Printer Friendly

Precautionary saving by young immigrants and young natives.

1. Introduction and Literature Review

Foreign-born population accounted for approximately 9% of the total population residing in the United States by the mid-1990s and, including their children, they were responsible for approximately 51% of the U.S. population growth (Martin 1995). These dramatic changes in the U.S. population have prompted much economic research on the immigrant population. But despite the great deal of attention that has been paid to the labor market progress of immigrants, little is known about their saving and wealth accumulation patterns. (1)

Many studies requiring information on the saving behavior of immigrants simply assume that immigrant savings parallel that of natives (MaCurdy, Nechyba, and Bhattacharya 1998). However, it may be the case that immigrants behave differently, have different motives, different opportunities, and different needs with respect to saving and wealth accumulation vis-a-vis natives.

Most of the limited literature on the saving behavior of immigrants is inspired by European and Middle Eastern guest-worker migration and consequently is geared to the behavior of temporary migrants. (2) Three papers that focus on the saving behavior of immigrants in general are the papers by Galor and Stark (1990), Carroll, Rhee, and Rhee (1994), and Dustmann (1997). Galor and Stark note that immigrants and natives differ from one another in their migration probabilities. Because immigrants have ties to other countries, it is more likely that they emigrate relative to the native-born worker. (3) Using an overlapping-generations model, Galor and Stark show that the higher the probability of emigration, the higher is the level of savings. Hence, they predict immigrants save more than the native-born.

Dustmann models precautionary savings and assumes that immigrants maximize expected lifetime utility from consumption in the host and home country proportionally to the lifetime spent abroad and at home. (4) Under appropriate restrictions, (5) Dustmann shows that, if labor markets in the home and host countries are uncorrelated, immigrants will carry out more precautionary savings relative to a comparable native worker, because immigrants will be subject to greater income risk in the host country. (6) However, it is also possible in Dustmann's formulation to have immigrants engaging in less precautionary saving relative to native workers. For example, if economic conditions across the two countries are correlated, immigrants could be subject to less lifelong income risk than natives, given their ability to diversify away labor market risk by operating in two markets rather than just one, as in the case of natives. Additionally, if the risk preferences of natives and immigrants are different, then the precauti onary saving motive could be more or less pronounced in one group relative to the other, all else equal.

Carroll, Rhee, and Rhee (1994) use the Canadian Survey of Family Expenditures to compare saving behavior of immigrants to Canada from different regions of the world. They conclude that immigrant saving does not significantly differ by country of origin. However, they do find that recent immigrants save less than do Canadian-born, and that immigrants' saving behavior approaches the pattern of the Canadian-born over time.

Overall the theoretical literature tends to favor the hypothesis that immigrants carry out more precautionary savings than natives, which could result in immigrants accumulating more wealth than natives. Immigrants face greater uncertainty regarding their labor market participation, labor market progress, and health care coverage. These factors, along with geographic distance, may make it more difficult to obtain financial support from family members in times of necessity, requiring that immigrants maintain larger stocks of precautionary saving. But Dustmann's model, in consonance with a widely held view, also allows for immigrants to be more risk tolerant than natives since they are willing to undergo higher uncertainties and risks when they decide to migrate to a new country (Barsky et al. 1997). Furthermore, the covariance of labor market conditions across the home and host country labor market could result in lower precautionary saving and wealth accumulation in the case of immigrants.

The bottom line is that one can devise various arguments regarding the precautionary saving and wealth accumulation behavior of immigrants relative to native workers. Nonetheless, there is little evidence to date on those relative saving patterns. Knowledge of immigrants' behavior in this regard may generate insights not only into immigrants' wealth accumulation patterns, but also immigrants' responsiveness to changes in eligibility to welfare and income programs, medical assistance and educational services, and their fiscal impacts.

We attempt to gain insight into the saving behavior of immigrants versus natives by comparing the precautionary savings and wealth accumulation patterns of immigrants and natives using data from the 1979 Youth Cohort of the National Longitudinal Surveys (NLSY79). The analysis investigates whether immigrants accumulate more, less, or the same wealth as natives, and the factors driving their wealth accumulation patterns over time; in particular, whether they carry out precautionary saving to face income uncertainty.

