations is also found. Those most

productive crop farms are less family driven and cultivate a larger area. However, these

farms are also less environmentally sustainable than the average crop farm in Ireland.

About a third of all crop farms show a low productivity and also very low environmental

sustainability scores based on the measures used in the empirical analysis. Those farms are

mainly family-driven and smaller than average, and they are the less innovative in the crop

sector.



Figure 8.4. Multi-dimensional indices for Irish crop farms



Table 8.4. Multiple characteristics of Irish crop farms, by class





Farm structure


Land (ha) 0.5239 -0.3606 -0.1572




Organic production (1=yes, 0=no) -0.0399 0.1090 -0.0729

Fuel per ha (€ per ha) 0.4338 -0.3309 -0.0964

Environmental subsidies per ha (€ per ha) -0.0025 -0.1309 0.1388

Tillage area (ha) 0.3797 -0.1841 -0.1942

REPS participation (1=yes, 0=no) -0.1135 -0.0928 0.2119

Water charges (EUR) 0.0989 -0.0646 -0.0333

Nitrate derogation (1=yes, 0=no) 0.0455 0.1050 -0.1555



Share contract farming -0.0330 0.0559 -0.0246


Energy crops area (ha) -0.0529 0.1125 -0.0633

Miscellaneous income (EUR) 0.1547 -0.0959 -0.0574




Technology

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8






Index 7 - Location

Index 8 - Household

Index 9 - Financial







Capital per land ratio (EUR per ha) 0.3801 -0.2120 -0.1656

Labour per land (AWU/ha) -0.1402 -0.0575 0.2023

Material per land (EUR per ha) -0.0185 0.0912 -0.0761

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) -0.1920 0.5714 -0.3996

Production diversity (yc/ΣY) 0.3264 -0.6547 0.3498

Forest area (ha) 0.1699 0.0319 -0.2059

Individual

Age (years) -0.0845 0.0994 -0.0176


0.3241 -0.2703 -0.0480

Gender operator (1-male, 0-female) -0.1561 0.1694 -0.0176

Marital status (1-married, 0-other) -0.3080 0.4311 -0.1357

Location

Subregion 1 nuts 2 0.1824 -0.0545 -0.1287

Subregion 2 nuts 2 -0.2006 0.0229 0.1800

Subregion 3 nuts 2 0.1869 -0.0424 -0.1459

Subregion 4 nuts 2 0.1104 0.0144 -0.1272

Subregion 5 nuts 2 -0.1102 0.2135 -0.1102

Subregion 6 nuts 2 -0.2641 -0.0240 0.2934

Subregion 7 nuts 2 -0.1620 0.0833 0.0779

Subregion 8 nuts 2 0.0000 0.0000 0.0000

(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) 0.1890 -0.1093 -0.0783

Less favoured area (1 = Not-disadvantaged, 2 = Less severely disadvantaged, 3 = Severely disadvantaged)

0.0013 -0.1130 0.1162

Household


Rural support (EUR) 0.2795 -0.3020 0.0302

Part-time farming (1-yes, 0-no) -0.6002 0.5944 -0.0087

Household size (n) 0.3075 -0.4355 0.1408

Financial

Total assets (EUR) 0.7127 -0.5744 -0.1265

Total subsidies per ha (EUR) 0.2251 -0.0066 -0.2220




8.3. Irish cattle farms

223. According to Table A D 8.9, which contains descriptive statistical measures for the

sample, the average livestock related output per cattle rearing farm was about EUR 29 000

in 2018 (with a total output of about EUR 44 000). For cattle other farms in the sample, the

livestock related output was about EUR 51 000 in 2018 (about EUR 30 000 in 2010). The

variable cost items increased over time for both cattle farm types and the share of hired

labour slightly increases for the average Irish cattle farm in the period investigated. The

average Irish cattle rearing farm in the sample operated a herd size of about 44 LU in 2018

(about 43 LU in 2010), and the average cattle other farm a herd size of about 72 LU in 2018

(about 63 LU in 2010). Cattle rearing farms exhibited a stocking density of about 1.09 LU



per ha in 2018 (about 0.97 LU per ha in 2010), and cattle other farms about 1.89 LU per ha

in 2018 (about 3.08 LU per ha in 2010).


components for the Irish beef production, the output variable is total output per farm and

year and input variables are land, livestock units, capital, labour and materials. The




8.3.1. Irish cattle rearing farms

225. For Irish cattle rearing farms, three distinct technology classes emerge from the

model estimates (Table A D.8.10). The individual farms are distributed very unevenly

across the three technology classes (Table 8.5), as Class 3 includes close to two-thirds of

all farms.

226. Cattle rearing farms in Class 1 show a higher productivity performance


in value per year — than their counterparts in classes 2 and 3 (Figure 8.5). Farms in Class 1

produce 25% more output per farm and year than farms in class 2 and 40% more output

than farms in Class 3. All three classes show positive rates of technical change, but it is

significantly higher for farms in Class 3, which are catching up with farms in the two other

classes, ceteris paribus. However, as farms in Class 2 have a slightly lower rate of technical

change than farms in Class 1, their productivity gap is expected to increase, ceteris paribus

(Figure 8.5).

227. If farms in Class 2 and Class 3 produced with the technology of farms in Class 1,

they would be able to increase their productivity by respectively 41% and 30% (Table 8.5).

Overall, this would result in an average 20% increase in productivity for all cattle rearing

farms (Table A C.5 in Part 1). In practice, however, natural and human resource constraints

may limit the adoption of the best technology by other classes.



Figure 8.5. Productivity and technical change for Irish cattle rearing farms,


Source: Table 8.5.

Table 8.5. Productivity characteristics of Irish cattle rearing farms, by class







51 103 54 650*** 40 123*** 51 103***

45 022 38 789*** 33 024*** 45 022***

39 431 40 280*** 30 897*** 39 431***


- Class 1 Technology 1.066** 1.044* 0.777*

- Class 2 Technology 6.123** 0.804* 0.380*





and cattle rearing farms in Class 1 exhibit decreasing returns to scale (of about 0.9541). For

farms in Class 2 significantly increasing returns to scale (of about 1.2108) are found, and

for farms in Class 3 close to constant returns to scale (of about 1.0105) (Table A D.8.14).

1.07

0.80

2.15

0.0

0.5

1.0

1.5

2.0

2.5

0

10

20

30

40

50

60


%'000 EUR




229. Figure 8.6 summarises the various cattle rearing farm classes in terms of estimated


230. Grouped in Class 1, the most productive farms are also the most environmentally

sustainable and account for over a quarter of all farms. They are the largest operations in

terms of herd size and have the most diversified production. They have a higher share of

hired labour than average and are more likely to be managed by younger farmers. They

have higher levels of investment in new technologies and activities and higher capital

intensity than average. They receive more rural support but generate less off-farm income

than average.

231. Class 2 farms are the least environmentally sustainable and medium productivity.

They account for less than 10% of all farms. They are the largest operations in terms of

herd size, with higher stocking density, lower capital intensity but higher investment in new

technologies than average. They are less likely to be located in less favoured areas. They

are likely to be managed by older men that have higher levels of off-farm income, and

higher assets and they receive more subsidies than average.

232. Class 3 is the largest class with close to two-thirds of all farms. Class 3 farms are

the least productive and achieve average levels of environmental sustainability. They are

smaller, more specialised operations, with lower investment in new technologies than

average, and have lower assets and receive fewer subsidies than average. They have close

to average scores on most other variables.

233. In summary, Irish cattle rearing farms in the three identified technology classes

differ significantly with respect to their economic performance as well as technical

development over time. Most productive farms are comparatively larger in terms of herd

size and land endowment. These farms show a high environmental sustainability of their

production activities based on the measures used in the empirical analysis. These farms are

also highly innovative. The production intensity of cattle rearing in Ireland seems positively

correlated with herd size, ceteris paribus. The rest of the cattle rearing farms score

significantly lower on environmental sustainability and also lower on productivity.

Medium productive farms in the sector (nearly 10%) show significant innovation activities

and financial stability. However, those farms score poorly on environmental sustainability

indicators.



Figure 8.6. Multi-dimensional indices for Irish cattle rearing farms



Table 8.6. Multiple characteristics of Irish cattle rearing farms, by class





Farm structure


Herd size (LU) 0.3171 0.3831 -0.1866


Stocking density (LU per ha) -0.2425 0.2732 0.0616

Chemicals use EUR per ha) -0.3181 0.1703 0.1078


Fuel per ha (EUR per ha) 0.1140 0.2292 -0.0803

Environmental subsidies per ha (€ per ha) 0.9537 -0.4207 -0.3361

Tillage area (ha) -0.0360 0.6593 -0.0796

REPS participation (1=yes, 0=no) 0.9957 -0.4337 -0.3517


Nitrate derogation (1=yes, 0=no) -0.0466 -0.0545 0.0272


Net investment ratio (per total assets) 0.3441 0.1956 -0.1711





Insurance expenses (EUR) 0.3415 0.2210 -0.1737


Professional/Advisory expenses (EUR) 0.4549 0.0305 -0.1935

Technology

-0.6

-0.4

-0.2

0

0.2

0.4

0.6






Index 7 - Location

Index 8 - Household

Index 9 - Financial







Capital / LU (EUR per LU) 0.0151 -0.0610 0.0025

Fodder / LU (EUR per LU) -0.2402 -0.1038 0.1147

Labour / LU (AWU per LU) -0.3425 0.7149 0.0398

Materials / LU (EUR per LU) 0.1548 -0.1339 -0.0452

Diversity



Forest area (ha) 0.3444 0.3926 -0.1995

Individual

Age (years) -0.1410 0.1537 0.0366


0.1868 -0.0798 -0.0662

Gender operator (1-male, 0-female) 0.0268 0.3359 -0.0593

Marital status (1-married, 0-other) -0.3425 0.7149 0.0398

Location

Subregion 1 nuts 2 -0.1037 -0.2026 0.0721

Subregion 2 nuts 2 0.1769 0.0631 -0.0826

Subregion 3 nuts 2 -0.0203 0.3217 -0.0377

Subregion 4 nuts 2 0.1238 0.0706 -0.0616

Subregion 5 nuts 2 0.0109 0.3129 -0.0494

Subregion 6 nuts 2 -0.0330 0.0216 0.0106

Subregion 7 nuts 2 0.0192 0.0876 -0.0205

Subregion 8 nuts 2 -0.0295 -0.3712 0.0655

(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) 0.0870 0.1330 -0.0552



0.0662 -0.4076 0.0310

Soil classification (Classes 1 and 2 - Soil Group I,

Classes 3 and 4 Soil Group II, Classes 5 and 6 Soil Group Ill) -0.1503 -0.2609 0.0999

Household

Off-farm income (EUR) -0.3022 0.4396 0.0626


Part-time farming (1-yes, 0-no) -0.3053 -0.2987 0.1698

Household size (n) 0.4245 -0.3996 -0.1191

Financial

Total assets (EUR) 0.4432 0.7229 -0.2879

Total subsidies (EUR) 0.2021 0.9375 -0.2185




8.3.2. Irish “cattle other” farms

234. For Irish cattle other farms, three distinct technology classes emerge from the model


technology classes (Table 8.7), as Class 1 includes more than half of all farms.

235. “Cattle other” farms in Class 3 show a higher productivity performance


in value per year — than their counterparts in classes 1 and 2 (Figure 8.7). Farms in Class 1

produce per farm and year 40% of the output of farms in Class 3, while farms in Class 2



produce less than 20% of the output of farms in Class 3. However, the latter show a much

higher rate of technical change than farms in the other two classes. Moreover, the most

productive farms have the lowest rate of technical change, suggesting that the productivity

difference across classes will narrow down in the future, ceteris paribus (Figure 8.7).

236. If “cattle other” farms in Class 1 and Class 2 produced with the technology of farms

in Class 3, they would be able to increase their productivity by 50% and 137% respectively

(Table 8.7). Overall, this would result in an average 37% increase in productivity for Irish

“cattle other” farms, suggesting focusing efforts on productivity improvement in Class 2

farms would bring the highest benefits (Table A C.5 in Part 1). In practice, however,

natural and human resource constraints may limit the adoption of the best technology by

other classes.

237. The estimated production functions for each farm class are significant and all

“cattle other” farm classes exhibit decreasing returns to scale: of about 0.9325 for farms in

Class 1, of about 0.8835 for farms in Class 2, and of about 0.8349 for farms in Class 3

(Table A D.8.19).

Figure 8.7. Productivity and technical change for Irish “cattle other” farms,


Source: Table 8.7.

0.76

2.68

0.180.0

0.5

1.0

1.5

2.0

2.5

3.0

0

20

40

60

80

100

120

140


%'000 EUR




Table 8.7. Productivity characteristics of Irish “cattle other” farms, by class











- Class 1 Technology 0.762** 0.953** -0.046*

- Class 2 Technology 3.235*** 2.680*** 3.298**

- Class 3 Technology 0.093* -0.334* 0.180*



238. Figure 8.8 summarises the various “cattle other” farm classes in terms of estimated


239. Class 1 groups over half of “cattle other” farms, which are medium productive and

achieve slightly above average environmental sustainability. They have slightly below

average size in terms of herd, and have relatively lower investment and capital intensity

than average. They are managed by farmers with a lower agricultural training and are more

likely to be in hilly and less favoured areas. They also generate higher than average off-

farm income levels. They are close to the average of all farms for most other variables.

240. Class 2 farms are the most environmentally sustainable and the least productive.

They are less diversified, more family-driven operations, with much smaller herds and

lower stocking density. They use the most land extensive farming practices, are less capital

intensive and invest less in new technologies and activities. They have the lowest assets

and receive the least subsidies.

241. Class 3 groups the most productive farms and the least environmentally sustainable,

as they use land the most intensively, accounting for 22% of all farms. They are larger,

more diverse operations, with higher levels of investment in new technologies and

activities, and higher than average capital intensity. They have higher assets, receive more

subsidies and are managed by younger farmers with better agricultural training than

average. They are less likely to be in less-favoured areas and they have lower than average

amounts of off-farm income.

