attitudes to inequality and povertydarp.lse.ac.uk/subjective/presentation.pdf · dalton’s...

Attitudes to Inequality and Poverty

Frank CowellJune 2001

Some Basic Questions

• Where do ideas about inequality and poverty comparisons come from?

• How are they related to other distributional concepts?

• Are there systematic differences in attitudes amongst groups of the population?

Theoretical Approaches

• Ad hoc measures of inequality and poverty• Social welfare approaches• Axiomatic approaches

Outline of present approach

• Focus on specific issues of the inequality or poverty map.

• Encapsulate them in a series of questions• Ask people…

Example. Let’s look at the core of inequality analysis, which is also used in modern approaches to poverty comparisons

Pigou’s Approach to Income Inequality

0 1 2 3 4 5 6 7 8 9 10$

P R

0 1 2 3 4 5 6 7 8 9 10$

P R

0 1 2 3 4 5 6 7 8 9 10$

P R

0 1 2 3 4 5 6 7 8 9 10$

RP

Incr

easi

ng In

equa

lity

Pigou’s Approach: Limitations

• Comparability of incomes• Absence of History• Restriction to two persons

Dalton’s Approach• The Principle of Transfers is often referred to as the

“Pigou-Dalton” condition

• This is a bit misleading

• Hugh Dalton strengthened the Pigou idea considerably (Econ Journal 1920)

• This was to give a neat form for the n-person economy

• It formed the basis of all the modern work on Distributional Analysis, including Social Welfare and Poverty

But do you think it’s reasonable? You do? Perhaps think again…

Dalton’s Approach

Which group seems to have the more unequal distribution?

0 1 2 3 4 5 6 7 8 9 10 11 12 13

0 1 2 3 4 5 6 7 8 9 10 11 12 13

$

$

The “Wrong” Answer?

0 1 2 3 4 5 6 7 8 9 10 11 12 13

0 1 2 3 4 5 6 7 8 9 10 11 12 13

$

$

The “Truth”?

0 1 2 3 4 5 6 7 8 9 10 11 12 13

0 1 2 3 4 5 6 7 8 9 10 11 12 13

$

$

Issues to be Investigated

• What form of inequality comparisons do people commonly make?

• Do they have the same structure as risk comparisons?

• What about social welfare?•The Atkinson analogy

•Rothschild-Stiglitz

•The Harsanyi paradigm

•The Atkinson analogy

•Rothschild-Stiglitz

•The Harsanyi paradigm

Outline of Approach

• Questionnaire responses of international group of over 1100 students

• Questionnaire experiments were run during 1998

• Each session run during lecture/class time

• Questionnaire consisted of a combination of (related) numerical problems and a verbal question

• Although the experiment was anonymous individuals were asked about personal characteristics

The Questionnaire

INEQUALITY QUESTIONNAIRE

This questionnaire concerns people's attitude to inequality. We would be interested in yourview, based on hypothetical situations. Because it is about attitudes there are no "right"answers. Some of the possible answers correspond to assumptions consciously made byeconomists: but these assumptions may not be good ones. Your responses will help to shedsome light on this, and we would like to thank you for your participation. The questionnaireis anonymous.

Alfaland consists of five regions that are identical in every respect other thanthe incomes of their inhabitants. Everyone within a given region receives thesame income, but personal incomes differ from region to region.

Two economic policy proposals A and B are being considered forimplementation in Alfaland next year. It is known that - apart from theirimpact on personal incomes - the two policies would have the same effect onthe population. The impact upon the regions’ incomes would depend uponthe particular state of the Alfaland economy at the time the policy (A or B) isto be introduced.

In each of questions (1) to (6) two alternative lists of incomes A and B (inAlfaland local currency) are given. Each of these pairs represents theoutcomes of the A-policy and the B-policy on the five regions in each of sixdifferent situations in which Alfaland might find itself next year. In eachcase please state which policy you consider would result in higherinequality in Alfaland by circling A or B. If you consider that the twopolicies will result in the same inequality then circle both A and B.

RISK QUESTIONNAIRE

This questionnaire concerns people's attitude to risk. We would be interested in your view,based on hypothetical situations. Because it is about attitudes there are no "right" answers.Some of the possible answers correspond to assumptions consciously made by economists: butthese assumptions may not be good ones. Your responses will help to shed some light on this,and we would like to thank you for your participation. The questionnaire is anonymous.