A drawback to using the NLSY79 data is that confidentiality concerns prevent us from learning the country of origin of the immigrants in the survey. Therefore we cannot test propositions regarding the wealth accumulation behavior of immigrants as they relate to country of origin variables. For example, we cannot test hypotheses that relate wealth accumulation of immigrants to the covariance of conditions at home to conditions in the United States. However, there are advantages to using the NLSY79. We are able to follow immigrants and natives over time and, thereby, better separate out changes in wealth accumulation due to changing economic conditions and to household-specific heterogeneity. Furthermore, working with longitudinal data (in contrast to simply cross-sectional data) allows us to construct a time-varying, conditional measure of income uncertainty. This measure of income uncertainty fluctuates over time in accordance with the economic circumstances instead of relying on a constant, unconditional, su mmary measure of income uncertainty for the entire period under consideration. Finally, an additional bonus to using the NLSY79 is that we examine the behavior of young natives and young immigrants. Much of the literature on U.S. saving patterns has focused primarily on wealth accumulation patterns of mature men and women nearing retirement (Kazarosian 1997; Lusardi 1998). In contrast, the analysis of young households' accumulation patterns as they finish college, start their careers, raise families, buy homes, send their children to college, and start planning for retirement has not received much attention (Engelhardt 1998).

We first analyze the composition of young natives' and immigrants' assets and liabilities. Assuming a buffer-stock model of savings (Carroll and Samwick 1998), we then estimate the responsiveness of wealth accumulation to income uncertainty once we control for other factors, including personal characteristics and attitudes towards risk.

The paper proceeds as follows. Section 2 discusses the theoretical framework. A detailed description of the data, as well as some descriptive evidence on immigrants' and natives' wealth composition, are outlined in sections 3 and 4 of the paper. The empirical methodology and our findings are presented in sections 5 and 6, respectively. Section 7 concludes the paper with a summary of our findings.

2. The Model

The main challenge of precautionary saving models is that they do not allow for a well-defined, closed-form solution to be implemented empirically (Carroll and Samwick 1998). An exception is the work by Caballero (1990), who derives a closed-form solution using a constant absolute risk aversion utility function. Following Deaton (1991), Carroll (1992, 1997), and Lusardi (1998), we assume consumers have a wealth-to-permanent income ratio as a target. When wealth is above the target, consumption will exceed permanent income and wealth will fall. If wealth is below the target, permanent income will exceed consumption and wealth will rise. This "buffer-stock" model appears to be consistent with macro- and microeconomic evidence on savings (Carroll and Samwick 1998).

Wealth accumulation, captured by the household wealth-to-permanent income ratio, can be expressed as a function of personal characteristics of the household ([Z.sub.h]) and of income uncertainty facing such households ([[sigma].sub.h.sup.2] Therefore,

[W.sub.h]/[Y.sup.p.sub.h] = f([Z.sub.h], [[sigma].sup.2.sub.h]) + [e.sub.h]. (1)

The personal characteristics of the household include factors related to the household heads (Browning and Lusardi 1996; Lusardi 1998). Income uncertainty is proxied by the sum of the squared residuals of annual income regressions up to the year in question, normalized by permanent income. According to the buffer-stock model, as income uncertainty rises, individuals will accumulate more so as to maintain their desired wealth-to-permanent income target. The magnitude of the coefficient on the income uncertainty proxy will capture the extent of precautionary savings carried out by respondents. Differences between immigrants' and natives' wealth accumulation behavior may also result from different attitudes toward risk. We, therefore, control for the latter when examining respondents' wealth accumulation patterns.