242. In summary, Irish “cattle other” farms in the three identified technology classes

differ significantly with respect to their economic performance as well as technical

development over time. In the “cattle other” sector in Ireland, over half of farms produce

with a medium productivity and an average level of environmental sustainability based on

the measures used in the empirical analysis. The most productive “cattle other” farms are

less family driven but full-time operations with a high level of innovativeness but the lowest

scores on environmental sustainability. Most environmentally sustainable “cattle other”



farms in Ireland show a low productivity with a strong family driven focus but low levels

of intensity and financial stability.

Figure 8.8. Multi-dimensional indices for Irish “cattle other” farms



Table 8.8. Multiple characteristics of Irish “cattle other” farms, by class





Farm structure


Herd size (LU) -0.1081 -0.6010 0.9871


Stocking density (LU per ha) -0.0037 -0.5512 0.6790

Chemicals use (EUR per ha) -0.0814 -0.5341 0.8422

Organic production (1=yes, 0=no) 0.0897 0.0329 -0.2528

Fuel per ha (EUR per ha) 0.0077 -0.1364 0.1476

Environmental subsidies per ha (€ per ha) 0.0888 -0.3197 0.1782

Tillage area (ha) -0.1954 -0.3128 0.8437

REPS participation (1=yes, 0=no) 0.0880 -0.2573 0.1041

Water charges (EUR) 0.0308 -0.1577 0.1188

Nitrate derogation (1=yes, 0=no) -0.0595 0.0106 0.1282


Net investment ratio (per total assets) -0.1453 -0.5776 1.0467

Share contract farming -0.1480 -0.4465 0.8939


Energy crops area (ha) -0.0144 -0.0389 0.0815

Miscellaneous income (EUR) -0.0336 -0.1304 0.2383

-1.5

-1

-0.5

0

0.5






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Insurance expenses (EUR) -0.1178 -0.6025 1.0119

E-Profit monitoring (1=yes, 0=no) -0.0796 -0.2497 0.4924

Professional/Advisory expenses (EUR) -0.0519 -0.5752 0.8223

Technology


Capital / LU (EUR per LU) -0.0218 -0.1204 0.1981

Fodder / LU (EUR per LU) -0.0178 -0.1380 0.2100

Labour / LU (AWU per LU) 0.0229 0.1377 -0.5334

Materials / LU (EUR per LU) 0.0245 -0.0755 0.0338

Diversity



Forest area (ha) -0.0296 -0.2089 0.3241

Individual

Age (years) 0.0939 -0.0345 -0.1808


0.1925 -0.2467 -0.1566

Gender operator (1-male, 0-female) -0.0542 -0.0453 0.1835

Marital status (1-married, 0-other) 0.0229 0.1377 -0.2217

Location

Subregion 1 nuts 2 0.0563 0.0954 -0.2494

Subregion 2 nuts 2 -0.0385 0.0458 0.0358

Subregion 3 nuts 2 -0.0709 0.0961 0.0513

Subregion 4 nuts 2 -0.0280 -0.1259 0.2193

Subregion 5 nuts 2 -0.0115 -0.0810 0.1257

Subregion 6 nuts 2 -0.0880 -0.1647 0.4090

Subregion 7 nuts 2 0.1010 -0.0236 -0.2108

Subregion 8 nuts 2 0.0503 0.2223 -0.3896

(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) 0.1409 -0.2520 -0.0278



0.1542 -0.0838 -0.2638


Classes 3 and 4 Soil Group II, Classes 5 and 6 Soil Group Ill) 0.0353 0.2152 -0.3454

Household



Part-time farming (1-yes, 0-no) 0.0667 0.5877 -0.8726

Household size (n) -0.1057 -0.2354 0.5368

Financial

Total assets (EUR) -0.1219 -0.6657 1.0985







8.4. Irish sheep farms


the sample, the average livestock related output per farm in 2018 was about EUR 37 000

(with a total output of about EUR 61 000). The variable cost items increased over time and

the share of hired labour significantly increases for the average Irish sheep farm over the

period investigated. The average sheep farm in the sample operated with a herd size of

about 66 LU in 2018 (of about 63 LU in 2010) and a land area of about 71 ha (compared

to about 72 ha in 2010). This results in a stocking density of about 1.23 LU per ha for 2018

(compared to a density of about 1.18 LU per ha in 2010).


components for the Irish sheep production, the output variable is total output per farm and

year and input variables are land, livestock units, capital, labour and materials. The




245. For Irish sheep farms, three distinct technology classes emerge from the model

estimates (Table A D.8.22). The individual farms are distributed relatively evenly across

the three technology classes (Table 8.9).

246. Sheep farms in Class 1 show a higher productivity performance — measured as the


than their counterparts in classes 2 and 3 (Figure 8.9). Sheep farms in Class 1 produce per

farm and year about twice the output of farms in class 2 and 3, which are similar levels of

productivity. However, while farms in Class 2 show a significant positive technical change,

technical change is negative for farms in both Class 1 and Class 3. As a result, farms in

Class 2 are reducing their productivity gap with farms in Class 1, while farms in Class 3

are increasing their productivity gap with both Class 1 and Class 2 farms, ceteris paribus

(Figure 8.9).

247. If sheep farms in Class 2 and Class 3 produced with the technology of farms in

Class 1, they would be able to increase their productivity by about 45% each, indicating

that technologies in Class 2 and Class 3 are relatively close (Table 8.9). Overall, this would

result in an average 26% increase in productivity for the Irish sheep farms (Table A C.5 in

Part 1). In practice, however, natural and human resource constraints may limit the

adoption of the best technology by other classes.



Figure 8.9. Productivity and technical change for Irish sheep farms,


Source: Table 8.9.

Table 8.9. Productivity characteristics of Irish sheep farms, by class











- Class 1 Technology -0.807* -0.753* -0.273*

- Class 2 Technology 1.216* 0.257* -0.027*

- Class 3 Technology 1.085* -1.113* -1.058**



248. The estimated production functions for each farm class are significant and sheep

farms in Class 1 and Class 2 exhibit decreasing returns to scale of about 0.9426 and 0.9424,

respectively. Sheep farms in Class 3, however, show significantly increasing returns to

scale of about 1.2231 (Table A D.8.24).

249. Figure 8.10 summarises the various sheep farm classes in terms of estimated


-0.81

0.26

-1.06

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0

10

20

30

40

50

60

70

80

90


%'000 EUR




250. Class 1 farms are the most productive farms (about a quarter of all farms) and the

least environmentally sustainable, based on the variety of indicators used to derive the

environmental sustainability score. They are the largest operations in terms of herd size and

most diversified. They have the highest investment in new technologies and activities than

and are the most capital intensive farms. They are managed by younger farmers, with a

higher education in farming, and are the least likely to be located in less-favoured or hilly

areas. They have higher asset levels, receive more subsidies and generate less off-farm

income than average.

251. Class 2 groups the least productive farms that are more environmentally sustainable

than average and account for 38% of all farms. They are the most likely to be located in

less favoured areas and are endowed with the lowest levels of asset and subsidies, but

receive more rural support. They are more likely to be managed by a man, to be part-time

farms, and to generate high levels of off-farm income.

252. Class 3 groups the most environmentally sustainable farms, which account for 27%

of all farms, and achieve productivity levels that are slightly higher than the lowest. They

are smaller, more specialised and family-driven operations. They are less capital intensive

than average and have lower investment in new technologies and activities than average.

They are the least capital and labour intensive of all farms. Their levels of assets and

subsidies is close to the lowest in Class 2.

253. In summary, Irish sheep farms in the three identified technology classes differ


over time. Most productive farms are mainly full-time operations, which are non-family

driven and highly innovative. These farms produce with a lower environmental

sustainability than the sample average and a relatively high capital intensity. They score

high on financial stability and liquidity. More than half of the farms in the Irish sheep sector

produce with lower productivity but score significantly higher on environmental

sustainability based on the measures used in the empirical analysis. About 13% of all farms

were found to operate with a very low productivity and innovativeness. However, these

farms, which are mainly part-time with lower input intensity, score highest on

environmental sustainability measures.



Figure 8.10. Multi-dimensional indices for Irish sheep farms



Table 8.10. Multiple characteristics of Irish sheep farms, by class





Farm structure


Herd size (LU) 0.6588 -0.3356 -0.1002



Chemicals use (EUR per ha) 0.1147 -0.2759 0.2036



Environmental subsidies per ha (EUR per ha) -0.0704 0.1235 -0.0784

Tillage area (ha) 0.4251 -0.2177 -0.0635

REPS participation (1=yes, 0=no) 0.0617 0.0136 -0.0552


Nitrate derogation (1=yes, 0=no) -0.0905 0.0346 0.0254










-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Technology


Capital / LU (EUR per LU) 0.2621 0.1211 -0.2986

Fodder / LU (EUR per LU) 0.0478 0.1614 -0.1961

Labour / LU (AWU per LU) -0.3934 0.3303 -0.1156

Materials / LU (EUR per LU) 0.2949 0.1246 -0.3242

Diversity



Forest area (ha) 0.2288 -0.0256 -0.1273

Individual

Age (years) -0.1215 0.0453 0.0353


-0.1953 0.0404 0.0897

Gender operator (1-male, 0-female) -0.0695 0.1280 -0.0835

Marital status (1-married, 0-other) -0.3934 0.1280 -0.2857

Location

Subregion 1 nuts 2 0.0708 0.0396 -0.0877

Subregion 2 nuts 2 -0.0889 -0.0106 0.0703

Subregion 3 nuts 2 0.3702 -0.1437 -0.1019

Subregion 4 nuts 2 0.1089 -0.1025 0.0312

Subregion 5 nuts 2 0.0072 -0.0034 -0.0013

Subregion 6 nuts 2 0.0099 -0.0857 0.0805

Subregion 7 nuts 2 -0.2719 0.3194 -0.1425

Subregion 8 nuts 2 -0.2485 -0.0415 0.2086

(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) -0.2277 0.0042 0.1112



-0.1587 0.1196 -0.0369


Classes 3 and 4 Soil Group II, Classes 5 and 6 Soil Group Ill) -0.2022 0.2037 -0.0716

Household


Rural support (EUR) -0.1855 0.2484 -0.1282

Part-time farming (1-yes, 0-no) -0.3985 0.3644 -0.1034

Household size (n) 0.3295 -0.5147 0.3024

Financial

Total assets (EUR) 0.9000 -0.3065 -0.2913





9. Italy

9.1. Italian crop farms

254. In Italy, the analysis applies to a sample of crop farms covering the period 2008 to

2015. The northern part of Italy produces primarily grains, soybeans, meat, and dairy

products, while the south specialises in fruits, vegetables, olive oil, wine, and durum wheat.



255. Italian agriculture is characterised by small-sized farms where more than 50% of

the agricultural holdings have less than 5 ha with an average farm size of about 12 ha (EU28

average of about 16 ha). Farms are managed by relatively old farmers with only about 5%

below 35 years of age (European Commission, 2014). Italy’s agricultural production is very

diversified, the majority of output stemming from vegetables and horticultural products,

fruits and wine, milk and cereals. Other important products relate to cattle and pig

production, poultry and forage plants. In 2015 the major share of agricultural utilised area

devoted to crops was used for the production of forage, wheat and olive trees (in total about

5.2 million ha), followed by maize, grapevine, citrus and fruit trees, oil seeds and

vegetables production. In addition, minor shares of land are used to grow barley, oats,

potatoes, and legumes. Wheat and maize production are the strongest in the North of Italy

whereas citrus and fruit trees as well as olive trees and tomato production dominate in the

South and the Islands.

256. According to Table A.B.9.1, which summarises descriptive statistical measures for

the Italian sample of crop farms, the average crop output per farm did not significantly

increase over the period 2008-15. The average farm size in terms of land endowment did

not change significantly and the variable input costs only slightly increased. The labour

structure with about 30-35% hired labour remained more or less stable throughout the

period, however, an increase in the share of family labour is observed in the last years.

Finally, the probability for organic crop farming has been constant throughout the eight

years considered.


components for the Italian crop production, the output variable is the total output per farm

and year and the inputs variables are land, capital, labour, chemicals, and materials. The

Class identification vector for the Italian case study consists of all seven indices as outlined

in Table A.B.9.2).

258. Crop farms are distributed unevenly across the three classes, with Class 1

comprising over half the farms in the sample (about 51.5%), Class 2 about 7% of farms and

Class 3 about 41.5% of all farms (Figure 9.1).

259. Farms in Class 1 show the highest productivity performance — measured as the


followed by crop farms in Class 3 and Class 2. The first class produces nearly three times

more per farm and year than those farms in Class 2 (Figure 9.1).



Figure 9.1. Productivity and technical change for Italian crop farms, by class, 2008 to 2015

Source: Table 9.1.


would be able to significantly increase their productivity by more than 50% (to about

EUR 26 226 per year), those in Class 3 by about 33% (Table 9.1). Overall, this would result

in a 12% increase in the average productivity of crop farms in Italy (Table A C.5 in Part 1).

261. Despite higher productivity performance in Class 1, technical change slightly

decreases (by about -0.7% per year), while crop farms in the other two classes experience

a positive rate of technical change (of about 1.5% per year for Class 3 farms and about

1.8% for Class 2 farms). Hence, despite crops farms in Class 2 and Class 3 being far less

productive than their counterparts in the first class, they seem to catch up in their production

performance based on a significant positive technical change development over time

(Table 9.1).

262. Significantly decreasing returns to scale were found for farms in Class 2 (of about

0.9163), slightly decreasing returns to scale for farms in Class 3 (of about 0.9612), and

increasing returns to scale of about 1.0512 for farms in Class 1 (Table A D.9.4).

Understanding the determinants of decreasing returns to scale would require further

investigation. Crop farms in these classes should remove those factors that are limiting their

productivity.

-0.675

1.779

1.501

-1

-0.5

0

0.5

1

1.5

2

0

5

10

15

20

25

30

35

40

45

50

Class 1 (51.5%) Class 2 (7%) Class 3 (41.5%)

%EUR '000




Table 9.1. Productivity characteristics of Italian crop farms, by class



Number of observations (% of sample farms) 10 756 (51.5%) 1 432 (7%) 8 659 (41.5%)



Productivity level (EUR per year.) 1











while Table 9.2 contains the estimates for variables used to construct the indices. Nearly

all dimensions considered to distinguish crop farms showed to be statistically significant

(Table A D.9.2).