Alfaland consists of five regions that are identical in every respect other thanthe incomes of their inhabitants. Everyone within a given region receives thesame income, but personal incomes differ from region to region. An immigrantto Alfaland would be assigned at random, with equal probability, to any one ofthese five regions. Such a person would therefore have a 20% chance of beingon any one of five income levels.

Two economic policy proposals A and B are being considered forimplementation in Alfaland next year. It is known that - apart from their impacton personal incomes - the two policies would have the same effect on thepopulation. The impact upon the regions’ incomes would depend upon theparticular state of the Alfaland economy at the time the policy (A or B) is to beintroduced.

In each of questions (1) to (6) two alternative lists of incomes A and B (in Alfalandlocal currency) are given. Each of these pairs represents the outcomes of the A-policy and theB-policy on the five regions in each of six different situations in which Alfaland might find itselfnext year. In each case please state which policy you consider would result in higher risk for aperson immigrating to Alfaland by circling A or B. If you consider that the two policies willresult in the same risk to a potential immigrant then circle both A and B.

Features of Questionnaire: 1

• Two questionnaires for the price of one: risk questionnaire generated from inequality by Ctrl-H

• Students ranked six pairs of income vectors (A and B) in terms of risk and inequality

• For each question B was obtained from A by anequalising income transfer from a rich to a poor region

• According to the Principle of Transfers, anequalising transfer reduces risk/inequality; this implies that A is riskier/more unequal than B in all six questions

Questionnaire: Numerical Part

1) A = (2, 5, 9, 20, 30) B = (2, 6, 8, 20, 30)

2) A = (2, 5, 9, 20, 30) B = (3, 5, 9, 20, 29)

3) A = (2, 5, 9, 20, 30) B = (2, 6, 9, 20, 29)

4) A = (2, 5, 9, 20, 30) B = (2, 10, 9, 15, 30)

5) A = (10, 10, 10, 10, 30) B = (10, 10, 10, 20, 20)

6) A = (2, 5, 9, 20, 30) B = (2, 6, 9, 19, 30)

Features of Questionnaire 2

• Check the numerical responses with a verbal question

• Using the same story we present the issue of the principle of transfers

• Then see if they want to change their minds on the numerical problems

Questionnaire: Verbal Part

risk

risk

riskrisk

…and for risk

Questionnaire: A Check

Data Set: Students by Country

Total Inequality RiskArgentinaAustraliaBelgiumGermanyIsraelSwedenUK

7412413726418063

311

386287

1269331

150

366250

138 8732

161OVERALL 1150 475 441

NumberTotal Inequality Risk

8 8 814 13 1415 18 1123 21 2420 20 20

7 7 734 32 37

100 100 100

Percentage

Students from 7 countries answering inequality and risk questionnaires in equal proportions

The Questionnaire: Personal Characteristics

Finally, we would be grateful for some information about yourself:

• Are you male or female? M/F• What is your age? _____ years• What is your special subject of study? __________• Were you employed before university? Yes / No

• How would you rate your political views? Please puta 4on this scale.

• How would you rate your family’s income in 1990?Please put a 4on this scale.

• How would you rate your own income prospects inthe year 2010? Please put a 4on this scale.

“very Poor”

“veryRich”

“extreme left”

“extremeright”

“very Poor”

“veryRich”

Personal Characteristics of Students

MaleEconomic SubjectAgeEmploymentPolitical OpinionIncome 1990Income Change 2010

DefinitionDummyDummyYears

Dummy1-7 (right)

1-7, +1-7, +

P-value

66463248393575

Mean SD0.63 0.480.29 0.4522.9 4.40.43 0.504.0 1.14.4 1.00.7 1.2

InequalityMean SD0.61 0.490.30 0.4623.2 4.00.45 0.504.0 1.14.3 1.10.7 1.4

Risk

Inequality and risk samples are similar across all personal characteristics

Equality in means across risk and

inequality subgroup.







Higher proportion of A responses for risk than inequality, and for male than female, especially for risk

Transfer Principle: Summary Statistics

Inequality Risk

Consistency with Transfer Principle?Total 17 23Male 21 31Female 10 11

proportion of answer A in all questions

(N=6918).

proportion of answer A in all questions

(N=6918).

Equalising Transfer Reduces Inequality?Total 59 61Male 61 67Female 57 53

proportion of students answering A

in all six questions(N=1153).

proportion of students answering A

in all six questions(N=1153).