3. The Data

We use microlevel data from the NLSY79. The longitudinal nature of the NLSY79 provides information regarding the time profile of natives' and immigrants' wealth accumulation instead of a snapshot, as a cross-sectional set would. In addition, by surveying 12,686 respondents ages 14 to 21 years old as of December 31, the NLSY79 constitutes a primary source of to learn about changes in income and wealth of young versus mature men and women nearing retirement (Kazarosian 1997; Lusardi 1998). This distinction is of interest since consumption and saving theories predict different wealth accumulation patterns for the young than for individuals over the age of 50 (Carroll 1992; Carroll 1997; Carroll and Samwick 1998).

To estimate our model, we require information on income, wealth, and various household characteristics. Our dependent variable, the wealth-to-permanent income ratio, captures the extent of wealth accumulation carried Out by individuals in the sample. From its first round, the NLSY79 contained information on income used to create net family income by the Center for Human Resource Research staff. (7) Net family income is first deflated to reflect its value in 1997 dollars and, second, used to compute a measure of permanent income, as will be discussed later in the paper. Detailed questions on assets and wealth were introduced in 1985. (8) The measure of wealth used in this empirical analysis is net household worth, computed as the difference between total assets and liabilities during the survey year. (9) Given the varying degree of liquidity of the assets in which respondents may invest and the difficulty of distinguishing between the consumption and investment value of some assets, such as main home property, we also use a liquid measure of wealth typically used in the literature in our final model estimation. The latter equals the monetary assets described in Appendix B. These assets can be liquidated on short notice with small transaction costs and, therefore, are expected to be more sensitive to uncertainty.

As is often the case with other surveys, data on wealth and income in the NLSY79 are top-coded for confidentiality reasons. Following a common practice among labor economists (e.g., Katz and Revenga 1989; Murphy and Welch 1992; Juhn, Murphy, and Pierce 1993; Buchinsky 1994), we multiply the top-coded value by 1.45. (10) Both measures of wealth are also deflated to reflect their value in 1997 dollars. Finally, the survey contains detailed data on respondents' demographic characteristics including educational attainment, labor market experience and work outcomes, family statistics, and health status.

This paper uses eight waves (1985-1990 and 1992-1993) of the NLSY79 along with some questions on personal background and characteristics (e.g., place of birth, race, and gender) asked in the 1979 wave. (11) Of the initial 12,686 respondents surveyed in 1979, 11,811 were bom in the United States or its territories and 874 were born outside the United States and its territories. Ninetyone percent of the 874 respondents born outside the United States were born from non-U.S.-born parents. Consequently, there were 796 immigrants or respondents born outside the United States and its territories from non-U.S.-born parents. By 1993, after discontinuing the subsample of economically disadvantaged non-blacks and non-Hispanic in 1990 and significantly reducing the military subsample because of funding constraints in that year, the NSLY79 sample counted with information on 9011 respondents (NLSY79 Users' Guide 1997, p. 36). (12)

The analysis controls for other personal, family, and work aspects that might alter the respondent's wealth accumulation patterns, such as the respondents' attitude toward risk. A risk-loving variable is created using an array of questions in the survey that addressed the respondent's attitude towards risk in the 1993 wave. The described risk variable coincides with that constructed by Lusardi (1998) and discussed by Barsky et al. (1997) using Health and Retirement Survey (HRS) data. (13) Table 1 provides the distribution of natives and immigrants according to their attitudes toward risk. Immigrants appear to display marginally more extreme attitudes toward risk, particularly with respect to risk loving, and the differences in the distributions are statistically significant.

To examine how wealth accumulation is affected by income uncertainty while controlling for attitudes toward risk, we first derive a measure of income uncertainty faced by the respondent in each of the surveyed years according to changing personal and job-related characteristics. (14) Our measure is similar in spirit to the one used by Kazarosian (1997). Using panel data, Kazarosian (1997) constructs the standard deviation of the residuals from household income regressions to proxy for each individual's income uncertainty. Our measure is the average sum, up to the year in consideration, of the squared residuals of annual regressions of the logarithm of income on the respondent's personal and job-related characteristics, health condition, and attitude toward risk, normalized by permanent income. (15) In this manner, we measure unexplained income as a fraction of predicted or explained income and, thus, capture the extent of unpredictability of income or income uncertainty in a time-varying fashion. Hence, as in Kazarosian (1997), we obtain measures of household income uncertainty that control for anticipated growth in household incomes. Our methodology, however, goes one step further in that we construct a time-varying measure of income uncertainty to gauge how precautionary saving by households responds to changes in uncertainty over time. Kazarosian (1997), instead, uses the overall summary measure of individual income uncertainty to examine how precautionary saving varies across the Older Men cohort of the NLS at one point in time.

