شرح صاحبان ثروت گروه های مختلف: رشد بهره وری و تامین اجتماعی

In this paper, we explore the reasons why different generations accumulate different amounts of wealth. We use basic economic theory to propose two indicators of the economic conditions under which households accumulate wealth. The first one represents productivity differences across cohorts: The aggregate level of gross national product per capita around the time the head of the household entered the labor market. The second measure summarizes the changes in Social Security during the head of household's working life. Using panel data from the Dutch Socio-Economic Panel, we show that productivity growth can explain all the cohort effects present in income data, while productivity growth and the generosity of Social Security can explain all the cohort effects present in household net worth. We also find a limited offset of Social Security on wealth holdings.

There exists an important debate in the literature on the determinants of saving and wealth accumulation and on how one can explain, for example, the sharp decline in saving that many developed countries witnessed during the 1980s. Some researchers have argued that it is simply the aging of the population that has caused the change in saving. These might be called age effects. Others have argued that people coming of age in different times have different preferences. They argue, for example, that generations born after the Great Depression are less thrifty or less alert to risk than previous generations. An alternative view is that preferences may be identical across cohorts, but that the economic conditions of the past are very different from the present. Whether it is preferences or economic conditions, these considerations lead to the supposition of cohort or generation effects. Yet another group of researchers have argued that it is the capital gains in the stock market and the housing market that explain the movement in saving. These might be called time effects. While all these theories have strengths and weaknesses, the critical issue is: How can we distinguish among age, cohort, and time effects in saving? 1 Empirically, one cannot disentangle age and cohort effects in wealth data in a single cross-section. Shorrocks (1975) was the first to point out that productivity growth creates differences in household wealth holdings across generations and one cannot simply assume that the elderly provide a good representation of the current young generation when they get older. Thus, solely on the basis of cross-sectional data, one cannot study issues such as whether the elderly decumulate wealth after retirement. A few authors have used time series of cross sections to study the behavior of wealth or saving.2 They estimate a wealth or saving equation as a function of age dummies (or a polynomial in age) and cohort dummies. Additionally, one would like to include time dummies, for example to allow for macro shocks. However, this introduces the identification problem mentioned above: Calendar time is simply equal to year of birth (cohort) plus age. Some authors, such as Attanasio (1998), simply acknowledge this identification problem and show that one can only identify the age profile of the changes in saving but not the age profile of saving itself. Others impose restrictions on the time dummies. The leading approach is the one of Deaton and Paxson 1994a and Deaton and Paxson 1994b in the context of the life-cycle permanent income hypothesis (LC-PIH) for consumption. They assume that the coefficients corresponding to the time dummies add up to zero and are orthogonal to a time trend. One possible justification for this assumption is that time effects are only due to macroshocks and average out over time. This assumption might be reasonable if the LC-PIH provides a good characterization of household behavior and one has panel data or a time series of cross-sections available with many waves. In our dataset, as is generally the case in existing empirical studies, the number of waves is modest (T=12). Furthermore, our sample period is characterized by dramatic changes in house and stock prices. Since net worth obviously depends on these prices, we have to rely on other identifying assumptions than the one suggested by Deaton and Paxson. We address the identification problem by exploiting the predictions of a fairly standard version of the LC-PIH to explicitly model the cohort effect.3 We show that productivity growth and changes in Social Security (SS) together can explain the differences in wealth across cohorts. Productivity differences generate differences in permanent income across cohorts, which then feed into the wealth accumulation of households belonging to different generations. Changes in SS provisions alter the time path of income over one's life cycle and hence affect the need to save for retirement. In our empirical work, productivity of a cohort is proxied by the level of real gross national product per capita (RGNPC) when the head of the household entered the labor market (which we take to be between age 16–25). The effect of SS is proxied by a measure summarizing the changes in the SS system during the head of household's life. The inclusion of productivity growth in the regression for household non-capital income renders cohort dummies statistically insignificant. Thus, our productivity measures appear to capture all cohort effects in income. In turn, permanent income, which include these productivity effects, and our constructed SS variable can explain all cohort differences in net worth. The advantages of using these measures rather than cohort dummies are several. First, we can determine more clearly the causes for the differences in income and wealth holdings across cohorts. While many potential reasons have been proposed for explaining these differences, simple cohort dummies cannot distinguish, for example, between differences in economic circumstances and differences in preferences. Second, the simple theoretical framework described in the paper highlights that it is very restrictive to use cohort dummies to model cohort effects, since it is easy to envisage cases where the effect is not simply additive, but, as we illustrate for the case of SS, there are interactions between cohort and age effects. Finally, cohort dummies can be rather difficult to interpret when some past economic conditions (productivity growth) lead to an increase in wealth across cohorts, while others (increases in the generosity of SS) lead to a decrease. For policy purposes, it may be very important to disentangle those effects in the data. Our simple economic framework as a basis for the empirical model allows for that. We estimate our model using panel data from the Netherlands. This is a country whose historical conditions are ideal to study the effects of productivity growth and SS. The Netherlands experienced a steady growth after World War II. At the same time, it also built up a generous welfare system.4 We find that past economic circumstances can explain the variation in paths of wealth accumulation across cohorts. Thus, we do not need to rely on differences in preferences, e.g. differences in thriftiness or impatience, to explain differences in wealth across cohorts. In particular, our empirical work reinforces the findings of others, such as Shorrocks (1975) and Feldstein 1974 and Feldstein 1996, that productivity growth and Social Security are important determinants of wealth. Furthermore, we find evidence of a limited offset of SS on wealth holdings. The paper is organized as follows: In Section 2, we present a simple theoretical framework in which we illustrate the effects of productivity growth and SS on wealth. In Section 3, we introduce the data set and describe the main features of wealth and income over the life cycle. In Section 4, we describe the econometric specification for after-tax household non-capital income and present our empirical results. In Section 5, we report the econometric specification and empirical results for wealth and in Section 6 we provide some brief conclusions.

We have examined the income and wealth holdings of different cohorts. While many explanations have been given for observed differences among cohorts, disentangling these explanations is difficult. For example, simple cohort dummies cannot distinguish between differences in preferences and differences in economic circumstances. Our strategy consists of devising indicators that can summarize the economic conditions over time. A good proxy for the differences in the income profiles of households is the level of aggregate income per capita around the time the head of the household entered the labor market. Similarly, we have devised a proxy for the changes in the SS system at the beginning of or during the working life of the head of the household. We find that the inclusion of a measure of productivity growth in our equations for income obviates the need to include cohort dummies. Similarly, productivity growth in permanent income and changes in SS can explain all cohort effects in wealth. Thus, past economic conditions can explain why generations differ in their wealth holdings and we do not need to resort to differences in preferences or other reasons to explain the differences in wealth holdings across generations. Obviously, there are other economic conditions that could be important for explaining the differences in wealth accumulation, including precautionary motives, longevity, changes in house prices, bequest motives, changes in demographics and female labor supply. Undoubtedly, these factors play a role. The strategy in this paper has been, however, to simplify the model of wealth accumulation as far as possible while still being able to explain important features of the data. The fact that our two proposed economic indicators for productivity growth and SS are effective in explaining the wealth holdings of different cohorts suggests that either these other factors are very strongly correlated with our indicators, or that the other factors are of secondary importance in explaining the wealth accumulation of different cohorts.