Thailand Law Journal 2009 Spring Issue 1 Volume 12

The Mincerian Equation

Our measure of earnings is monthly earnings in January of each year, including overtime pay and excluding bonuses. The stock of training at time t is defined as the sum of average monthly hours of training from 1998 to the calendar year t-1. Therefore, we associate in (3) wages in year t with the training undertaken up to year t-1, and treat the latter as a predetermined variable. As in the case of the training equations, we pool the available information and include in the vector of controls X firm dummies, a linear trend, and a quadratic trend. Since we estimate (3) using fixed-effects, the coefficients associated with time-invariant variables, such as education, gender, tenure, and experience in 1998, cannot be estimated. Moreover, the inclusion of trends precludes the identification of age effects.

Our results in Table 9 are organized in four columns. In the first and third columns we assume that unobserved and time-invariant talent does not affect the private returns to training and estimate (3) on the full sample and on the subsample of production workers using fixed-effects. In the second and fourth columns we allow private returns to training to be affected by unobserved talent and replace the observed years of education with the predicted years from the estimate of equation (4). By so doing, we combine fixed effects with instrumental variables.[FN18]

 Table 9. Fixed-Effects Regression of the Mincerian Equation, Full Sample and 

      Production Workers Only, 1998-2001. (robust standard errors within      

                    parentheses, p-values within brackets)                    

-------------------------------------------------------------------------------
Description           All Workers   All Workers     Production     Production  

                                     with Years      Workers      Workers with

                                    of Education                    Years of  

                                     Treated as                    Education  

                                     Endogenous                    Treated as  

                                                                   Endogenous  

-------------------------------------------------------------------------------
Sum of OJT from 1998    -0.003        -0.010**        -0.004       -0.013***  

  to Year t (SOJT)                                                            

                        (.002)         (.004)         (.003)         (.005)    

Sum of OFFJT from       -0.019*       -0.018*         -0.006          .002    

  1998 to Year t                                                              

  (SOFFJT)                                                                    

                        (.011)         (.011)         (.016)         (.013)    

Years of Education     0.0005**       0.001***       0.0006**       .0013***  

  * SOJT                                                                      

                        (.0002)       (.0003)        (.0003)        (.0004)    

Years of Education     0.0017**       0.0016*         0.0006         .00003    

  * SOFFJT                                                                    

                        (.0007)       (.0008)         (.010)         (.001)    

Wald Test for the                                                              

  Interactions of                                                              

  Training                                                                    

and Education           [.002]        [.0002]         [.004]         [.005]    

eln W                   0.0024         0.0022         0.0025         0.0016    

eSOJT                   [.000]         [.001]         [.005]         [.074]    

eln W                   0.0000         0.0009         0.0010         0.0024    

e SOFFJT                [.990]         [.680]         [.779]         [.569]    

                                                                              
Observations             4,536         4,481          2,979          2,943    

R2                     0.033         0.035          0.018          0.019    

-------------------------------------------------------------------------------
   Notes: See notes to Table 5. Each regression includes a linear and a  quadratic trend.    

[FN18]. As suggested by a referee, we also experimented with a non-linear specification of the wage equation, using the square root of training rather than training. This trial produced no substantial qualitative change in the results.

This article is published with the kind permission of Kenn Ariga and Giorgio Brunello. The article originally appeared in Volume 59, Issue 4, July 2006, of the Industrial and Labor Relations Review. Copyright Cornell University.