264. Farms in Class 1 are the most productive operations, but show a negative annual

technical change rate over the period investigated. These crop farms have the lowest share

of family labour in the sample and operate with a higher than average farm size and still

experiencing increasing returns to scale. Farms in Class 1 are most probably non-single

owners. They score slightly lower on environmental sustainability than the average crop

farm in Italy (depending on the selected indicators used: chemicals used, organic

production, and environmental subsidies received). These farms show the highest share of

rented land and a slightly higher net investment rate than the average crop farm. They are

more likely to co-operate, have the highest share of irrigated land but are least likely to be

engaged in agritourism. They have a high capital and material intensity, are most

specialised and are least likely to diversify into non-agricultural, e.g. forestry, production

etc. The farm managers are younger and better educated than the average crop farmer in

Italy, and off-farm income is of average importance (Table 9.2).



Figure 9.2. Multi-dimensional indices for Italian crop farms



265. Crop farms in Class 2 represent about 7% of the total sample and are the least

productive with the lowest but a considerable positive technical change (of about 1.8% per

year). They have the highest share of family labour across all crop farms but are also the

smallest in terms of land size exhibiting significantly decreasing returns to scale (of about

0.9163). Crop farms in this class have most likely a single owner and score lowest on

environmental sustainability (depending on the selected indicators used: chemicals used,

organic production, and environmental subsidies received). Those farms show a high input

intensity but are also most likely and significantly diversified in their production and have

a high probability of being engaged in forestry production. The farm manager is slightly

older than average and the farm is most likely located in less-favoured and high altitude

areas of Italy (Table 9.3).

266. Class 3 farms (about 41.5% of all crop farms) are nearly half as productive as farms

in Class 1 and show a positive rate of technical change of about 1.5% per year. For those

crop farms family labour is important and likely, a single person or family owns the farm.

Farms in this class are slightly less endowed with land than the average crop farm in Italy

and interestingly also exhibit negative scale effects (of about 0.9612). Nevertheless, those

farms operate with the highest environmental sustainability of all farms in the sample

(depending on the selected indicators used: chemicals used, organic production, and

environmental subsidies received). They have a lower than average share of rented land

and are least input intensive. However, farms in Class 3 are less diversified than the average

crop farm in Italy, their farm managers are significantly older, and these farms are less

likely to be located in less favoured areas and areas of high altitude.

267. In summary, Italian crop farms in the three identified technology classes differ

significantly with respect to their economic performance and their technical development

-1.5

-1

-0.5

0

0.5

1






Index 7 - Location






over time. Innovative farms are also more likely to be productive. Family driven farms and

also smaller farms in terms of acreage are not necessarily more environmentally

sustainable. Highly environmentally sustainable crop farms in Italy are most likely located

in lower altitudes and least likely in less-favoured areas. Capital intensive operations are

negatively correlated with the farms’ environmental sustainability, whereas diverse farms

are not necessarily operating with a higher sustainability. The productivity of Italian crop

farms is clearly linked to the farm's structure (approximated by farm size, family labour

share, and form of ownership). Finally, the results suggest that environmentally sustainable

crop farming in Italy is not clearly correlated with a lower or higher productivity in

economic terms.

Table 9.2. Multiple characteristics of Italian crop farms, by class




Class 3: Medium productive

(41.5%)

Farm structure


Land (ha) 0.0287 -0.0714 -0.0242

Form ownership: (1-self-employment, 2-legal person, 3-cooperative form)

0.0839 -0.2041 -0.0715



Organic (probability) 0.0811 -0.1071 -0.0833

Environmental subsidies (EUR per ha) 0.0259 0.0277 -0.0365




Cooperation (probability) 0.0469 -0.0641 -0.0478

Irrigated area ratio 0.1956 -0.4756 -0.1663

Agritourism (probability) -0.1055 0.3419 0.0761

Technology

Capital / labour ratio (EUR per hour) 0.0459 0.0495 -0.0648

Capital per ha (EUR per ha) 0.0534 0.1047 -0.0827

Materials per ha (hour per ha) 0.1158 0.0083 -0.1446

Total assets (EUR) 0.0005 0.0189 -0.0036

Diversity



Forestry (probability) -0.0626 0.3334 0.0243

Individual

Age (years) -0.1401 0.0221 0.1699

Education (1:primary, 2: secondary, 3: high,

4: college 1st, 5: college 2nd)

0.0688 0.0648 -0.0955

Gender (1-male, 0-female) 0.0122 -0.1770 0.0131

Location

Less Favoured Area(1 not to- 3 severely disadvantaged)

0.0831 0.2331 -0.1401

Altitude (1: <300m, 2: 300-600m, 3: >600m) 0.0716 0.2747 -0.1324

Note: 1. Deviations from sample means (=0), z-scores based, scaled values.




10. Korea

10.1. Korean rice farms

268. In Korea, the analysis applies to a sample of rice farms covering the period 2003 to

2015. The structure of agricultural production in Korea has significantly changed during

the last decades with the production of rice still accounting for approximately half of

Korea’s cultivated land area. With the development and distribution of new rice varieties,

rice production per ha increased from about 4.5 tonnes in 1995 to 5.4 tonnes in 2015. About

60% of the 1.09 million farms in Korea are engaged in rice production with the average

rice farm cultivating about 1.25 ha in 2015.

269. According to Table A D.10.1, which summarises various statistical measures for

the sample of Korean rice farms, the average rice output per farm increased between 2003

and 2015, by about 50% also due to a significant increase in land endowment per farm. The

variable cost items also increased over the period, however, the share of hired labour

decreased for the average rice farm in Korea. Total assets per farm significantly increased

and finally the probability of being engaged in part-time rice farming increased in the

period considered.


components for the Korean rice production, the output variable is total output per farm and

year whereas the input variables are land, capital, chemicals, materials, and labour. The


sustainability, innovation, diversity and location (Table A D.10.2). The individual farms

are unevenly distributed over the three identified technology classes (Figure 10.1). Class 1

covers about 58% of all rice farms, Class 2 about 33% and Class 3 about 9%.

271. Farms in Class 2, however, show a significantly higher productivity performance


in EUR per year — than their rice-producing counterparts in Classes 1 and 3. Rice farms

in Class 2 produce about 20% more per farm and year than farms in Class 1, and about

116% more per farm and year than rice farms in Class 3. If rice farms in Class 1 produced

with the technology of farms in Class 2, they would be able to increase their productivity

by about 2% (Table 10.1). However, if farms in Class 3 adopted this technology, their

productivity would decrease by -4%.

272. Rice farms in Class 2 also show a significantly positive technical change (estimated

at 1.74% per year) compared to a modest positive technical change rate for farms in Class 1

(estimated at 1.44% per year) and a negative rate in Class 3 (estimated at -1.37% per year).

Hence, although rice farms in Class 2 are already far more productive, they also seem to

increase their productivity at a higher rate based on a significant positive technical change

development, ceteris paribus (Table 10.1).



Figure 10.1. Productivity and technical change for Korean rice farms, by class, 2003 to 2015

Source: Table 10.1.

Table 10.1. Productivity characteristics of Korean rice farms, by class











- Class 1 Technology 1.438*** -0.278*** 0.406***





273. The estimated production functions are highly significant and all rice farm classes

exhibit increasing returns to scale: 1.2104 for Class 1, 1.0522 for Class 2, and 1.5505 for

farms in Class 3 (see Table A D.10.4 for complete class related elasticities). Switching

technologies could — ceteris paribus — result in higher productivity per rice farm and a

significantly higher technical change rate for most farms (especially regarding a switch to

Class 2 technology). Various indices, reflecting the different dimensions along which rice

farms can be distinguished, to robustly identify the reported farm classes: Farm structure,


individual characteristics and locational specifics. Table A D.10.2 summarises the

estimates for the various indices that were used as components for the class identifying

1.44 1.74

-1.37

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

0

1 000

2 000

3 000

4 000

5 000

6 000

7 000


%EURProductivity (EUR per year) Technical change (% per year)



vector in the latent class estimation. Most of the indices considered showed significant

estimates. To decide on the empirically most appropriate number of classes to be estimated,

various statistical quality tests have been performed (most prominently the Akaike

Information Criteria AIC) which holds for all country cases considered in this study.

274. Figure 10.2 summarises the various rice farm classes in terms of estimated indices,


275. As discussed earlier, rice farms in Class 2 are the most productive and exhibit the

highest positive technical change rate per year. These farms have the lowest share of family

labour (i.e. a significantly lower family per hired labour share than the average rice farm in

the sample) and a significantly above average acreage size, with slightly increasing returns

to scale. Rice producing farms in Class 2, however, score relatively low on environmental

sustainability indicators (such as fertiliser and chemicals use). On the other hand, these

farms show the highest scores on innovation and commercialisation with a significantly

above average investment rate and rented land share. Class 2 farms show a higher than

average capital per labour intensity, while using more capital per land than the average rice

farm in Korea based on high levels of total assets endowment. These rice farms are less

diversified, and their managers are younger than the average rice farmer in Korea. Finally,

rice farms in Class 2 are most likely located in more environmentally beneficial areas and

off-farm income is important for those farms (Table 10.2).

276. Rice farms in Class 1 are nearly as productive as Class 2 rice farms, but show a

lower, positive technical change per year. Family labour is important for Class 1 farms,

which are smaller than the average rice farm in Korea in terms of land endowment. Rice

farms in Class 1 should significantly increase the size of their production operations due to

measured economies of scale. These farms are found to have above average environmental

sustainability performance based on the various indicators used (such as chemical use per

ha). However, rice farms in Class 1 invest slightly less than the average rice farm in Korea

and show a lower than average rented land share. Their capital intensity is still above

average, and they employ most capital per land. These rice farms are, however, the least

specialised and their farm managers are older than average. Finally, rice farms in Class 1

are likely located in less-favoured areas and show the lowest off-farm income of all rice

farms in Korea (Table 10.2).

277. Class 3 farms are the least productive and show a significant negative average

technical change rate per year. Family labour is most important for those rice farms, which

are considerably smaller than the average rice farm in Korea in terms of land endowment.

Rice farms in Class 3 should increase the size of their production operations to become

more profitable due to identified positive economies of scale (of around 1.55). However,

these farms are found to be among the most environmentally sustainable rice producers

based on various indicators used (such as fertiliser and chemicals use per ha). Rice farms

in Class 3 invest far less than the average rice farm in Korea and rely less on rented land.

Their capital intensity is the lowest of all rice farms and consistently with this notion is the

finding that farms in Class 3 employ least capital per land. The diversification of activities

in these rice farms is about average and their farm managers are of average age. Finally,

rice farms in Class 3 can be located in environmentally less-favoured detrimental areas but

also more favourable areas (Table 10.2).



Figure 10.2. Multi-dimensional indices for Korean rice farms



278. In summary, Korean rice farms in the three identified technology classes differ


over time. Family driven farms and comparatively smaller farms show a higher

environmental sustainability based on the measures used in the empirical analysis. Highly

environmentally sustainable rice farms most likely show a more diverse Farm structure, are

most likely located in less-favoured areas, and have a lower than average capital intensity.

Farms’ capital intensity is positively correlated with land endowment, while farms’

productivity is correlated with acreage size and the share of hired labour. Finally, Korean

rice farms producing more environmentally sustainably can also exhibit a very high

productivity, ceteris paribus.

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6



Index 3 -Innovation/Commercialisation


Index 5 - Diversity


Index 7 - Location

Index 8 - Household






Table 10.2. Multiple characteristics of Korean rice farms, by class





Farm structure


Land (ha) -0.0952 0.2801 -0.4031


Inorganic fertiliser (kg per ha) -0.1289 0.2780 -0.1741

Organic fertiliser (kg per ha) -0.2097 0.3984 -0.0855

Agri-chemicals Dry (kg per ha) -0.0196 0.0309 0.0151

Agri-chemicals liquid (kg per ha) -0.1901 0.4150 -0.2755



Share land rented -0.0391 0.1119 -0.1551

Technology

Capital / labour ratio (EUR per AWU) 0.0598 0.0226 -0.4763

Capital per land (EUR per ha) 0.0521 0.0465 -0.5134

Total assets (EUR) -0.0664 0.2027 -0.3083

Diversity


Production diversity (yc/ΣY) -0.3562 0.6468 -0.0344

Individual

Age (years) 0.1303 -0.2526 0.0713

Education (1-high school, 2-college, 3-graduate) -0.0512 0.0862 0.0197

Location

Suitability Indicator (points) 0.1983 -0.3425 -0.0452

Household

Female/male labour ratio 0.0821 -0.0929 -0.1979

Family/Total household member ratio -0.2896 0.4419 0.2802

Off-farm income (share) -0.3565 0.4873 0.5529





11. Norway

279. In Norway, the analysis applies to a sample of dairy farms, a sample of crop farms

and a sample of livestock farms, all covering the period 2005 to 2016. Agricultural

production in Norway is characterised by its unique environmental and climatic conditions.

A relatively long winter season as well as much lower yields linked to a small-scale

structure and generally high level of production cost (Norwegian Ministry of Agriculture

and Food, 2017). Only 3% of the total land area is used for agricultural production with

livestock farming mainly in the Western and Northern parts of the country and crop related

production activities mainly in the south-eastern parts of the country. In 2017 there were

about 40 000 farms in Norway cultivating nearly 1 million ha. The average land

endowment per farm has been about 24 ha (Hemmings 2016, Norwegian Ministry of

Agriculture and Food 2017).

