Labour Empirics (useful) 2 Flashcards
Signalling model
Spence 1973. Explains how educ is merely a signal of high productivity, rather than causing it.
Tests of signalling
Supported!
Altonji and Pierret, 2001: E gets more credit before much experience! Indicates power of signal dies down.
Arcadino et al (2010): College grads have already successfully signalled ability so marginal effect of experience does not rise with exper!
Sheepskin effects?
Jaeger and Page, 1996: Controlling for years of schooling, still a 22% wage premia for BA
Curriculum effects under signalling?
Quality of educ shouldn’t matter if signalling. These studies face self selection issues.
Altonji, 1995: Returns to extra HS courses small
Rose and Betts, 2001: Large math effects on earnings
Kane and Rouse, 1995: If you fail a degree, number of credits matters!
GED
High school equivalent qualification for dropouts (assume it is a good ability measure).
GED test for Spence!
Tyler, Murrane, Willet, 2000
Exploit variation in passing standards. Use DiD to find variation in identically abled students. Find GED has a 19% impact on wages
Challenge to TMW 2000
Jepsen et al, 2016
As individuals can retake, it is not a valid abil control. If we only allow first try, no effect!
Compulsory school attendace laws.
Lang and Kropp 2016
If pure HC, only marginal decisions affected. If Spence, all affected as value of signal down. Appears pos effect for all, but stat not significant
Hysteresis
Kahn, 2010: Graduating in a weak economy leads to wage loss as high as 2.5% 15 years later.
Benefits and job search
Card, Chetty, Weber 2007. Regression discontinuity. Extension of potential duration of benfits lowers finding rate in first 20 weeks by 5-9%. This is because length of potential benefits depends positively on number of months employed in last 5 years
Katz and Meyer, 1990 Far higher chance of finding a job the week benefits expire
Min Wage Diff in Diff
Card and Krueger (1994)
DiD for changes to min wage Pennsylvania vs New Jersey (min wage here). Assume parallel trends. Suggests Ld elasticity of +0.7. Specific industries!
Issues with CK 94
- Measurement errors: Phone survey vs actual administrative data.
- Fast food not representative? (hungry teenager theory) (Keenan, 1995). Teenagers employed and stim demand.
- Adjustment takes time! (Baker et al, 1999)
Seattle min wage
We see politicised debate! UC Berkeley study backed by mayor’s office so incentive to support min wage.
Key of debate is elasticity of Ld close to min wage barrier
Taste based discrimination
Becker, 1957
Employer has a discrim coefficient and acts as if wage = (1+a)w
1 - If a is common, then B members pay a ‘discrimination tax’ and so wage down.
2 - If a is from a distribution, then the a at which there is indifference rules wage differential.
Implies discrim will be driven out as not profitable!
Must consider wage differentials given same vector of productivity characteristics
Cain, 1986 on discrimination. Consider Oaxaca Blinder Decomposition
Statistical discrimination.
Phelps (1972) / Arrow (1973)
Observable which doesn’t directly matter (eg race) is corr with performance (via unobservables).
Set up using a noisy evaluation process.
Implies return to training is lower for B types! Self fulfilling?
Empirics on race differentials are generated how?
Correspondence: Fictitious paper applications.
Audits: Trained auditors for job interviews
Audit and critiques of audit studies
Pager, Western, Bonikowski (2009)
Blacks less likely to get a callback offer. B roughly equal to Whites with 18 month drug felony
Heckmann Siegelmann, 1992. Assumes unobservables identical across groups and not double blind so unconcious actions by auditors? Sample sizes inevitably small due to cost
Visible hand
Doleac and Stein (2013)
iPod sales with hand visible. Black sellers: Fewer and lower offers. Largest difference in thin markets / more racial isolation so appears consistent with this channel
Correspondence
Bertrand and Mullaithan (2004). Random characteristics and change name. Roughly 50% gap in B/W callback rate!
Fryer and Levitt (2004). No compelling effect of life outcomes of black names after we control for background. Suggests perhaps little LR effect?
Oaxaca - Blinder
Decompose differences in wages into due to discrimination and due to skills gap
Return to educ lower for blacks due to school quality, not discrimination?
Card and Krueger 1992
Should we control for industry / occupation?
Yes? Should make comparisons across equivalent occupations/ industries!
No? Choices are at least partly endogenous!
B/W wage ratio over time
US
Men: 1940=0.4 vs 2009=0.77
Sig closer for women
Reasons for closing gap in B/W wage ratio?
