karsten staehr. minimum wages and the wage distribution in estonia
TRANSCRIPT
\16minw-show3.doc 1
Open seminar, Eesti Pank
20 September 2016
Minimum Wages and the Wage Distribution
in Estonia �
KARSTEN STAEHR
Tallinn University of Technology, Eesti Pank
All opinions personal!
Simona Ferraro, Jaanika Meriküll & Karsten Staehr (2016): “Minimum wages and
the wage distribution in Estonia”, Working Papers of Eesti Pank, no. 6/2016
\16minw-show3.doc 2
Menu
1. Introduction
2. Results from literature
3. Methodology
4. Data and summary statistics
5. Estimation results
6. Final comments
NB: Positive / descriptive analysis!
\16minw-show3.doc 3
1. Introduction
“The new normal” � distributional concerns
� Piketty (2014)
� ECB and monetary policy (Mersch, 2014)
� IMF and fiscal policy (Dabla-Norris et al., 2015)
� Changes to person income taxation in EE and LV
Minimum wages
� Politically contested topic in USA, UK
� … and recently Germany
� IMF (2016): “Cross-country report on minimum wages”, IMF Country Report, no.
16/151 (http://www.imf.org/external/pubs/ft/scr/2016/cr16151.pdf)
� Latvia, Lithuania, Poland and Romania
\16minw-show3.doc 4
Questions
a) How do minimum wages affect employment?
b) How do changes in minimum wages affect wage distribution?
b1) How do minimum wages affect wages for wage-earners directly affected by
changes?
b2) How do minimum wages affect wages for wage-earners not directly affected,
i.e. above changed minimum wage?
~ Spill-over or ripple effect
�
Effect on average wage depends on spill-over effects ⇒⇒⇒⇒ macroeconomic
implications
This paper
� Address b2) in isolation!
� How do changes in the minimum wage affect wages at different percentiles of the
wage distribution at or above the changed minimum wage?
Use (modified version of) standard methodology � Lee (1999)
\16minw-show3.doc 5
Contribution
� Estonia � post-communist (until now only detailed studies for Ukraine, Russia)
� EU country from CEE
� Market-oriented, flat tax, low social transfers, little collective bargaining, rather
wide wage and income distributions
� Consider pre-crisis, crisis, post-crisis periods separately ☺☺☺☺
\16minw-show3.doc 6
2. Results from literature
Methodological complication
� In given period � “everybody” typically face same minimum wage (“same
treatment”)
� Changes from period to period ⇒⇒⇒⇒ weak identification
Methods
� Early studies � plots of wage distributions before and after
� From mid-1990s � semi-parametric methods
� Lee (1999) � smart identification strategy & econometrics
� Various other methods
\16minw-show3.doc 7
Results
USA
� Spill-over effects of minimum wage up to 25th
percentile
� Gradual decline of the real value of the federal minimum wage ⇒⇒⇒⇒ lower tail
inequality ↑ (DiNardo et al. 1996, Lee 1999, Autor et al. 2016)
UK
� Generally small or no spill-over effects (Stewart 2012, Dickens & Manning 2004b)
Continental Europe
� Few studies (no minimum wage in many countries)
Post-communist / emerging markets
� Mexico (Bosch & Manacorda 2010), Vietnam (Hansen et al. forthcoming) �
substantial spill-over effects
� Ukraine � large spill-over effects, largest for women (Ganguli & Terrell 2006,
JCE)
� Russia � large spill-over effects, largest for women (Lukiyanova 2011, NSE)
\16minw-show3.doc 8
3. Methodology
Identification problem � lack of variation in minimum wage across wage earners
� Cross-sectional dimension
� Time dimension
Lee (1999)
Consider various “labour markets” / “cells” �
� Lee’s labour market / cell � state, year
� Our labour market / cell � region, year, sector
\16minw-show3.doc 9
Wage distribution differs across cells �
Effect of minimum wage on wage distribution depends on size of the minimum wage
relative to the wages in the particular cell:
� If minimum wage high relative to wage distribution in cell ⇒⇒⇒⇒ binding or effective
for many ���� large effect on wage distribution
� If minimum wage low relative to wage distribution in cell ⇒⇒⇒⇒ binding or effective
for only few ���� little effect on wage distribution
Measure of “bindingness” or effectiveness of minimum wage in cell
=