280. The main agricultural income activities in Norway relate to dairy, crops and

livestock production whereas the majority of farms is engaged in a mix of activities. In

2017, about 26% of all farms in Norway were cultivating around 10 to 20 ha per farm,

about 17% around 5 to 10 ha, and also about 17% around 20 to 30 ha of land. Only about

11% of all farms in Norway use more than 50 ha for production. About 60% of all farms

are engaged in some kind of livestock production, about 35% in crop related activities and

about 15% in dairy production. 30% of all farms active in milk production in 2017 have

30 or more cows, around 26% 20-29 cows, and around 20% of these farms produce with a

dairy herd size of about 15-19 cows. Around 30% of all livestock farms use around 20-

49 sheep units per herd, 28% use 50-99 sheep units per herd, and around 15% more than

150 sheep units per herd. Finally, in 2017, 24% of all farms engaged in pig production had

up to 19 sows for breeding, 17% around 40-59 units and about 23% more than 100 sows.

In terms of land use, main crops cultivated in Norway in 2017 are wheat, oats, and barley

as well as potatoes and vegetables (Statistics Norway 2018).

11.1. Norwegian dairy farms

281. According to Table A D.11.1, which gives descriptive statistical measures for the

sample of Norwegian dairy farms, the average milk output per farm increased over the

period 2005-16, partly due to a significant increase in herd size. The variable cost items

increased over the period, as did the share of hired labour for the average Norwegian dairy

farm. The stocking density remained more or less at the same level whereas the probability

of being engaged in organic dairy production slightly increased.


components for the Norwegian dairy production, the output variable is total output per farm

and year, and the input variables are land, cows, capital, materials, and labour. The

Class identification component in each estimation is based on the indices related to

structure, environmental sustainability, innovation, technology, diversity, individual and

household characteristics, location and financial aspects (Table A D.11.2).

283. For Norwegian dairy farms, three distinct technology classes emerge from the

model estimates (Table A D.11.2). The individual farms are distributed unevenly across

the three identified technology classes (Table 11.1). Class 1 covers about 64.5% of all dairy

farms, Class 2 about 16.3%, and Class 3 about 19.2%.

284. Farms in Class 1 show a higher productivity performance — measured as the




than their milk-producing counterparts in Classes 2 and 3 (Figure 11.1). Dairy farms in

Class 1 produce about 100% more per farm and year than farms in Class 2 but only about

5% more per farm and year than dairy farms in Class 3.

Figure 11.1. Productivity and technical change for Norwegian dairy farms,


Source: Table 11.1.


would be able to increase significantly their productivity by about 30%, but if they did the

same, farms in Class 3 would experience a slight decline in productivity (Table A C.4 in

Part 1).

286. Dairy farms in Class 1 also show a significantly positive technical change

(estimated at 2.52% per year) compared to a modest positive technical change rate for farms

in Class 3 (estimated at 0.49% per year) and a negative rate in Class 2 (estimated at -0.41%

per year). Hence, although dairy farms in Class 1 are already far more productive, they also

seem to increase their productivity at a higher rate given the significant positive technical

change development ceteris paribus (Table 11.1).

287. The estimated production functions are highly significant and all dairy farm classes

exhibit significantly increasing returns to scale: 1.0918 for Class 1, 1.2371 for Class 2, and

1.1413 for farms in Class 3 (see Table A D.11.4 for complete class related elasticities).

Switching technologies could — ceteris paribus — result in higher productivity per farm

and also a significantly higher technical change rate for some farms (especially regarding

a switch to Class 1 technology). Various indices, reflecting the different dimensions along

which dairy farms can be distinguished, are used to robustly identify the reported farm


characteristics, degree of diversity, individual and household characteristics and locational

as well as financial specifics. Table A D.11.3 summarises the estimates for the various

indices that were used as components for the class-identifying vector in the latent class

estimation. Most of the indices considered showed significant estimates. To decide on the

empirically most appropriate number of classes to be estimated, various statistical quality

tests have been performed (most prominently the Akaike Information Criteria AIC) which

holds for all country cases considered in this study.

2.52

-0.41

0.49

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0

10

20

30

40

50

60

70

80

90


%EUR '000




Table 11.1. Productivity characteristics of Norwegian dairy farms, by class



Number of observations (% of sample farms) 3 583 (64.5%) 902 (16.3%) 1 064 (19.2%)













288. Figure 11.2 summarises the various dairy farm classes in terms of estimated


289. As discussed earlier, dairy farms in Class 1 are the most productive and exhibit the

most significant positive technical change rate per year. They have a relatively medium

share of family labour and herd size, and a lower than average acreage size, but still

experience increasing returns to scale. Dairy farms in Class 1 score slightly lower on

environmental sustainability indicators (such as stocking density, chemicals use per ha and

probability of producing organic) compared to the average Norwegian dairy farm. These

farms show medium scores on innovation and commercialisation based on indicators such

as investment rate, rented land share, and income generated by biofuels production. Class 1

farms show a slightly lower than average capital per labour and capital per cow intensity,

while using slightly more than average fodder per cow. These dairy farms are least

diversified, their managers are younger than the average dairy farmer in Norway. Finally,

dairy farms in Class 1 generate only a low amount of off-farm income and operate with a

medium assets endowment (Table 11.2).



Figure 11.2. Multi-dimensional indices for Norwegian dairy farms



290. Dairy farms in Class 3 are almost as productive as farms in Class 1, but show a

significantly lower technical change per year. Hired labour is very important for those

farms, which are significantly larger than the average dairy farm in Norway in terms of

herd size and land endowment. Dairy farms in Class 3 should significantly increase the size

of their production operations given the measured economies of scale. These farms are

found to be the most environmentally sustainable based on indicators used (such as

stocking density, chemicals use per ha and probability of producing organic). However,

dairy farms in Class 3 score average on innovation and commercialisation criteria such as

net investment and share of land rented. Their capital intensity is the highest of all dairy

farms, and they employ the highest rate of capital per cow. The specialisation of these dairy

farms is medium and their farm managers are slightly older than the average dairy farmer.

Finally, dairy farms in Class 3 are likely located in advantageous regions for dairy

production in Norway (Table 11.2).

291. Class 2 farms are the least productive of all and show a negative technical change

rate per year. Family labour is most important for those dairy farms, which are considerably

smaller than the average dairy farm in Norway in terms of herd size and land endowment.

Dairy farms in Class 2 should increase the size of their production operations to become

more profitable due to significantly positive economies of scale. However, these farms are

found to have an average environmental sustainability performance based on indicators

used (such as stocking density, chemicals use per ha and probability of producing organic).

Dairy farms in Class 2 show an average investment level but higher than average share of

land rented. Their capital intensity is the lowest of all dairy farms in Norway. They are

highly diversified and their farm managers are of average age. Finally, dairy farms in

Class 2 are less likely located in favourable areas and operate with a lower than average

assets endowment (Table 11.3).

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8





Index 5 - DiversityIndex 6 – Individual









292. In summary, Norwegian dairy farms in the three identified technology classes differ


over time. Family driven farms and comparatively smaller farms not necessarily show a

higher environmental sustainability based on the measures used in the empirical analysis.

Highly environmentally sustainable dairy farms most likely show a more diverse structure

and are most likely located in a dairy production advantageous area. Farms’ capital

intensity is positively correlated with herd size, while farms’ productivity is correlated with

herd size and the share of hired labour. Finally, a strong share (nearly 20%) of more

sustainably producing Norwegian dairy farms also exhibit a very high productivity, ceteris

paribus.

Table 11.2. Multiple characteristics of Norwegian dairy farms, by class





Farm structure

Family/hired labour ratio -0.0018 0.1554 -0.1255

Herd size (LU) 0.0286 -0.4126 0.2536

Land (ha) -0.1056 -0.0661 0.4117




Organic (probability) -0.2023 -0.0801 0.7492

Environmental subsidies (EUR per ha) -0.1086 0.0344 0.3365




Contract farming (prob) -0.0136 0.1641 -0.0933

Technology

Capital / labour ratio (EUR per hour) -0.0647 -0.0449 0.2559

Capital per cow (EUR per LU) -0.0458 0.1231 0.0476

Fodder per cow (EUR per LU) 0.0967 -0.0777 -0.2599

Diversity



Individual

Age (years) -0.0076 0.0143 0.0134

Gender (0-male, 1-female) -0.0479 0.1857 0.0041

Location

Dairy zone (0-10, increasingly disadvantaged) 0.1245 0.1031 -0.5065

Forest income (EUR) -0.1403 -0.0768 0.5377

Household

Female/male labour ratio -0.0461 0.1238 0.0502


Financial

Total assets (EUR) -0.0198 -0.2412 0.2811


Note: LU: Livestock Unit. 1. Deviations from sample means (=0), z-scores based, scaled values.




11.2. Norwegian crop farms

293. According to Table A D.11.6, which summarises descriptive statistical measures

for the sample of Norwegian crop farms , the crop output of the average Norwegian crop

farm increased over the period investigated 2005 to 2016, mainly due to productivity

improvements as, for example, the acreage per farm only slightly increased over the

11 years considered. The variable input costs also increased, and the labour structure

changed to some extent with the share of hired labour higher on average per farm than the

share of family labour in 2016. Finally, the probability of being engaged in organic

production slightly increased.


components for the Norwegian crop production, the output variable is total output per farm

and year per farm and year and the input variables are land, capital, chemicals, materials.

The Class identification component in each estimation is based on the indices related to

structure, environmental sustainability, innovation, technology, diversity, individual and

household characteristics, location and financial aspects (Table A D.11.7).

295. For Norwegian crop farms, three distinct farm classes of crop farms emerge from

the model estimates (Table A D.11.7). The individual farms are distributed relatively

unevenly across the three identified technology classes (Table 11.3). Class 1 covers about

42% of all crop farms, Class 2 about 45%, and Class 3 about 14%.



in EUR per year — than their crop-producing counterparts in Classes 2 and 3 (Figure 11.3).

Crop farms in Class 1 produce about 1.6 times more per farm and year than farms in Class 2

and about 1.5 times more per farm and year than crop farms in Class 3.

Figure 11.3. Productivity and technical change for Norwegian crop farms,


Source: Table 11.3.

297. Assuming that crop farms in Class 2 produced with the technology of farms in

Class 1, they would be able to significantly increase their productivity by nearly 10%, and

farms in Class 3 would increase their productivity even by nearly 50% (Table 11.3). If all

2.28

1.47

0.93

0.0

0.5

1.0

1.5

2.0

2.5

0

10

20

30

40

50

60


%EUR '000




farms adopted the technology of the most productive farms in Class 1, the overall

productivity of crop farms in Norway would increase by less than 10% (Table A C.5 in

Part 1).

298. Crop farms in Class 1 also show a significantly positive technical change (estimated

at 2.28% per year) compared to a still considerable positive technical change rate for farms

in Class 2 (estimated at 1.47% per year) and in Class 3 (estimated at 0.93% per year).

Hence, although crop farms in Class 1 are already more productive, they also seem to

increase their productivity at a higher rate based on a significant positive technical change

development ceteris paribus (Table 11.3).

Table 11.3. Productivity characteristics of Norwegian crop farms, by class













- Class 3 Technology -6.264*** 2.869*** 0.929***



299. The estimated production functions are highly significant, and all crop farm classes

exhibit (significantly) increasing returns to scale: 1.3096 for Class 1, 1.1674 for Class 2,

and 1.7737 for farms in Class 3 (see Table A D.11.9 for complete class related elasticities).

Switching technologies could ceteris paribus result in higher productivity per farm and a

significantly higher technical change rate for the majority of farms (especially regarding a

switch to Class 1 technology). Various indices, reflecting the different dimensions along

which dairy farms can be distinguished, are used to robustly identify the reported farm


characteristics, degree of diversity, individual and household characteristics as well as

locational and financial specifics. Table A D.11.9 summarises the estimates for the various

indices that were used as components for the class-identifying vector in the latent class

estimation. Most of the indices considered showed significant estimates. To decide on the

empirically most appropriate number of classes to be estimated, various statistical quality

tests have been performed (most prominently the Akaike Information Criteria AIC) which

holds for all country cases considered in this study.





301. As discussed earlier, crop farms in Class 1 are most productive and exhibit a

significant positive technical change rate per year. They show a lower than average share

of family labour (i.e. a lower family per hired labour ratio than the average farm in the

sample) and a below average acreage size, and therefore experience increasing returns to

scale. Crop farms in Class 1 score relatively low on environmental sustainability indicators

(such as chemicals use per ha and probability of producing organic). These farms show

above average scores on innovation and commercialisation with a slightly higher than

average investment rate and rented land share. Class 1 farms show a significantly higher

than average capital per labour intensity compared to the average crop farm in Norway.

Their capital per land intensity is slightly above average based, however, on relatively low

levels of total assets endowment. These crop farms are less diversified and their managers

are older than the average crop farmer in Norway. Finally, crop farms in Class 1 generate

below average off-farm income and receive below average subsidies (Table 11.4).

Figure 11.4. Multi-dimensional indices for Norwegian crop farms



302. Crop farms in Class 3 are less productive and show a lower than average technical

change per year. Hired labour is important for those farms, which are, however,

considerably smaller than the average crop farm in Norway in terms of land endowment.

Crop farms in Class 3 should significantly increase the size of their production operations

given the measured economies of scale. These farms are found to be less environmentally

sustainable based on indicators used (such as chemicals use per ha and probability of

producing organic). However, crop farms in Class 3 invest more than the average crop farm

in Norway and show a higher than average share of land rented. Their capital intensity is

the highest of all crop farms, and they employ the highest rate of capital per labour. The

specialisation of these crop farms is medium and their farm managers are younger than the

average crop farmer. Finally, crop farms in Class 3 generate about average off-farm income

but receive significantly below average subsidies (Table 11.4).

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8














303. Class 2 farms are the least productive but only slightly less productive than their

counterparts in Class 3. They show, however, a considerable positive technical change rate

per year. Family labour is very important for those crop farms, which are considerably

larger than the average crop farm in Norway in terms of land endowment. Crop farms in

Class 2 should increase the size of their production operations to become more profitable

given the estimated positive economies of scale (which are, however, considerably smaller

than for their counterparts in other classes). These farms are found to be below average

environmentally sustainable based on indicators used (such as chemicals use per ha and

probability of producing organic). Crop farms in Class 2 invest less than the average crop

farm in Norway and show a lower than average share of land rented. Their capital intensity

is relatively low and consistently with this notion is the finding that farms in Class 2 employ

the lowest rate of capital per labour. These crop farms are medium diversified and finally

crop farms in Class 2 generate well above average levels of off-farm income and receive

significantly above average levels of subsidies (Table 11.4).

