Presentation on theme: "The Returns to Experience and Job Tenure About how wages vary over the life-cycle."— Presentation transcript:
The Returns to Experience and Job Tenure About how wages vary over the life-cycle
The Returns to (Potential) Experience in the UK
Note Normalized on male log earnings at 0 years of experience being 0 Note gender difference – will return to this later Part of profile may be cohort effect but not most of it Mincer first drew attention to earnings profile Why did he use experience not age?
Mincer Specification Earnings Function of the form: Has more or less stood test of time (should perhaps be quartic not quadratic) E=(a-s) so implies interaction between a and s. Earnings profiles for different education groups should be parallel.
The Returns to Experience by Education (1 is high, 4 is low)
Mincer’s Interpretation Return to experience is return to general human capital Productivity rises with experience so wages do as well Why should productivity rise with experience? –learning-by-doing –Explicit investments in training Why do earnings of older workers decline? –Depreciation of human capital –Not worthwhile replacing it
Problems with Mincer Interpretation Mincer explanation probably has some truth in it but is it the whole story? Displaced workers – older workers who lose jobs through plant closures seem to suffer larger earnings losses Empirical evidence on whether older workers really have higher productivity a bit weak e.g. Medoff-Abraham QJE 1980
Part of this may be due to losses in specific human capital But suggestion that not all of it can be explained in this way
Search Explanations of Return to Experience If jobs vary in wages then older workers have had more chance to find the good jobs Can think of this as accumulation of search capital This can explain earnings losses of displaced workers – a loss of search capital But there almost certainly is some earnnigs growth within jobs
A More Formal Model Assume there are only two types of job, that we will call bad and good Wages are w g and w b If fraction of people in good jobs is θ then: So can express everything in terms of θ Note ‘true’ returns to experience and job tenure are zero
Assume that new job offers arrive at a rate λ A fraction α of new jobs are good ones Only workers in bad jobs will accept new job offers and they will only do it if the new job is a good one - workers in bad jobs will leave for good ones at a rate λα Jobs get destroyed at a rate δ – assume everyone gets a new job straight away (simpler)
Denote by θ(a) the fraction of workers of experience a who are in good jobs. The change in this fraction must equal the inflow of workers into good jobs minus the flow out of it. Using the assumptions on job offer arrival rates and job destruction rates described above we must then have that: In a long-run steady state we will have:
Can write differential equation as: With solution: i.e. θ(a) will be increasing in a but at a decreasing rate. This implies that average wages will be increasing in a but at a decreasing rate i.e. we can explain the concave earnings function we observe in the data. But what about down-turn for older workers – changing transition rates
The fraction in good jobs
The Returns to Job Tenure (UK)
Interpretation (Mincer-Jovanovic) Higher job tenure leads to greater accumulation of job-specific human capital Workers manage to extract some share of these returns This the return to job tenure
Other Interpretations May be efficient to offer delayed compensation: –Efficiency wages (Lazear, JPE 79) –Monopsony – deters quits Reverse causality – those in good jobs are less likely to leave Can investigate this using the search model we used above – effects quite subtle - a high level of job tenure indicates that: –the job has lasted a long time so is likely to be a good one –the job started a long time ago so is less likely to be a good one.
Empirical Evidence on Returns to Job Tenure and Experience Aim is to get estimate of ‘true’ return to experience and job tenure An estimate of how much earnings grow on the job is a joint combination of returns to experience and job tenure Get estimate of how much earnings increase on ‘new’ jobs – estimate of the return to experience
A Model Denote by w i (a,t) the log wage of an individual with experience a and job tenure t – we will just consider 2 periods Individual i enters the labour market with experience 0 and job tenure 0 and with wage – wage is w i (0,0) Can either stay in same job, S i =1 with wage w i (1,1) or move, S i =0, with wage w i (1,0)
Defining Returns to Experience and Job Tenure Return to experience will be: Return to job tenure will be: Problem is that only observe one wage at date 1 so cannot compute these returns directly Let’s consider OLS on cross-section data
The OLS Cross-Section Estimate The OLS estimate of the return to experience involves comparing the average wages among those with age one and tenure zero with the average wages among those with age zero and tenure zero. One way of writing the OLS estimate of return to experience is:
Can write this as: First term may be what you want to estimate but second is a bias term which is the correlation between the initial wage and whether you move jobs. It is likely that this bias is negative – with panel data one can generally show that it is.
One way of putting this is that the OLS estimate is the average return to experience if –the return to experience is uncorrelated with staying in jobs –the initial level of wages is uncorrelated with staying in jobs. What about the OLS estimate of the return to job tenure?
OLS estimate of return is given by: This again has biases caused by the potential correlation of initial wages and the returns to experience with mobility. How can we do better – panel data?
Estimates Using Panel data This is essentially the approach of Topel, JPE 91, and Altonji and Shakotko, ReStud, 97 (though these differ in details) Essentially estimate returns using data for whom it is available e.g. for experience use the movers:
And use the stayers to have: With return to tenure as: These are consistent estimates if there is no correlation between moving and the return to experience. Also this is average treatment effect if no correlation between return to tenure and return to experience.
Unresolved Problems Returns to experience and job tenure likely to be very heterogeneous There are ‘good’ job moves (quits) and ‘bad’ job moves (fires) with very different returns May be that those who have high wage growth remain in their job None of existing studies especially convincing