Regression Approach

• Consider equation of the form: • Prob(answer A)= Φ(b1x1 +b2x2 +…+bnxn)• Estimate this using probit if Φ is standard

normal • Personal characteristics can be used as

dummies

Probit Regression of Response A

P-value

0163755757240

Coef P>|z|0.06 10.10 00.00 630.02 350.00 520.00 750.01 19

InequalityCoef P>|z|0.13 00.15 00.00 500.00 94-0.01 290.01 400.00 85

RiskVariable

MaleEconomic SubjectAgeEmploymentPolitical opinionIncome 1990Income Change 2010

Explanatory variables include dummy variables for countries.Number of observations 6767.

Only “male” and “economic subject” lead to higher share of A responses, especially for risk

Equality of coefficients across subgroups.


Probit Regression of 6 x Response A

P>χ2

16566058907415

Coef P>|z|0.12 10.10 90.00 900.02 54-0.01 450.00 880.02 21

InequalityVariable


Number of observations 1153.Explanatory variables include dummy variables for countries.

As before: male and economic subject lead to higher share of A responses, especially for risk

Coef P>|z|0.21 00.14 10.01 41-0.01 88-0.01 330.00 76-0.01 48

Risk



Numerical Questions: Detail

1) A = (2, 5, 9, 20, 30) B = (2, 6, 8, 20, 30)

2) A = (2, 5, 9, 20, 30) B = (3, 5, 9, 20, 29)

3) A = (2, 5, 9, 20, 30) B = (2, 6, 9, 20, 29)

4) A = (2, 5, 9, 20, 30) B = (2, 10, 9, 15, 30)

5) A = (10, 10, 10, 10, 30) B = (10, 10, 10, 20, 20)

6) A = (2, 5, 9, 20, 30) B = (2, 6, 9, 19, 30)

Type of Income Transfer: Summary

RISKQuestion 1 49 55 39Question 2 68 74 60Question 3 59 66 52Question 4 68 74 60Question 5 67 71 60Question 6 55 62 46

Total Male FemaleINEQUALITY

Question 1 40 44 34Question 2 74 74 74Question 3 61 62 60Question 4 60 64 56Question 5 72 70 76Question 6 48 53 42

Higher share of response A if transfer involves richest region (Q2 and Q5), is to poorest region (Q2); lower share if transfers among middle-ranking regions (Q1 and Q6)

Percentage of “A” answers.


Familiar pattern: male and economic subject are usually significant, leading to higher share of response A. Coefficients are higher for risk

Transfer Type: Probits of Response AVariable

QUESTION 1 MaleEconomic Subject






P>χ2

6869

584

1763

559

121

3159

Coef P>|z|

0.13 10.11 8

0.00 990.11 4

0.05 310.14 2

0.09 40.06 27

-0.07 120.06 31

0.08 80.14 3

InequalityCoef P>|z|

0.18 00.15 2

0.11 10.12 1

0.13 00.18 0

0.12 10.18 0

0.09 40.14 0

0.15 00.10 10

Risk



What if we had used welfare?

• Following Atkinson, inequality rankings should derive from social welfare rankings

• Likewise risk rankings should derive from preference rankings

• What would have happened if we changed the context of the question?

Type of Income Transfer: Summary

PreferenceQuestion 1 44 44Question 2 72 68Question 3 57 54Question 4 65 54Question 5 65 67Question 6 58 51

Male FemaleSoc Welfare

Question 1 58 53Question 2 72 70Question 3 63 60Question 4 60 51Question 5 72 66Question 6 55 48

Higher conventional response if transfer involves richest region (Q2 and Q5), is to poorest region (Q2); lower share if transfers among middle-ranking regions (Q1 and Q6)



Conclusions: Numerical Part• More male than female students view equalising transfer as

risk/inequality reducing, on each question separately

• Also true for consistency with Principle of Transfers (AAAAAA)

• Male/female differences are larger for risk than inequality

• Probit regressions confirm pattern while controlling for other personal characteristics

• Respondents are more likely to view equalising transfers as risk/inequality reducing when occurring from upper to lower end of distribution rather than ‘within’ the distribution

• The transfer type matters more for female and for risk

Main finding: gender matters for attitudes towards risk and inequality, and matters more for risk than inequality