304. In summary, Norwegian crop farms in the three identified technology classes differ


over time. Innovative crop farms in Norway can be expected to be also more productive

than the average of their peer group. Family driven farms and comparatively smaller farms

show a lower environmental sustainability based on the measures used in the empirical

analysis. Highly environmentally sustainable crop farms do not necessarily show a more

diverse structure but likely a lower than average capital intensity. Farms’ productivity is

obviously not per se positively correlated with land endowment but with the share of hired

labour. Finally, Norwegian crop farms producing more environmentally sustainably show

a (relatively) lower productivity, ceteris paribus.



Table 11.4. Multiple characteristics of Norwegian crop farms, by class





Farm structure


Land (ha) -0.2253 0.4031 -0.6271






Net investment ratio (per total assets) 0.0179 -0.0889 0.2357


Contract farming (prob) 0.0261 -0.0901 0.2148

Technology

Capital / labour ratio (EUR per hour) 0.3527 -0.4153 0.2766

Capital per land (EUR per ha) 0.0801 -0.0371 -0.1239

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) 0.1838 0.0083 -0.5902

Production diversity (yc/ΣY) 0.1781 0.0103 -0.5793

Individual

Age (years) 0.2145 -0.1006 -0.3282

Gender (0-male, 1-female) -0.1175 0.0185 0.2996

Location

Dairy zone (0-10, increasingly disadvantaged) 0.2071 -0.4639 0.8815

Meat zone (0-5, increasingly disadvantaged) 0.2206 -0.4887 0.9213

Forest income (EUR) -0.0549 0.0266 0.0813

Household

Female/male labour ratio -0.0255 -0.0864 0.3604


Financial

Total assets (EUR) -0.2546 0.2838 -0.1708

Total subsidies (EUR) -0.2431 0.3052 -0.2527

Note: 1. Deviations from sample means (=0), z-scores based, scaled values.


11.3. Norwegian cattle farms

305. According to Table A D.11.11, which shows descriptive statistical measures for the

sample of Norwegian cattle farms, the average livestock output of these farms significantly

increased over the period 2005 to 2016, partly due to an increase in livestock herd size. The

variable cost items increased over the period, as well as the share of hired labour for the

average Norwegian cattle farm, whereas the share of family labour remained more or less

constant. The stocking density increased, and the probability of being engaged in organic

production slightly decreased.


components for the Norwegian beef production, the output variable is total output per farm

and year and the input variables are land, capital, livestock units, fodder, materials, fuel,

and labour. The Class identification component in each estimation is based on the indices



related to structure, environmental sustainability, innovation, technology, diversity,

individual and household characteristics, location and financial aspects (Table A D.11.12).

307. For Norwegian specialised cattle farms, two distinct technology classes emerge

from the model estimates (Table A D.11.12). The individual farms are distributed very

unevenly across the two identified technology classes (Table 11.5). Class 1 covers about

81% of all livestock farms while Class 2 contains the remaining 19%.

308. Cattle farms in Class 2 show a slightly higher productivity performance


in EUR per year — than their milk-producing counterparts in Class 1 (Figure 11.5). They

produce about 13% more per farm and year than farms in Class 1.

Figure 11.5. Productivity and technical change for Norwegian cattle farms,


Source: Table 11.5.

309. The empirical findings paradoxically suggest that only farms in the slightly more

productive Class 2 could increase their productivity further, assuming they would switch

to Class 1 technology. This might be due to the unique production environment for some

cattle farms in Norway and the minor difference being found between class related

productivity levels. However, if livestock farms in Class 2 produced with the technology

of farms in Class 1, they would be able to further increase their productivity by about 24%.

310. Cattle farms in Class 1 show a positive technical change (of about 1.566% per year)

compared to a negative technical change rate for farms in Class 2 (of about -1.19% per

year). Hence, farms in Class 1 are slightly less productive, they nevertheless seem to

increase their productivity at a higher rate given the significant positive technical change

development ceteris paribus (Table 11.5).

1.56

-1.20

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

28

29

30

31

32

33

34

35

36

Class 1 (81%) Class 2 (19%)

%EUR '000




Table 11.5. Productivity characteristics of Norwegian cattle farms, by class


Class 1 Class 2

Number of observations (% of sample farms) 1 340 (80.9%) 315 (19.1%)








- Class 2 Technology -2.821*** -1.197***

Note: * significant at 10%, ** significant at 5%, *** significant at 1%.

1. Fitted values at sample means.


311. The estimated production functions are highly significant and both livestock farm

classes exhibit increasing returns to scale: 1.0777 for Class 1, and 2.0563 for farms in

Class 2 (see Table A D.11.14 for complete class related elasticities). Switching

technologies could — ceteris paribus — result in higher productivity per farm and also a

significantly higher technical change rate a minority of farms only (especially regarding a

switch to Class 1 technology). Various indices, reflecting the different dimensions along

which cattle farms can be distinguished, are used to robustly identify the reported farm


characteristics, degree of diversity, individual and household characteristics and locational

as well as financial characteristics. Table A D.11.14 summarises the estimates for the

various indices that were used as components for the class-identifying vector in the latent

class estimation. Most of the indices considered showed significant estimates. To decide

on the empirically most appropriate number of classes to be estimated, again various

statistical quality tests have been performed (most prominently the Akaike Information

Criteria AIC) which holds for all country cases considered in this study.

312. Figure 11.6 summarises the various cattle farm classes in terms of estimated


313. As discussed earlier, cattle farms in Class 2 are slightly more productive but exhibit

a negative technical change rate per year. They have the lowest share of family labour (i.e.

a significantly lower family per hired labour ratio than the average farm in the sample) and

a significantly above average herd and acreage size, but still experience (modest) increasing

returns to scale. Cattle farms in Class 2 score relatively high on environmental

sustainability indicators (such as stocking density, chemicals use per ha and probability of

producing organic). These farms show lower than average scores on innovation and

commercialisation with a lower than average investment rate and rented land share. Class 2

farms show a higher than average capital per labour intensity, while using more capital per

livestock unit than the average cattle farm in Norway based on high levels of total assets

endowment. These farms are, however, more diversified, and their managers are younger

than the average cattle farmer in Norway. Finally, cattle farms in Class 2 generate slightly



more off-farm income and receive a significantly higher level of subsidies than the average

cattle farm (Table 11.6).

Figure 11.6. Multi-dimensional indices for Norwegian cattle farms



314. Cattle farms in Class 1 are slightly less productive but show a positive technical

change per year. Family labour is more important for those farms, which are smaller than

the average cattle farm in Norway in terms of herd size and land endowment. Livestock

farms in Class 1 should still increase the size of their production operations given the

measured economies of scale. These farms are found to be less environmentally sustainable

based on indicators used (such as stocking density, chemicals use per ha and probability of

producing organic). However, cattle farms in Class 1 invest more than the average cattle

farm in Norway and use more rented land for production. Their capital intensity is lower

than average, and they employ the highest rate of labour per capital and livestock unit.

These cattle farms are less diversified than average and their farm managers are slightly

older than the average cattle farmer in Norway. Finally, cattle farms in Class 1 generate

slightly less off-farm income and receive a lower level of subsidies than the average cattle

farm in Norway (Table 11.6).

315. In summary, Norwegian cattle farms in the two identified technology classes differ


over time. Innovative farms are not per se more productive compared to their peer group.

Family driven farms and comparatively smaller farms show a lower environmental

sustainability based on the measures used in the empirical analysis. Highly environmentally

sustainable cattle farms most likely show a very diverse farm structure and most likely

receive a higher amount of subsidies. Farms’ capital intensity is positively correlated with

herd size, while farms’ productivity is only slightly correlated with herd size and the share

-1.2-1

-0.8-0.6-0.4-0.2

00.20.40.60.8









Class 1: Least productive (81%) Class 2: Most productive (19%)



of hired labour. Finally, Norwegian cattle farms producing more environmentally

sustainably can also exhibit a very high productivity, ceteris paribus.

Table 11.6. Multiple characteristics of Norwegian cattle farms, by class




Farm structure

Family/hired labour ratio 0.0634 -0.2697

Herd size (LU) -0.1414 0.6017

Land (ha) -0.1943 0.8266


Stocking density (LU per ha) -0.0109 0.0461

Chemicals use (EUR per ha) 0.0336 -0.1430

Organic (probability) -0.1044 0.4442

Environmental subsidies (EUR per ha) -0.0831 0.3524


Net investment ratio (per total assets) 0.0129 -0.0549

Share land rented 0.1045 -0.4446

Contract farming (prob) 0.0332 -0.1412

Technology

Capital / labour ratio (EUR per h) -0.0291 0.1236

Capital per livestock unit (EUR per LU) -0.0043 0.0187

Fodder per livestock unit (EUR per LU) 0.0068 -0.0289

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) 0.1529 -0.6507

Production diversity (yc/ΣY) 0.1571 -0.6683

Individual

Age (years) 0.0245 -0.1041

Gender (0-male, 1-female) 0.0623 -0.2649

Location

Dairy zone (0-10, increasingly disadvantaged) 0.1005 -0.4342

Meat zone (0-5, increasingly disadvantaged) 0.0957 -0.4071

Forest income (EUR) -0.0864 0.3675

Household

Female/male labour ratio 0.0317 -0.1347

Off-farm income (EUR) -0.0118 0.0503

Financial

Total assets (EUR) -0.0688 0.3028

Total subsidies (EUR) -0.1401 0.5961

Note: LU: Livestock Unit. 1. Deviations from sample means (=0), z-scores based, scaled values.




12. Sweden

316. For Sweden, the analysis applies to a sample of crop farms and dairy farms

covering the period 1997 to 2017.

12.1. Swedish dairy farms


the sample, the average milk output per farm was about EUR 230 000 in 2017 (with a total


of hired labour significantly increases for the average Swedish dairy farm. The average

Swedish dairy farm in the sample in 2017 operated with a herd size of about 118 dairy cows

and an average stocking density of about 0.69 LU per ha.


components for the Swedish dairy production the output variable is total output per farm

and year and input variables are capital, materials, fuel and labour. The Class identification


innovation, technology, diversity, individual, location, household and financial aspects

(Table A D.12.2).

319. For Swedish dairy farms, three distinct technology classes emerge from the model


technology classes (Table 12.1), with Class 2 including close to half of all farms.




classes 1 and 3, which have similar levels of productivity, produce per farm and year about

a third of the output for farms in Class 2.

321. All three farm classes show significant positive technical change, but the growth

rate for the most productive farms in Class 2 is lower than that for other classes, suggesting

that they will lose some of their productivity advantage over time, ceteris paribus.

Productivity developments between Class 1 and Class 3 farms will differ significantly as

Class 3 farms show a rate of technical change per farm and per year much higher than that

for farms in Class 1 (Figure 12.1).


Class 2, they would be able to increase their productivity by 180% and 112% respectively.

As with developments in technical change, this suggests that technologies in Class 1 and

Class 3 differ significantly, although current productivity levels are close (Table 12.1).

Overall, the adoption of Class 2 technologies by all farms would result in an average 45%

increase in productivity for Swedish dairy farms (Table A C.4 in Part 1).



Figure 12.1. Productivity and technical change for Swedish dairy farms,


Source: Table 12.1.

Table 12.1. Productivity characteristics of Swedish dairy farms, by class



Number of observations (% of sample farms) 36% 46.9% 17.1%

Prior probability of class membership .11368 .58747 .29885

Posterior probability of class membership .4173 .381617 .201082






- Class 1 Technology 1.059* 2.523*** 4.1579***

- Class 2 Technology 4.545* 0.720** 0.487*

- Class 3 Technology 2.118* 1.096*** 2.674***



323. The estimated production functions for each farm class are highly significant. Dairy

farms in Class 1 exhibit slightly increasing returns to scale of about 1.0174

(Table A D.12.4). Dairy farms in classes 2 and 3, however, exhibit decreasing returns to

scale of about 0.8429 and 0.9472, respectively.



1.06

0.72

2.67

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0

50

100

150

200

250

Class 1 (36%) Class 2 (46.9%) Class 3 (17.1%)

%EUR '000




Figure 12.2. Multi-dimensional indices for Swedish dairy farms



325. Farms in Class 1 are the least productive and most environmentally sustainable of

all dairy farms in Sweden. They are smaller operations, more specialised, with lower

investments in new technologies and lower capital intensity. They are more likely to be

operated by older men with lower education levels. They receive fewer subsidies than

average and are more reliant on off-farm income than average.

326. Class 2 groups more productive, least environmentally sustainable farms in Class 2,

which account for close to half of all dairy farms. They are larger farms, with larger assets,

which invest in new technologies and are capital intensive. They are relatively diversified

and receive higher levels of subsidies. They are more likely to be operated by women, with

higher education levels than average.

327. Class 3 includes farms that achieve a productivity level barely above the lowest

performers, and their environmental sustainability score is slightly below average. They

are smaller, more diversified operations, with lower investment than average. They are

more likely to be in less-favoured areas and to have a female operator, with higher

education levels than average.

328. In summary, Swedish dairy farms in the three identified technology classes differ


over time. Comparatively smaller farms achieve a higher environmental sustainability

based on the variables used in the empirical analysis. However, these dairy farms are

significantly less productive than other farms in the sector. Highly environmentally

sustainable dairy farms most likely produce with a lower intensity but are not necessarily

are more diverse. The capital intensity of farms seems positively correlated with herd size,

while their productivity seems correlated with herd size but not with the share of hired

labour, ceteris paribus. Almost half of Swedish dairy farms show a higher productivity

linked to higher innovativeness. However, around 17% of all dairy farms are less

-1

-0.5

0

0.5

1



Index 3 -Innovation/commercialisation



Index 7 - Location

Index 8 - Household

Index 9 - Financial






innovative than the average and achieve lower productivity and environmental

sustainability.

