کمک های جستجو و سرمایه انسانی برای رشد درآمد در طول چرخه زندگی

This paper presents and estimates a unified model where both human capital investment and job search are endogenized. This unification enables us to quantify the relative contributions of each mechanism to life cycle earnings growth, while investigating potential interactions between human capital investment and job search. Within the unified framework, the expectation of rising rental rates of human capital through job search gives workers more incentive to invest in human capital. In addition, unemployed workers reduce their reservation rental rates and increase their search effort to leave unemployment quickly to take advantage of human capital accumulation on the job. The results show both forces are important for earnings growth and the interactions are substantial: human capital accumulation accounts for 50% of total earnings growth, job search accounts for 20%, and the remaining 30% is due to the interactions of the two.

A well established fact in labor economics is that the life cycle earnings profile is increasing and concave. In their review, Rubinstein and Weiss (2006) discuss three leading sources for this pattern of earnings growth: human capital accumulation, job search and learning about job, worker or match quality. Human capital theory argues that workers invest in human capital when they are young thus forgoing earnings and reaping the returns to investment when they become old. Search theory argues that workers climb up a job ladder, moving from low-paying to high-paying jobs. When they are young, workers are more likely to be in the lower tail of the earnings distribution. This triggers job-to-job mobility associated with higher growth. As they age, the chance of accepting better outside options declines and fewer job-to-job transitions and lower growth results. With learning earnings on the job change as information is revealed and workers move from poor matches to better ones resulting in across job earnings growth as well. Once the information has been revealed and a good match attained growth subsides. All three of these explanations have been studied extensively on their own and in isolation each can reproduce the observed shape of the life cycle earnings profile if not the full amount of growth. Recently a new literature on quantifying the relative contributions of these sources of life cycle earnings growth has developed. For the most part this new literature has focussed on modeling the combination of human capital accumulation and job search.1 Understanding the relative contributions of human capital accumulation and job search to life cycle earnings growth is important since they have different policy implications concerning training on the one hand and labor market mobility on the other hand. With very few exceptions this new literature uses structural models that treat the human capital accumulation process as deterministic through an exogenous learning-by-doing framework. Furthermore, they also commonly treat the search process as exogenous with constant job offer arrival rates. There are two papers that treat the search process as exogenous but endogenize the human capital accumulation process. In a largely conceptual rather than quantitative analysis, Rubinstein and Weiss (2006) endogenizes human capital using a Ben-Porath style investment function. Michelacci and Pijoan-Mas (2012) endogenizes the hours decision within a learning-by-doing human capital accumulation model with exogenous search. However, their focus is on inequality and not life cycle earnings growth. A common result in this literature is that human capital is found to explain more earnings growth over the life cycle than search. In contrast to this prior literature, we develop and estimate a unified model where both human capital investment and search effort are endogenized to quantitatively examine the relative contributions of both mechanisms and their potential interactions to earnings growth over the life cycle. The decision-making in terms of human capital investment and job search are likely to be different within a unified framework. Consider a unified model where workers, facing a distribution of the rental rate of human capital, decide how much time to invest in general human capital within a Ben-Porath style investment model and how much effort to spend searching for a better job. In this setting, there are likely three interactions between human capital accumulation and job search. First, workers will likely invest more in human capital than they would without job search and with only a fixed rental rate of human capital. This is due to the upward drift in the distribution of the rental rate of human capital, inherent in the search model. Second, workers will likely spend more effort searching with human capital accumulation than without. This is because, without human capital accumulation, the return to search is only realized for a fixed level of human capital. With human capital accumulation, the return to search is greater since it is now realized for growing human capital. Third, because of human capital accumulation on the job, workers will likely reduce their reservation rental rate while unemployed in order to get a job to start accumulating human capital.2 For the most part the existing literature has been able to identify only a subset of these interactions.3 In addition to quantifying the full extent of the interactions between human capital investment and job search effort over the life cycle, we examine whether allowing for the interactions within a unified model changes the implications for earnings growth. Since Rubinstein and Weiss (2006) established that the amount of human capital accumulated with (exogenous) job search and without is different, it follows from the arguments above that the amount of human capital accumulated will be different if the job search process is then endogenized. Furthermore, the job search process will also change if it is endogenized. Mortensen (2003) shows that search effort is a decreasing function of wages. The workers who earn higher wages search less because the chance for them to climb further up the job ladder is smaller. It is reasonable to believe that workers with different human capital levels also have different incentives to search. Unlike the existing literature, in our model search effort depends not only on the current rental rate but also on the level of human capital. In this way, the job offer arrival rates differ across worker attributes including human capital levels, rental rates, experience, and age.4 Both of these factors will contribute to job search and human capital accumulation having different roles over the life cycle and potentially a different contribution to overall earnings growth. To conduct this study we combine a partial equilibrium search model with a human capital investment model.5 The search component of the model includes a search effort decision that determines the arrival rate of job offers both on and off the job. As is standard in search models reservation strategies are used to determine optimal transition patterns. In contrast to standard search models, here the reservation strategies depend on the level of human capital as well as current earnings and employment states. In the model human capital is governed by a Ben-Porath investment model where workers spend some of their working time investing in human capital and thus trade off current earnings for future growth in earnings. We chose an investment model for several reasons. First, it is arguably the most common human capital accumulation specification in labor economics. Second, it allows for the human capital accumulation process to be a function of the current rental rate as well as expected future rental rates derived through search. Third, there is evidence that the learning-by-doing model can be rejected by the data in favor of an investment-style model (see, for example, Belley, 2012). Finally, it allows us to contrast our results with the other work in this area that assumes a learning-by-doing human capital accumulation process. We estimate two extended versions of the model. In an attempt to better fit the transition data we first augment the model with unobserved heterogeneity in the search parameters (see Liu, 2009 for details). Then, in order to better match earnings growth on the job, we add unobserved heterogeneity in the ability to accumulate human capital. We estimate both versions of the model using indirect inference and using a relatively homogeneous cohort from the 1979 National Longitudinal Survey of Youth (NLSY), with the aid of additional information from all ages in the Survey of Income and Program Participation (SIPP). We use two data sets because Monte Carlo simulations of the estimation procedure showed that the finite-sample properties of the structural parameter estimates rely on how much transition and earnings information is available over the full life cycle. In particular, the Monte Carlo exercises show that the precision of estimates from using information from partial histories (e.g. the first half of the life cycle) can be improved substantially with additional information on transitions and earnings from the second half of the life cycle.6 This is because earnings and transitions for older workers provide information for identifying search parameters due to their limited human capital investment. Our preferred model is the one with both search and human capital heterogeneity. We show that it fits many of the transition and earnings growth features of the data over the life cycle. Based on the estimates from this model, the interactions between human capital accumulation and job search are well supported by the data. Because of job search, on average, workers invest more in human capital throughout the life cycle: 54% more at the beginning of the life cycle than without job search. Because of human capital accumulation, on average, workers reduce their reservation rates while unemployed throughout the life cycle and put more effort into search to generate higher offer job arrival rates while unemployed and on the job. Furthermore, counterfactual experiments show that a pure human capital model can generate around 50% of the total earnings growth, while a search model without human capital accumulation can generate around 20% of the total earnings growth. The remaining 30% is due to the interactions. Over the life cycle the strength of the components differs with job search and human capital equally contributing to earnings growth early in the life cycle, and human capital accumulation dominating later in the life cycle. Our findings are similar in magnitude and ordering to the rest of the decomposition literature, with the major exception that endogenizing both components yields an important role for the interaction effects of each component on the other. The rest of the paper is organized as follows. The model without unobserved heterogeneity is presented in Section 2. Section 3 discusses the identification and estimation strategies. Details on sample selection and construction of labor market histories are presented in Section 4. Section 5 discusses estimation results and model fit for the two extended versions of the model. Simulations of individual behavior and counterfactual experiments that examine the interactions and quantify the relative contributions of each mechanism to life cycle earnings growth are presented in Section 6. Section 7 concludes.