Reminder of Verbal Part

The Verbal Responses: 1

All respondentsboth ineq risk

a 8.7% 8.7% 8.7%b 29.3% 30.6% 27.9%c 17.5% 17.3% 18.0%d 23.6% 22.8% 24.4%e 8.7% 8.5% 8.9%

at least d 27.9% 27.7% 28.1%

The (strict) orthodox answer is d, but less than a quarter of respondents chose that.Even the “lenient” orthodox answer gets the support of less than 30 percent of the respondents.The most favoured response (particularly for inequality) is the “Pigou” answer b.


malesboth ineq risk

a 8.3% 8.2% 8.3%b 30.0% 31.6% 28.2%c 14.6% 14.8% 14.4%d 24.9% 22.2% 27.9%e 9.1% 9.8% 8.3%

at least d 29.5% 27.8% 31.4%

Males answer more conventionally on risk than on inequality


femalesboth ineq risk

a 9.0% 8.7% 9.2%b 28.6% 29.6% 27.5%c 22.1% 20.9% 23.5%d 22.0% 24.3% 19.4%e 7.2% 5.3% 9.2%

at least d 25.9% 28.2% 23.5%

Females answer more conventionally on inequality than on risk


InequalityAll males females

a 8.7% 8.3% 8.7%b 30.6% 31.7% 29.6%c 17.0% 14.8% 20.9%d 22.8% 22.2% 24.3%e 8.5% 9.8% 5.3%

at least d 27.7% 27.8% 28.2%

Females answer more conventionally than males on inequality!


RiskAll males females

a 8.7% 8.3% 9.2%b 27.9% 28.2% 27.5%c 18.0 % 14.4% 23.5%d 24.4% 27.8% 19.4%e 8.9% 8.3% 9.2%

at least d 28.1% 31.4% 23.5%

Males answer more conventionally than females on risk

Probit Regression of Response d

P>χ2

1041

4916574

Coef P>|z|-0.02 560.01 810.01 1-0.05 18-0.01 730.00 94-0.02 23

InequalityVariable


Number of observations 1153.Explanatory variables include dummy variables for countries.

For the orthodox answer on the verbal question gender is not significant, but economic subject is important for risk

Coef P>|z|0.07 90.18 0-0.00 38-0.02 69-0.03 30.01 490.04 4

Risk



Conclusions: Verbal Part

• People favour the “Pigou-but-not-Dalton” response

• The conventional interpretation of the transfer principle is made by less than 25 percent of respondents

• Gender is not such a good predictor for orthodoxy, but there are remarkable differences in gender responses as between inequality and risk.

Where next?

• Clearly we could apply the same methodology to poverty orderings

• When we do we find even less “support” for the transfer principle. •People see it in terms of

a straight “headcount” approach

•People see it in terms of a straight “headcount” approach

• We can also examine other individual axioms in both inequality and poverty.

• But maybe we should do something more fundamental

Where next?

Are the results an artefact

• …of the context?

• …of the respondent population?

• …or of the design?

• … or of the medium?

• No – see the results on Social welfare and Poverty

• No – see the results on Social welfare and Poverty

• More difficult. But we have plans for ordinary - and extraordinary - people

• More difficult. But we have plans for ordinary - and extraordinary - people

• Our next move…•…to the Internet

• Our next move…•…to the Internet

Characteristics by Gender: Inequality

Male students are older, have more work experience, are more conservative, and have a higher family income in 1990

Economic SubjectAgeEmploymentPolitical OpinionIncome 1990Income Change 2010

750022

11

P-value

0.30 0.4623.4 5.00.5 0.54.1 1.14.4 1.00.8 1.2

Mean SDMale

0.28 0.4522.1 3.30.3 0.53.9 1.04.2 1.10.6 1.3

Mean SDFemale





Characteristics by Gender: Risk

(Broadly same as for inequality subsample)

Economic Subject 0.33 0.47 0.27 0.44 10Age 23.8 4.5 22.3 2.8 0Employment 0.48 0.50 0.36 0.48 0Political Opinion 4.0 1.1 3.9 1.0 8Income 1990 4.3 1.2 4.2 0.9 66Income Change 2010 0.73 1.4 0.48 1.2 0

P-value*Mean SD Mean SD

Male Female

attitudes to inequality and povertydarp.lse.ac.uk/subjective/presentation.pdf · dalton’s...

Documents