Table 12.2. Multiple characteristics of Swedish dairy farms, by class





Farm structure


Herd size (LU) -0.3516 0.3535 -0.2283


-0.6218 0.4300 0.1290





Fuel per LU (EUR per LU) -0.2472 0.1648 0.0681

Environmental subsidies per ha (EUR per ha) 0.0161 -0.0143 0.0053

Innovation-Commercialisation




Biofuel income (EUR) -0.0576 0.0531 -0.0243

Miscellaneous income (EUR) -0.0172 0.0110 0.0061

Technology


Capital per cow (EUR per LU) -0.3085 0.2217 0.0410

Cow per labour (LU per AWU) -0.5262 0.4830 -0.2159

Fodder per cow (EUR per LU) -0.3208 0.2201 0.0711

Materials per cow (EUR per LU) -0.3844 0.2746 0.0557

Diversity



Individual

Age (years) 0.0633 -0.0400 -0.0234

Gender (1-female, 0-male) -0.2313 0.2725 0.2682

Education -0.1942 0.2630 0.1986



Location

Subregion 1 0.0502 -0.0126 -0.0708

Subregion 2 0.0000 0.0000 0.0000

Subregion 3 0.0400 0.0077 -0.1047

Subregion 4 0.0278 -0.0208 -0.0014

Subregion 5 0.0162 0.0279 -0.1101

Subregion 6 -0.0500 0.0276 0.0294

Subregion 7 -0.0188 -0.0212 0.0970

Subregion 8 -0.0165 0.0543 -0.1136

Subregion 9 -0.0160 -0.0101 0.0609

Subregion 10 0.0500 -0.0151 -0.0636

Subregion 11 0.0055 -0.0517 0.1296

Subregion 12 -0.0549 -0.0113 0.1458

Subregion 13 -0.0543 -0.0996 0.3856

Subregion 14 -0.0315 -0.0529 0.2105

Subregion 15 -0.0361 -0.0817 0.2986

Subregion 16 -0.0401 -0.0793 0.3005

Subregion 17 0.1218 -0.0684 -0.0684

Subregion 18 0.1258 0.0294 -0.3438

Subregion 19 -0.0716 0.0833 -0.0773

Subregion 20 -0.0721 0.1033 -0.1310

Subregion 21 0.1148 -0.0645 -0.0645

Subregion 22 0.1601 -0.0899 -0.0899

Subregion 23 -0.0658 -0.0104 0.1663

Subregion 24 0.0087 -0.0185 0.0323

Subregion 25 0.0199 0.0093 -0.0672

(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) 0.0652 -0.0448 -0.0142

Less favoured area (subsidies in EUR) -0.0327 -0.2372 0.7158

Household


Spouse labour (AWU) 0.0331 -0.0712 -0.0135

Relative labour (AWU) -0.1149 0.2518 0.0429

Financial

Total assets (EUR) -0.3637 0.3992 -0.3278

Total subsidies (EUR) -0.2529 0.1448 0.1347

Equity/debt ratio 0.0013 -0.0438 0.0314




12.2. Swedish crop farms


the sample, the average total output per farm was about EUR 461 000 in 2017 (with a total

output of about EUR 97 400 in 1997). The variable cost items increased over time and the

share of hired labour significantly increases for the average Swedish crop farm. The

average crop farm in the sample operates about 115 ha over the period (from about 94 ha

in 1997 to about 157 ha in 2017 on average).


components for the Swedish crop production the output variable is total output per farm

and year and input variables are land, chemicals, labour, and materials. The






331. For Swedish crop farms, three distinct technology classes emerge from the model

estimates (Table A D.12.7). The individual farms are distributed relatively evenly across

the three technology classes (Table 12.3).



than their counterparts in classes 1 and 3 (Figure 12.3). Crop farms in Class 1 produce

three-quarters of the output per farm and year produced by farms in Class 2, while farms

in Class 3 produce only a quarter of Class 2 output.

333. The three classes of Swedish crops farms show strong, positive rates of technical

change. The highest growth rate (about 5%) is estimated for the most productive farms in

Class 2, and the lowest for the least productive farms in Class 3. This suggests that

differences in productivity between classes will further increase, ceteris paribus

(Figure 12.3).

Figure 12.3. Productivity and technical change for Swedish crop farms,


Source: Table 12.3.


would be able to almost double their productivity (Table 12.3). However, if farms in

Class 1 adopted the technology of the most productive farms in Class 2, their productivity

would decrease by 34%, suggesting that they already produce with a technology which is

best adapted to their conditions. Overall, the productivity of Swedish crop farms would

increase by 4% (Table A C.5 in Part 1). The highest productivity gains would be obtained

if all farms adopted the technology of farms in Class 1, which is based on higher capital

intensity and more intensive farming practices than the average of all Swedish crop farms

(Table 12.4).

3.2

4.9

2.1

0

1

2

3

4

5

6

0

50

100

150

200

250

class 1 (25.1% of farms) class 2 (34.3% of farms) class 3 (40.6% of farms)

%EUR '000




Table 12.3. Productivity characteristics of Swedish crop farms, by class











- Class 1 Technology 3.153*** 2.289** 4.280***


- Class 3 Technology 2.028* 1.132* 2.145**




crop farm classes exhibit constant or decreasing returns to scale: 0.9208 for farms in

Class 1, 1.0142 for farms in Class 2, and 0.9427 for farms in Class 3 (Table A D.12.9).



Figure 12.4. Multi-dimensional indices for Swedish crop farms



-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6



Index 3 -Innovation/commercialisation



Index 7 - Location

Index 8 - Household

Index 9 - Financial






337. Class 1 farms achieve productivity levels slightly higher than least performers, but

are the least environmentally sustainable, as they use more intensive farm practices than

average. They are of average size and more specialised operations than the average Swedish

crop farm. They are more capital intensive but invest less in new technologies and

activities. They generate below average off-farm income and have higher debt ratios.

338. Class 2 groups the most productive farms, accounting for 44% of all farms, which

are also the most environmentally sustainable. They are larger operations, more likely to

be partnerships. The have more diversified productions, with higher investment in new

technologies and activities. They have larger assets than average but higher debt ratios.

339. Class 3 farms are least productive and achieve average environmental

sustainability, as measured by the intensity of farming practices. They are smaller and more

diverse operations than average. They are more likely operated by women and to generate

a higher share of off-farm income. They are less capital intensive than average, invest less

in new technologies, and have lower debt ratios.

340. In summary, Swedish crop farms in the three identified technology classes differ


over time. Innovative farms are also more likely to be more productive and more

environmentally sustainable than average. Comparatively, smaller farms and to a certain

extent family driven farms show a lower environmental sustainability. Productivity is

correlated positively with land endowment but not significantly with the share of hired

labour. For crop farms in Sweden, crop farms producing more environmentally sustainably

are also found to exhibit higher productivity, ceteris paribus.

Table 12.4. Multiple characteristics of Swedish crop farms, by class





Farm structure


Land (ha) -0.0422 0.3636 -0.2812


-0.5021 0.2682 0.0831





Environmental subsidies per ha (EUR per ha) -0.2111 0.1610 -0.0058





Biofuel income (EUR) -0.0743 0.1265 -0.0611


Technology


Capital per land (EUR per ha) 0.2449 -0.0652 -0.0960

Labour per land (AWU per ha) 0.2171 -0.0766 -0.0692

Materials per land (EUR per ha) 0.3925 0.0813 -0.1910

Diversity





Individual

Age (years) 0.0278 0.0143 -0.0293

Gender (1-female, 0-male) -0.3183 0.0508 0.1534

Education -0.2936 0.0715 0.1207

Location

Subregion 1 -0.0410 0.0055 0.0206

Subregion 2 0.0000 0.0000 0.0000

Subregion 3 -0.0155 -0.0050 0.0138

Subregion 4 0.0446 -0.0218 -0.0091

Subregion 5 0.0004 0.0058 -0.0051

Subregion 6 -0.0364 0.1172 -0.0766

Subregion 7 -0.0352 -0.0151 0.0345

Subregion 8 -0.0190 0.0992 -0.0722

Subregion 9 -0.1243 0.0821 0.0073

Subregion 10 0.0679 -0.0078 -0.0353

Subregion 11 0.0260 -0.0309 0.0101

Subregion 12 -0.0624 -0.0314 0.0650

Subregion 13 -0.1108 0.0049 0.0642

Subregion 14 -0.0900 -0.0185 0.0712

Subregion 15 -0.0187 0.0035 0.0086

Subregion 16 -0.0900 0.0182 0.0402

Subregion 17 0.0905 -0.0357 -0.0256

Subregion 18 0.2195 -0.0677 -0.0782

Subregion 19 0.1396 0.0040 -0.0895

Subregion 20 -0.1122 -0.0275 0.0925

Subregion 21 0.0623 -0.0389 -0.0055

Subregion 22 0.1337 -0.0284 -0.0585

Subregion 23 -0.0089 -0.0376 0.0372

Subregion 24 -0.0404 0.0101 0.0164

Subregion 25 -0.0468 -0.0639 0.0829

(all: 1=yes, 0=no)

Altitude (1- <300m, 2- 300-600m, 3- >600m) 0.1179 -0.0259 -0.0509

Less favoured area payments (EUR) -0.1839 0.0719 0.0527

Household


Spouse labour (AWU -0.0961 0.0388 0.0265

Relative labour (AWU) 0.0118 0.0775 -0.0727

Financial

Total assets (EUR) 0.1133 0.3191 -0.3396

Total subsidies (EUR) -0.1593 0.0477 0.0579






13. United Kingdom

341. For the United Kingdom, the analysis applies to a sample of crop farms, dairy

farms, pig farms and mixed farms, all covering the period 1995 to 2017.

13.1. UK dairy farms


the sample, the average milk output per farm in 2017 was about GBP 374 000 (with a total

output of about GBP 496 000). The variable cost items increased over time and the share

of hired labour significantly increases for the average UK dairy farm in the period

investigated. The average herd size was about 170 dairy cows in 2017, against about

96 cows in 1995. The average stocking density decreased from 1.24 LU per ha in 1995 to

1.41 LU per ha in 2017.


components for the UK dairy production the output variable is total output per farm and

year and input variables are land, dairy cows, capital, labour and materials. The




344. For UK dairy farms, three distinct technology classes emerge from the model


technology classes (Table 13.1), with Class 2 including over half of all farms.




Class 2 produce per farm and year broadly twice the output of farms in classes 1 and 3,

although farms in Class 3 have a slightly higher productivity than farms in Class 1.

346. All three farm classes show significant positive technical change, but the growth

rate for the least productive farms in Class 1 is higher than that for other classes, suggesting

that they will catch up over time with both farms in Class 2 and farms in Class 3, ceteris

paribus. At the same time, farms in Class 3 will see their productivity deteriorating against

that in other farm classes, as they have the lowest rate of technical change (Figure 12.1).


Class 2, they would be able to increase their productivity by 127% and 76% respectively.

As with developments in technical change, this suggests that technologies in Class 1 and

Class 3 differ significantly, although current productivity levels are close (Table 13.1).

Overall, the adoption of Class 2 technologies by all farms would result in an average 30%

increase in productivity for UK dairy farms (Table A C.4 in Part 1).



Figure 13.1. Productivity and technical change for UK dairy farms,



Table 13.1. Productivity characteristics of UK dairy farms, by class













- Class 3 Technology 0.199*** 2.670*** 0.317*




dairy farms in classes 2 and 3 exhibit increasing returns to scale of 1.0856 and 1.0345

respectively. However, dairy farms in Class 1 show significantly decreasing returns to scale

of 0.7602 (Table A D.13.4).



350. Class 1 farms are the most environmentally sustainable and least productive ones.

They are smaller, more family-oriented and specialised farms, which are more likely to

adopt agri-environmental practices and have land in less favoured areas. They are less

2.31

1.96

0.32

0

0.5

1

1.5

2

2.5

0

50

100

150

200

250

300

350


%EUR '000




capital intensive than average, and invest less in new technologies and practices. Their

operators are older and less educated than average, and they have higher off-farm income.

351. Class 2 farms are the most productive farms achieving average environmental

sustainability scores accounting for 55% of all farms. They are larger farms (herd size) that

use hired labour, invest in new technologies and activities and use capital more intensively.

They are more likely to be in rural and hilly areas. Their operators are more educated than

average.

352. Class 3 farms are the least environmentally sustainable farms. They achieve

productivity levels that are just above the average of the least productive performers in

Class 1, with which they share similar characteristics, except that they are more diversified

and have younger operators with lower education. They are more likely to be close to urban

centres.

353. In summary, UK dairy farms in the three identified technology classes differ


over time. Innovative farms are likely to be more productive. Family operated and smaller

farms show a higher environmental sustainability based on the indicators used. Highly

environmentally sustainable dairy farms most likely produce with a lower capital and input

intensity as well as a higher diversity of production. However, their innovativeness and

financial stability is low, ceteris paribus. More than half of all dairy farms produce with a

high productivity level and a slightly lower than average environmental sustainability level.

Those farms are very innovative and produce with a relatively high capital intensity.