Minimum wage – median wage in cell (< 0)
=
“Effective minimum wage” in cell
Median wage in cell � measure of wage distribution in cell
Identifying assumption � the median wage (and above) in cells not affected by the
minimum wage
\16minw-show3.doc 10
Estimations
Find effect of the “effective minimum wage” on various percentiles of the wage
distribution
Our empirical model
ijtijttijttijtpijt wwwwww ε++−β+−β=− controls)()( 250
250
150
� i = region, j = sector, t = year
� pijtw = p-percentile of log wage in region i, sector j and year t
� 50ijtw = median log wage
� tw = log minimum wage in year t
NB: Run regression for any percentile p
� #observations = #regions × # sectors × #years
[Intuition � increase of wt]
\16minw-show3.doc 11
Controls � year dummies, regional dummies, GDP growth and unemployment rate
� Hopefully remove effects of other factors
� Quadratic terms allow for non-linear relationship � compute marginal effect at
averages of explanatory variables
\16minw-show3.doc 12
4. Data and summary statistics
Statutory minimum wage
� In principle set in tri-partite negotiations � in practice government has final say
0
100
200
300
400
02 04 06 08 10 12 140
100
200
300
400
Figure 0: Pre-tax minimum wage for full-time wage-earner, EUR per month
\16minw-show3.doc 13
Estonian Labour Force Survey (LFS)
� 2001-2014
� No panel
� Only full-time wage earners (e.g. self-employed excluded)
� 6000-7000 observations per year � in total 91,447 observations
� Wage net-of-tax
� Other information used
� 5 regions (including counties), 11 sectors for creating cells
� Gender, age for sample splits
� Each cell (region, sector, year) � at least 20 wage-earners
\16minw-show3.doc 14
NB1: All wages net of tax
NB2: LFS wage data ≠ Eesti Statistika wage data
\16minw-show3.doc 15
5. Estimation results
� For whole sample
� For various subgroups
� Males vs. females
� Age 45 or less vs. age above 45
� Boom 2001-2007, crisis 2008-2010, recovery 2011-2014
Percentiles p = 5, 10, 15, 20, 25, 30, 40 (and for checking: 60, 70, 80, 90)
Empirical notation:
� minw – p50 = log minimum wage – median log wage = effective minimum wage
� p5 – p50 = log wage at 5th
percentile – median log wage
� p10 – p50 = log wage at 10th
percentile – median log wage
� …
\16minw-show3.doc 16
\16minw-show3.doc 17
Marginal
effects
evaluated at
means of
explanatory
variables
\16minw-show3.doc 18
Marginal effect = percentage change in wage (at given percentile of wage
distribution) when minimum wage increases by 1 percent
\16minw-show3.doc 19
NB: Percentage change at different log wage levels!
� Next slide � marginal effects in euros!
\16minw-show3.doc 20
Wages = net-of-tax wages
\16minw-show3.doc 21
Calculate effect on average wage
� Marginal effects at different percentiles
� Average wage at different percentiles
� Number of persons in interval around each percentile �
Minimum wage ↑ € 1 ⇒⇒⇒⇒ average wage ↑ € 0.11
\16minw-show3.doc 22
NB: Wage distributions very different for men and women, for young and grown-ups
\16minw-show3.doc 23
Effect of minimum wage seems to be smaller during crisis than before and after
\16minw-show3.doc 24
6. Final comments
Standard exercise using standard methodology!
� But augmented with sectoral dimension for identification
Results
� Fairly large spill-over effects, at least to 20th
percentile
� Stronger spill-over effects for women than for men and for older than young
(reflecting different wage distributions)
� Weaker spill-over effects during crisis than during boom and recovery
In euros
� Substantial spill-over effects even beyond 20th
percentile
� Minimum wage ↑ € 1 ⇒⇒⇒⇒ average wage ↑ € 0.11 ☺☺☺☺ / ����
\16minw-show3.doc 25
Why large spill-over effects in Estonia?
� Minimum wage main collective wage setting mechanism
� Great awareness � negotiations, press, in time for wage adjustments in beginning
of year
� Indexation of fees and prices � kindergarten, child support, traffic fines
Overall conclusion � minimum wage seems to lift wages in lower tail of wage
distribution ⇒⇒⇒⇒ potential to affect wage distribution
\16minw-show3.doc 26
Extra slide