This paper presents and estimates a life cycle model that endogenizes both human capital investment and search effort to examine the interactions between them and quantify the relative contribution of each mechanism to earnings growth over the life cycle. Two notable interactions are (1) the expectation of rising rental rates due to job search over the life cycle induces more investment in human capital almost throughout the entire life cycle, and (2) because of human capital accumulation, workers reduce their reservation rental rates and increase their search effort to take advantage of human capital accumulation on the job. We estimate two versions of the model: one with unobserved search heterogeneity only and one with both unobserved search heterogeneity and unobserved human capital heterogeneity. We prefer the model with both types of unobserved heterogeneity, because it fits the data better on both the transition and duration dimension and the on the job earnings growth dimension than a model without unobserved heterogeneity in human capital. Importantly all four worker types are represented in our sample of white, male high school graduates. Workers with high search efficiency and those with high ability tend to invest more in human capital and search more intensively and thus have a higher earnings profile over the life cycle. Interestingly, the point estimate of the curvature parameter for the human capital production function is smaller compared to those found in the human capital literature. This is because job search generates some of the needed curvature to match the life cycle earnings profile. Based on the counterfactual experiments, a model with only heterogeneous human capital can generate 50% of the total earnings growth, while a heterogeneous search model without human capital accumulation can generate around 20% of the total earnings growth. The remaining 30% stems from interactions between search and human capital and indicates that this component is non-trivial. This paper adds to the existing literature on the decomposition of earnings growth between job search and human capital in two respects. First, unlike most of the existing literature, the model pairs search heterogeneity with human capital heterogeneity. The existing studies only allow for heterogeneity in human capital in terms of productivity levels but not in terms of human capital accumulation or in terms of search. We find that the inclusion of both types of heterogeneity is important in matching many features of the data including duration dependence, on the job earnings growth and cross-correlations between transitions and earnings growth. These results indicate that there are workers capable of learning who are also more efficient in searching as well as workers who are not very good at either. However, the majority of workers are good at one but not the other. Hence, in some cases the presumable higher earnings growth for workers with high learning ability comes not only from more investment in human capital but also from more job-to-job transitions, while for other workers earnings growth is driven purely by human capital accumulation or search but not both. Finally, there is a group of workers for which very little earnings growth occurs either through search or human capital accumulation and this helps to explain some of the low earnings growth outcomes over the life cycle in the data. In frameworks where there is only heterogeneity in human capital but not in search, much more heterogeneity in human capital is needed to explain the wide variation in outcomes whereas in this model we have another dimension by which to generate different levels of earnings growth over the life cycle. Second, the interactions or the spill-over effects between human capital accumulation and job search are substantial and should be considered when decomposing earnings growth or disentangling the effect of one from the other. Without endogenizing both human capital investment and search effort, these interactions are likely to be wrongly allocated to either human capital or job search. Our results suggest that it is the role of job search that has been overestimated in the literature when such interactions have been excluded.