Figure 13.2. Multi-dimensional indices for UK dairy farms



-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Table 13.2. Multiple characteristics of UK dairy farms, by class





Farm structure


Number of holdings (1 Sole trader (incl. farmer & spouse partnership) 2 Partnership (other family only) 3 Partnership (other) 4 Farming company 5 Farm company subsidiary)

-0.0123 0.0099 -0.0117

Herd size (LU) -0.4388 0.2871 -0.2085

Form of ownership (1 Sole trader (including farmer and spouse partnership) 2 Partnership (other family only) 3 Partnership (other) 4 Farming company 5 Farm company subsidiary)

-0.1595 0.1194 -0.1230



Chemicals use (GBP per ha) 0.1163 -0.0099 -0.0421

Organic production share -0.0106 0.0567 -0.1609

Fuel per LU (GBP per LU) -0.0641 0.0919 -0.1875

Environmental subsidies per ha (GBP per ha) 0.0276 0.0021 -0.0501




Share land rented -0.0059 -0.0101 0.0412

Biofuel income (GBP) 0.0215 0.0000 -0.0339

Miscellaneous income (GBP) -0.1079 0.1602 -0.3323

Insurance expenses (GBP) -0.3194 0.2657 -0.3296

Professional fees (GBP) -0.2702 0.2023 -0.2084

Technology

Capital / labour ratio (GBP per AWU) -0.1713 0.1821 -0.3011

Capital per cow (GBP per LU) -0.1462 0.1320 -0.1835

Cow per labour (GBP per AWU) -0.1667 0.1105 -0.0839

Fodder per cow (GBP per LU) -0.3105 0.3129 -0.3315

Materials per cow (GBP per LU) -0.2344 0.1984 -0.2528

Diversity



Woodland area (ha) -0.0060 0.0016 0.0044

Individual

Age (years) 0.0242 -0.0081 -0.0127

Gender (1-female, 2-male) -0.2613 0.3659 -0.7356

Education (0 School only 1 GCSE or equivalent 2 A level or equivalent 3 College / National Diploma/ certificate 4 Degree 5 Postgraduate qualification 6 Apprenticeship 9 Other)

-0.2525 0.2933 -0.5217



Location

Subregion 1 -0.0070 0.0254 -0.0686

Subregion 2 -0.0159 0.0411 -0.1038

Subregion 3 -0.0435 0.0361 -0.0448

Subregion 4 -0.1208 0.0821 -0.0671

Subregion 5 -0.0791 0.0655 -0.0806

Subregion 6 -0.0581 0.0650 -0.1121

Subregion 7 0.0000 0.0000 0.0000

Subregion 8 -0.1162 0.0379 0.0645

Subregion 9 -0.1196 0.0890 -0.0907

Subregion 10 0.1485 -0.0710 -0.0115

(all: 1=yes, 0=no)

Altitude (1 Most of holding below 300m 2 Most of holding at 300m to 600m 3 Most of holding at 600m or over 4 Data not available)

-0.2307 0.3273 -0.6628

Less favoured area (1 All land outside LFA 2 All land inside SDA 3 All land inside DA 4 50% + in LFA of which 50% + in SDA 5 50% + in LFA of which 50% + in DA 6 <50% in LFA of which 50% + in SDA 7 <50% in LFA of which 50% + in DA)

0.2180 -0.0972 -0.0388

Rural-urban classification (1 Urban > 10k – sparse 2 Town and fringe – sparse 3 Village - sparse 4 Hamlet & isolated dwellings - sparse 5 Urban > 10k – less sparse 6 Town & fringe – less sparse 7 Village – less sparse 8 Hamlet & isolated dwellings – less sparse)

-0.2934 0.3711 -0.7013

Household


Labour spouse (AWU) 0.0273 0.0244 -0.1196

Financial

Total assets (GBP) -0.3441 0.2736 -0.3154

Total subsidies (GBP) 0.1648 -0.1728 0.2821





13.2. UK cereal farms

354. According to Table A.B.13.6, which contains descriptive statistical measures for

the sample, the average crop output per farm was about GBP 356 000 in 2017 (with a total

output of about GBP 506 000). The variable cost items increased over time and the share

of hired labour significantly increases for the average UK crop farm. The average farm

operated about 283 ha in 2017 (an increase from an average of about 208 ha in 1995).


components for the UK cereal production the output variable is total output per farm and

year and input variables are land, capital, labour, chemicals, and materials. The




356. For UK cereal farms, three distinct technology classes emerge from the model


technology classes (Table 13.3), with Class 2 including only 8% of all farms, while the

remaining farms are shared evenly between Class 1 and Class 3.





than their counterparts in classes 2 and 3 (Figure 13.3). However, crop farms in Class 2 are

not far away as they produce 85% of the output per farm and year produced by farms in

Class 1, while farms in Class 3 produce half of Class 1 output.

358. UK crop farms in classes 1 and 2 show significant positive rates of technical

change, but the growth rate for Class 2 farms is three times higher, suggesting they will

rapidly catch up with farms in Class 1, ceteris paribus. The least productive farms in

Class 3, which use the most extensive farm practices, show a fast declining rate of technical

change, suggesting that their productivity lag with other farms will increase over time,


Figure 13.3. Productivity and technical change for UK cereal farms,




would be increase their productivity by 25% (Table 13.3). However, if farms in Class 2

adopted the technology of the most productive farms in Class 1, their productivity would

decrease by 20%, suggesting that they already produce with a technology, which is best

adapted to their conditions. Overall, the productivity of UK crop farms would increase by

5% (Table A C.5 in Part 1). At the same time, the productivity of farms in Class 1 would

be the highest if they adopted the technology of farms in Class 3, which involves higher

specialisation and more extensive farming practices. This could suggest that Class 3

technology is better adapted to the rural, hilly regions, where farms in Class 1 are more

likely located.

1.11

3.27

-2.39

-3

-2

-1

0

1

2

3

4

0

50

100

150

200

250

300


%EUR '000




Table 13.3. Productivity characteristics of UK cereal farms, by class



Number of observations (% of sample farms) 49.1% 8% 42.9%










- Class 3 Technology 13.410*** 6.249*** -2.394***




all dairy farm classes exhibit increasing returns to scale: 1.0358 for Class 1 farms, 1.0462

for Class 2 farms, and 1.0604 for Class 3 farms (Table A D.13.9).

361. Figure 13.4 summarises the various cereal farm classes in terms of estimated


362. Class 1 farms are the most productive, account for about half of all farms, and

achieve below average environmental sustainability. They are larger, more diversified

operations, which invest in new technologies and activities. They are more capital intensive

and achieve higher financial ratios. Their operators are more likely to be men, older than

average and a with better education level.

363. Class 2 farms are the 8% least environmentally sustainable and achieve close to

highest productivity levels. They are smaller and more specialised operations than average

and they use the most intensive farm practices. They are capital intensive and invest in new

technologies. They are more reliant on off-farm income and their financial performance is

lower than average.

364. Class 3 farms are the most environmentally sustainable, using the most extensive

farm practices, and the least productive. They account for 43% of all farms and are smaller

and more specialised than average. They are more likely to be operated by women, with

lower education levels. They are less capital intensive than average, and have lower

investment in new technologies.

365. In summary, UK cereal farms in the three identified technology classes differ


over time. The empirical findings largely confirm the results for other crop sectors: a strong

positive correlation between innovativeness and productivity. Nearly half of all cereal

farms show a high level of productivity, however, they also reveal low scores on

environmental sustainability based on the measures used in the empirical analysis. These

farms are less family driven and of larger size with a high financial stability. A strong share



of farms (nearly 43%) produce with a high environmental sustainability but a relatively low

productivity. Those farms exhibit low levels of innovativeness and technology intensity.

Figure 13.4. Multi-dimensional indices for UK cereal farms



Table 13.4. Multiple characteristics of UK cereal farms, by class





Farm structure

Family/hired labour ratio 0.0148 -0.0712 -0.0037

Number of holdings 0.1012 -0.1107 -0.0951

Land (ha) 0.1025 -0.3494 -0.0523


0.0216 -0.0742 -0.0109


Chemicals use (GBP per ha) 0.2629 0.8822 -0.4643

Organic production share 0.0774 0.1140 -0.1097

Fuel per LU (GBP per LU) 0.0402 1.2536 -0.2787

Environmental subsidies per ha (GBP per ha) 0.2539 -0.0659 -0.2780

Tillage area (ha) 0.0604 -0.3186 -0.0098





Biofuel income (GBP) 0.0439 0.0127 -0.0526

-1.5

-1

-0.5

0

0.5






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Miscellaneous income (GBP) 0.3500 -0.0278 -0.3949

Insurance expenses (GBP) 0.2139 0.1084 -0.2646

Professional fees (GBP) 0.1282 0.0068 -0.1478

Technology

Capital / labour ratio (GBP per AWU) 0.0397 -0.0172 -0.0421

Material per land (GBP per ha) 0.0449 0.3299 -0.1126

Labour per land (AWU per ha) -0.1448 1.3945 -0.0934

Capital per land (GBP per ha) -0.0013 1.4336 -0.2647

Diversity



Woodland area (ha) -0.0138 -0.0847 0.0315

Individual

Age (years) 0.0270 -0.0138 -0.0283

Gender (1-female, 2-male) 0.7555 0.1771 -0.8965


0.5774 0.0608 -0.6712

Location

Subregion 1 0.1042 -0.0568 -0.1086

Subregion 2 0.0990 0.1603 -0.1429

Subregion 3 0.0634 0.0109 -0.0745

Subregion 4 0.1854 -0.1638 -0.1815

Subregion 5 0.0813 0.2021 -0.1305

Subregion 6 0.2093 -0.1710 -0.2075

Subregion 7 0.0000 0.0000 0.0000

Subregion 8 0.1248 0.1133 -0.1637

Subregion 9 0.1105 0.1145 -0.1476

Subregion 10 0.0508 0.3058 -0.1148

(all: 1=yes, 0=no)


0.8156 0.1681 -0.9635


0.0195 0.0335 -0.0285


0.8026 0.1433 -0.9441

Household


Labour spouse (AWU) 0.0674 0.1127 -0.0979

Financial

Total assets (GBP) 0.2574 -0.0961 -0.2764

Total subsidies (GBP) -0.0483 -0.0510 0.0644







13.3. UK mixed crop and livestock farms


the sample, the average total output per farm was about GBP 357 000 in 2017 (about

GBP 223 000 in 1995). The variable cost items increased over time and the share of hired

labour increases for the average UK mixed farm over the period investigated. The average

mixed farm cultivated about 198 ha in 2017 and operated a total herd size of about 149 LU

(against 156 ha and 184 LU in 1995).


components for the UK crop and meat production the output variable is total output per

farm and year and input variables are land, capital, labour, and livestock units. The




368. For UK mixed crop and livestock farms, three distinct technology classes emerge

from the model estimates (Table A D.13.12). The individual farms are distributed relatively

evenly across the three technology classes (Table 13.5), although Class 3 is larger than the

other two classes.

369. Crop and livestock farms in Class 1 achieve a higher productivity performance —

measured as the potential output levels that could be achieved with a given input bundle,

in value per year — than their counterparts in classes 2 and 3 (Figure 13.5). Farms in

Class 1 produce about 3 times more than farms in Class 2, and twice as much as farms in

Class 3.

370. Farms in Class 3 show a significantly positive technical change, which growth

twice as fast as that for farms in Class 1, suggesting that the productivity difference between

the two classes will narrow over time. But the performance gap between farms in Class 2

and other crop and livestock farms in the UK is increasing as the former experience a

significant, negative technical change, ceteris paribus (Figure 13.5).

Figure 13.5. Productivity and technical change for UK mixed crop and livestock farms,



0.94

-1.96

2.10

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

0

50

100

150

200

250

300

350

400

Class 1 (28.3% of farms) Class 2 (29.2% of farms) Class 3 (42.5% of farms)

%EUR '000




371. If crop and livestock farms in classes 2 and 3 produced with the technology of farms

in Class 1, they would be able to increase their productivity by 85% and 23% respectively

(Table 13.5). Overall, if all farms were producing with Class 1 technology, the productivity

of UK crop and livestock farms would increase by 21%, on average (Table A C6 in Part 1).

Table 13.5. Productivity characteristics of UK mixed crop and livestock farms, by class






Productivity level (GBP per year)1





- Class 1 Technology 0.940** -0.225* 0.345*

- Class 2 Technology 10.144*** -1.962* 14.154***





all mixed farm classes exhibit increasing returns to scale: 1.0966 for Class 1, 1.2396 for

Class 2 and 1.1192 for Class 3 farms.

373. Figure 13.6 summarises the various mixed crop and livestock farm classes in terms

of estimated indices, while Table 13.6 contains the estimates for variables used to construct

the indices.

374. Class 1 farms are the most productive and the least environmentally sustainable.

They are larger, more specialised operations, which are more likely to be partnerships. They

invest more in new technologies and activities and are more capital intensive than average.

They are more likely to be operated by men with higher than average education levels.

375. Class 2 groups the least productive farms (about 40% of all farms), achieving

slightly below average environmental sustainability. They are smaller than average and

more family-driven operations. They are less capital intensive and invest less in new

technologies. They are more likely to be in less-favoured areas, receive more subsidies and

have a higher share of off-farm income than average.

376. Most environmentally sustainable farms are in Class 3. They achieve medium

productivity and account for about a third of all farms. They are the smallest, most

diversified operations. They are more likely to be operated by women with higher than

average education levels, and to be located in hilly, rural areas. They have lower debt ratios

than average and receive less subsidies.



Figure 13.6. Multi-dimensional indices for UK mixed crop and livestock farms



377. In summary, UK mixed crop and livestock farms in the three identified technology

classes differ significantly with respect to their economic performance as well as technical

development over time. For these farms, the analysis suggests that nearly a third of all farms

in the sector produce with a very high productivity linked to high levels of innovativeness

but significantly low levels of environmental sustainability based on the measures used in

the empirical analysis. These farms score relatively high on financial stability and liquidity

indicators with a capital and input intensive production. Nevertheless, a large share of likely

family driven mixed farms in the United Kingdom (about 43%) show a high environmental

sustainability and also medium productivity levels. Those farms are average innovative and

capital intensive.