Closing gap in educ: E.g. end of seg Brown v Board of Educ 1954.
Biased employment tests abolished?
Affirmative action?
Labour force participation?
Unobserved skill differences?
Biased employment tests abolished?
Affirmative action?
Black police wages up 10% following abolished tests
CRA (1964): Strengthened by executive order mandating blind treatment / deliberate favouring eg in college applications!
Labour force participation?
Unobserved skill differences?
- Roughly equal 1960s to 6% gap 2015. Wr up leads to avg w up no change in discrim (1/3 of differential?)
- Neal and Johnson 1996: 3/4 of wage difference is due to human capital, following work by Arrow, 1973: stat discrim
Oaxaca Blinder leads to wage differential explained due to human capital
O’neill O’neill 2006
Similar results to NJ 1996: Wage ratio is very small when we control for AFQT score
OVB? AFQT is not a pure abil measure. Childhood educ and race! - Household income?
Gender wage gap UK
39% 1978 / 14% 2022
Manning and Swaffield, 2008: Gender pay gap on entry = 0, but 25% 10 years after entry. 17% is characteristics, 8% discrim.
Characteristics and gender wage gap in UK
Manning Swaffield 2008
Human capital: 11% / Occupational crowding 1.5% / Psychology 4.5%
Human capital and gender pay gap
Mincer and Polachek, 1974 human capital model:
- Div of labour in family means women gain less experience.
- Women: anticipate shorter / less cont. work lives so less incentive to invest in training.
- Also, employers E(leave) is higher so less willing to hire
Occupational crowding and gender pay gap
- In female dominated fields, all wages 14% lower.
- Manning and Petrolongo, 2008: ‘Part time pay penalty’: 14% to 28% 1975-95. Disappears when controlling for occupation.
- Connolly and Gregory (2009): FT to PT. ‘hidden brain drain’, with many women moving to lower skill jobs.
Part time pay penalty has risen and women crowd into these areas!
Psychology and gender pay gap
Gneeze, Niederle, Rustichini (2003): Women vs men compete. Women perform worse if mixed! Men better at competing men vs women?
Negotiations:
- Babcock, 2002: 7% of women vs 57% of men negotiate wage (Carnegie Mellon MBA students)
- Babcock and Laschever, in lab experiment: 9x as many men vs women asked to be paid more
Income pooling?
Lundberg, Pollack, Wales (1997): Change of child benefit from father to mother. Ratio of children’s clothes to men’s up (but time trend?)
Duflo, 2003
Black SA can receive pensions 1993. Both weight for height and height for age significantly up for girls if Grandmother pension, not if grandfather
Thomas, 1990
Brazil 1974/75: Effect of maternal income on nutrition demand is 4-7x larger vs paternal!
Canonical model of skills.
Tinbergen, 1974
Leads to skill premium: Log(Wh/WL)= ((theta-1)/theta)*log(Ah/AL) - (1/theta)log(H/L).
Supply and demand side!!
Test Tinbergen
Katz and Murphy 1992: Data in period: 1963-87. theta hat = 1.6 and annual 2.7% increase in rel demand for college labour.
Acemoglu and Autor, 2012. KM breaks down since 2000! Extend to 2008. theta hat = 2.4, annual increase = 1.6
Means: Low, High skill more substitutable and rel demand for college labour increasing more slowly!
Breakdown of Tinbergen
Cannot explain fall in real wage as Ah (prod) up and cannot explain the U shaped changes in wage dist since 1990s as just low and high skills!
Non-cognitive skill measures and importance
‘Big 5’ personality dimensions: Extraversion, agreeablenesss etc
Bottom quartile are only 1/3 as likely to complete degree vs top
Team production model
Deming, 2017.
If costless trade: Result similar to Acemoglu, Autor (2012): Workers trade tasks, with I* by normal task based model
Costs of trade over tasks
Uses Krugman, 1991: Iceberg transport costs! Trading one unit leaves S<1 unit remaining.
Empirics of non-cognitives / social
Comparing change from 1979 to 1997 NSLY: 1 SD cognitive on wage 20.3% to 15.1% and 1SD social: 2% to 3.7%
Generally, returns increasing in both cognitive and non so complements!
Increasing marginal effect of social skills since 1979 if we look at NSLY!
Task based model conditions?
Equal expenditure / Law of one price for skills / perfect comp / markets clear / no arbitrage at critical task w*j(i) common here!