-2

-1.5

-1

-0.5

0

0.5






Index 7 - Location

Index 8 - Household

Index 9 - Financial






Table 13.6. Multiple characteristics of UK mixed crop and livestock farms, by class





Farm structure



Herd size (LU) 0.5448 -0.1565 -0.2552


0.4223 -0.1083 -0.2068




Organic production share -0.2155 -0.1910 0.2750

Fuel per LU (GBP per LU) 0.1577 -0.1729 0.0140

Environmental subsidies per ha (GBP per ha) -0.1501 -0.2853 0.2963



Share contract farming -0.2571 -0.0810 0.2207


Biofuel income (GBP) 0.0907 -0.0749 -0.0088

Miscellaneous income (GBP) 0.2580 -0.4624 0.1463

Insurance expenses (GBP) 0.4548 -0.4250 -0.0105

Professional fees (GBP) 0.3406 -0.2626 -0.0462

Technology

Capital / labour ratio (GBP per AWU) 0.1292 -0.2820 0.1080

Capital per LU (GBP per LU) 0.1587 -0.1990 0.0313

LU per labour (LU per AWU) 0.0622 -0.0745 0.0098

Materials per LU (GBP per LU) 0.2437 -0.1766 -0.0408

Fodder per LU (GBP per LU) 0.5557 -0.2236 -0.2163

Diversity

Herfindahl Index (sqrt[Σ(yi/Y)2]) 0.6379 0.1620 -0.5364


Woodland area (ha) 0.0851 -0.0287 -0.0369

Individual

Age (years) 0.0218 -0.0768 0.0383

Gender (1-female, 2-male) 0.2012 -0.9957 -0.5364


0.2372 -0.6936 0.3193



Location

Subregion 1 -0.0255 -0.1001 0.0859

Subregion 2 0.0366 -0.1243 0.0612

Subregion 3 0.0751 -0.1411 0.0471

Subregion 4 0.1018 -0.1949 0.0663

Subregion 5 0.0557 -0.2115 0.1084

Subregion 6 0.2340 -0.1742 -0.0360

Subregion 7 0.0000 0.0000 0.0000

Subregion 8 -0.0234 -0.2248 0.1703

Subregion 9 -0.0207 -0.2295 0.1717

Subregion 10 -0.1191 -0.0180 0.0917

(all: 1=yes, 0=no)


0.2308 -1.0981 0.6019


-0.1037 0.1034 -0.0021


0.2753 -1.0659 0.5501

Household


Labour spouse (AWU) -0.0709 -0.1594 0.1569

Financial

Total assets (GBP) 0.3899 -0.4173 0.0275

Total subsidies (GBP) -0.2712 0.7860 -0.3433

Equity/debt ratio -0.0944 0.0174 0.0521




13.4. UK pig farms


the sample, the average pig meat output per farm was about GBP 895 000 in 2017 (about

GBP 303 000 in 1995). The variable cost items increased over time and the share of hired

labour slightly decreased for the average UK pig farm in the period investigated. The

average farm operated a total herd size of about 793 LU with a stocking density of about

39.12 LU per ha (compared to 280 LU and a density of about 106 LU per LU in 1995).


components for the UK pig meat production the output variable is total output per farm and

year and input variables are land, capital, labour, livestock units, and materials. The




380. For UK pig farms, three distinct technology classes emerge from the model




technology classes (Table 13.7), with Class 2 including close to two-thirds of all farms,

while Class 3 only contains 7% of all farms.

381. Pig farms in Class 1 achieve a higher productivity performance — measured as the

potential output levels that could be achieved with a given input bundle, in value per year

— than their counterparts in classes 2 and 3 (Figure 13.7). Farms in classes 2 and 3 produce

respectively about 75% and 22% of the output of farms in Class 1.

382. All farm classes show a positive rate of technical change, with large differences

across classes. The highest rate is found for the least productive farms in Class 3 but the

most productive farms in Class 1 also show a strong growth in technical change. The

convergence is thus likely to be slow. The performance gap between farms in Class 2 and

farms in Class 1 is likely to increase, ceteris paribus, as the former have a much lower rate

of technical change (0.5% against 3.2%) (Figure 13.7).

383. If pig farms in Class 3 produced with the technology of farms in Class 1, they would

be able to increase their productivity by 51% (Table 13.7). But the technology of farms in

Class 2 seems well adapted to their conditions as switching technology would result for

them in a slight decline in productivity, and would probably deteriorate their environmental

sustainability performance as they have more extensive farming practices than farms in

Class 1. Overall, if all farms were producing with Class 1 technology, the productivity of

UK pig farms would increase by less than 1% (Table A C6 in Part 1).

Figure 13.7. Productivity and technical change for UK pig farms,



3.17

0.45

3.56

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

0

50

100

150

200

250

300

350

400


%EUR '000




Table 13.7. Productivity characteristics of UK pig farms, by class



Number of observations (% of sample farms) 28.3 64.8 6.9



Productivity level (GBP per year)1











all pig farm classes exhibit increasing returns to scale: 1.029 for farms in Class 1, 1.0448

for farms in Class 2, and 1.3557 for farms in Class 3 (Table A D.13.19).

385. Figure 13.8 summarises the various pig farm classes in terms of estimated indices,


Figure 13.8. Multi-dimensional indices for UK pig farms



-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6






Index 7 - Location

Index 8 - Household

Index 9 - Financial






386. Class 1 farms are the most productive farms, with below average environmental

sustainability. They are the largest, most diversified operations, and the most likely to be

partnerships, with managers having better education. They have the highest capital

intensity and investments in new technologies, building on large assets. They are more

likely to be located in a less-favoured and rural area. They receive the highest level of

subsidies and have lower debt ratios.

387. Class 2 groups the most environmentally sustainable farms (close to two-thirds of

all farms), which achieve below average productivity. They are smaller and more

specialised operations than average, with lower investment in new technologies and capital

intensity. Their managers are more likely to be younger, female, and with lower education

levels than average. They are below average in terms of asset endowment, subsidy received

and equity/debt ratio.

388. Class 3 group the 7% of farms that are the least environmentally sustainable and

least productive. They are the smallest operations, least capital intensive operations, which

invest the least in new technologies. Their managers are more likely to be older and male.

They are more likely to be located in hilly, rural areas, but not less favoured areas. They

have the lowest asset endowment and are the most reliant on off-farm income.

389. In summary, UK pig farms in the three identified technology classes differ


over time. More innovative farms are most likely more productive compared to their peer

group, and family driven and smaller farms generally show a higher environmental

sustainability, ceteris paribus. However, a small share of the farms in pig farming sector

(about 7%) are largely dependent on family labour. The analysis revealed very low scores

on environmental sustainability based on the measures used in the empirical analysis.

Those farms also show very low levels of productivity and a low intensity of production.

Table 13.8. Multiple characteristics of UK pig farms, by class





Farm structure



Herd size (LU) 0.4101 -0.1444 -0.3235


0.3872 -0.1167 -0.4889



Chemicals use (GBP per ha) 0.1537 -0.1120 0.4218

Organic production share -0.0056 0.0076 -0.0482

Fuel per LU (GBP per LU) 0.1264 -0.1108 0.5228

Environmental subsidies per ha (GBP per ha) -0.0221 -0.0052 0.1390



Share contract farming 0.1303 -0.0672 0.0923


Biofuel income (GBP) -0.0263 -0.0228 0.3217






Technology


Capital per LU (GBP per LU) 0.3613 -0.1335 -0.2258

LU per labour (GBP per AWU) 0.2566 -0.0749 -0.3469

Fodder per LU (GBP per LU) -0.2034 0.1230 -0.3219

Materials per LU (GBP per LU) -0.0540 -0.0164 0.3749

Diversity



Woodland area 0.3914 -0.1490 -0.2035

Individual

Age (years) 0.0505 -0.0318 0.0918

Gender (1-female, 2-male) 0.1612 -0.1003 0.2817


0.2769 -0.1283 0.0712

Location

Subregion 1 -0.0344 0.0239 -0.0837

Subregion 2 -0.0701 0.0503 -0.1849

Subregion 3 0.2103 -0.0873 -0.0414

Subregion 4 -0.1002 0.0391 0.0432

Subregion 5 -0.0553 0.0332 -0.0852

Subregion 6 0.0652 -0.0968 0.6416

Subregion 7 0.0000 0.0000 0.0000

Subregion 8 0.1084 -0.0341 -0.1236

Subregion 9 -0.0412 0.0447 -0.2504

Subregion 10 0.0000 0.0000 0.0000

(all: 1=yes, 0=no)


0.1770 -0.1005 0.2191


0.0957 -0.0260 -0.1473


0.1674 -0.1063 0.3122

Household


Labour spouse (AWU) -0.1177 0.0486 0.0251

Financial

Total assets (GBP) 0.5043 -0.1850 -0.3274

Total subsidies (GBP) 0.1984 -0.0796 -0.0905







13.5. UK poultry farms

390. According to Table A D 13.21, which contains descriptive statistical measures for

the sample, the average livestock related output per farm was about GBP 872 000 in 2017

(with a total of about GBP 965 000 per farm). The variable cost items increased over time

and the share of hired labour significantly increases for the average UK poultry farm over

the period investigated. The average farm operated a total herd size of about 406 LU in

2017 (compared to a total herd size of 495 LU in 1995). Hence, the stocking density

decreased from about 39.4 LU per ha in 1995 to about 28.4 LU per ha in 2017.


components for the UK poultry meat production, the output variable is total output per farm

and year and input variables are land, capital, labour, livestock units, and materials. The




392. For UK poultry farms, three distinct technology classes emerge from the model


three technology classes (Table 13.9), as Class 3 includes about three-quarters of all UK

poultry farms.

393. Poultry farms in Class 1 achieve a higher productivity performance — measured as

the potential output levels that could be achieved with a given input bundle, in value per

year — than their counterparts in classes 2 and 3 (Figure 13.9). Farms in classes 2 and 3

produce respectively about 19% and 64% of the output of farms in Class 1.

394. All poultry farm classes show a positive rate of technical change, which is

particularly strong (over 5%) for the least productive farms in Class 2. Though lower, the

rate of technical change of farms in Class 1 is also high (3.3%). The lowest rate is found

for farms in Class 3, suggesting that the productivity gap with Class 1 farms is increasing,

ceteris paribus. Differences in technical rate rates also suggest the convergence between

farms in Class 3 and Class 2 will be faster, than that between Class 2 and Class 1

(Figure 13.9).

395. If poultry farms in Class 2 produced with the technology of farms in Class 1, they

would be able to increase their productivity by 63% (Table 13.9). But the technology of

farms in Class 3 seems well adapted to their conditions and relatively close to that used by

Class 1 farms, as switching technology would not have a significant effect on productivity.

Overall, if all farms were producing with Class 1 technology, the productivity of UK

poultry farms would increase by less than 1% (Table A C6 in Part 1).



Figure 13.9. Productivity and technical change for UK poultry farms,




all poultry farm classes exhibit slightly to significantly increasing returns to scale: 1.0372

for Class 1, 1.4736 for Class 2, and 1.0548 for Class 3 (Table A D.13.24).

Table 13.9. Productivity characteristics of UK poultry farms, by class



Number of observations (% of sample farms) 21% 4% 75%









- Class 2 Technology -0.579* 5.284*** -1.060*




397. Figure 13.10 summarises the various poultry farm classes in terms of estimated


3.30

5.28

1.39

0

1

2

3

4

5

6

0

50

100

150

200

250

300

Class 1 (21%) Class 2 (4%) Class 3 (75%)

%EUR '000




Figure 13.10. Multi-dimensional indices for UK poultry farms



398. Class 1 groups the most productive farms, which are below average in terms of

environmental sustainability. They are the largest, most diversified operations, and the most

likely to be partnerships, with managers having better education. They have the highest

capital intensity and investments in new technologies, building on large assets. They are

more likely to be located in a less-favoured and a rural area. They receive the highest level

of subsidies and have lower debt ratios.

399. Most environmentally sustainable farms achieving below average productivity are

in Class 2, which accounts for close to two-thirds of all farms. They are smaller and more

specialised operations than average, with lower investment in new technologies and capital

intensity. Their managers are more likely to be younger, women, and with lower education

levels than average. They are below average in terms of asset endowment, subsidy received

and equity/debt ratio.

400. Class 3 farms (7% of all farms) are the least environmentally sustainable and least

productive. They are the smallest operations, least capital intensive operations, which

invest the least in new technologies. Their manager is more likely to be older and male.

They are more likely to be located in hilly, rural areas, but not less favoured areas. They

have the lowest asset endowment and are the most reliant on off-farm income.

401. In summary, UK poultry farms in the three identified technology classes differ


over time. The analysis revealed that most of these farms (around 75%) produce with a

medium level of productivity and are relatively environmentally sustainable based on the

measures used in the empirical analysis. These poultry farms are more family driven with

a medium capital and input intensity. The most productive farms in the sector, however,

score lower than average on environmental sustainability but show high levels of

innovativeness. Those farms are less likely family driven farms and operate large flocks of

poultry.

-1-0.8-0.6-0.4-0.2

00.20.40.60.8

11.2

Index 1 - Structure





Index 7 - Location

Index 8 - Household

Index 9 - Financial






Table 13.10. Multiple characteristics of UK poultry farms, by class





Farm structure



Herd size (LU) 0.4588 -0.3658 -0.1099


0.4454 -0.3267 -0.1081




Organic production share -0.0441 0.0305 0.0108

Fuel per LU (GBP per LU) 0.0864 0.3163 -0.0388

Environmental subsidies per ha (GBP per ha) 0.1332 -0.0141 -0.0363





Biofuel income (GBP) 0.0888 -0.0424 -0.0226




Technology


Capital per LU (GBP per LU) 0.0654 0.3850 -0.0362

LU per labour (GBP per AWU) 0.0774 -0.3260 -0.0061

Fodder per LU (GBP per LU) -0.2411 -0.0747 0.0703

Materials per LU (GBP per LU) -0.0738 0.2265 0.0098

Diversity



Woodland area 0.2268 0.0707 -0.0662

Individual

Age (years) 0.0644 0.1891 -0.0267

Gender (1-female, 2-male) 0.4066 -0.4012 -0.0938


0.3245 -0.3864 -0.0718



Location

Subregion 1 -0.0298 -0.0035 0.0084

Subregion 2 -0.0339 -0.0691 0.0127

Subregion 3 0.2002 -0.1485 -0.0485

Subregion 4 0.0116 -0.1649 0.0045

Subregion 5 0.0540 -0.1042 -0.0101

Subregion 6 0.3838 -0.1923 -0.0973

Subregion 7 0.0000 0.0000 0.0000

Subregion 8 0.1050 -0.0609 -0.0262

Subregion 9 0.1010 -0.1226 -0.0222

Subregion 10 -0.2446 0.1906 0.0588

(all: 1=yes, 0=no)


0.4695 -0.4467 -0.1091


-0.3420 0.3762 0.0771


0.4835 -0.4711 -0.1118

Household


Labour spouse (AWU) -0.1887 -0.3165 0.0672

Financial

Total assets (GBP) 0.4581 -0.3654 -0.1097

Total subsidies (GBP) -0.0084 0.4798 -0.0205

Equity/debt ratio -0.0338 0.4362 -0.0122




working party on agricultural policies and markets drivers of

Documents