Despite using a time-varying measure of income uncertainty, which supposes an improvement over constant summary measures of income uncertainty, this variable is still likely to be subject to measurement error. (16) Therefore, we instrument income uncertainty to obtain consistent coefficient estimates. (17) A detailed description of each of the variables used in the analysis is provided in Table 2. Table 3 contains their means and standard deviations.

4. Some Descriptive Evidence on Immigrants' and Natives' Wealth Accumulation Patterns

Table 4 displays the distributions of wealth by income quartile for 1985 and 1993. It seems that there is a systematic relation between wealth and income quartile with mean net worth rising with income. The marked differences between mean and median values for wealth suggest the existence of substantial wealth inequality. The latter is further substantiated by the worsening distribution of net worth for both immigrants and natives from 1985 through 1993. In the final year of our sample, 56% of natives' net worth and 52% of immigrants' net worth belonged to the highest income quartile, whereas the lowest income quartile owned only 7% and 9% of natives' and immigrants' wealth, respectively.

Ownership rates and median values of respondents' assets and liabilities at the beginning and end of our sample also differed, as shown in Table 5. In 1985, 19% of native and 14% of immigrant households reported home ownership. The median home value among native owners was $87,379, whereas that of their immigrant counterparts reached $116,505. By 1993, the percentage of respondents in both groups owning their homes had increased to approximately 50%, with median home values reaching $93,349 among natives and $140,023 among immigrants. An almost parallel rise in the percentage of respondents with mortgages and property-related debts is also observable. However, although the median value of property debts among natives rose along with the median value of their homes, immigrants actually reduced the median value of their mortgages during the 8-year period. A similar phenomenon is observed with respect to cars and other vehicle-related assets and debts, suggesting that immigrants possibly saved and paid off their debts to a greater extent than their native counterparts.

Greater saving by immigrants over time is also suggested by Table 6, which displays mean and median wealth values for natives and immigrants at the beginning and end of our sample. Although mean wealth values for natives exceeded those of immigrants in 1985, the pattern reversed by 1993, with immigrants displaying greater mean and median wealth values than their native counterparts at the end of the sample period regardless of civil status or family size. Additionally, as previously found in the literature, we also find that married households enjoy substantially higher median wealth than their single counterparts (Hurst, Luoh, and Stafford 1998). Finally, median wealth is lower for households with at least four children in 1985; however, by 1993, respondents with larger-sized households enjoyed higher mean and median wealth than smaller households.

5. Empirical Implementation

We want to learn about immigrant and native wealth relative to their permanent income, that is, wealth accumulation as explained by buffer-stock models of saving. Thus, we estimate Equation 1 using the Baltagi and Chang (1994) two-stage least-squares random-effects estimator for unbalanced panels. Since the dependent variable is wealth to permanent income, we first need to obtain a permanent income measure. We use an augmented Mincer's equation and regress the log of deflated household income on a set of demographic and work-related characteristics, including occupation, industry, tenure, tenure squared, previous unemployment experience, work limitations due to a health condition, and risk preferences. The results from the variance-components income regressions are contained in Table 7.

As in the case of wealth in Table 6, marital status and family size are often linked to income. The main difference between the two groups is the importance of gender and race in determining income for natives versus immigrants. As hypothesized by human capital theory, educational attainment, tenure, and unemployment experience resulting in obsolete human capital affect employees' earned income. Work limitations due to a health condition result in reduced incomes. Risk preferences affect the extent of income uncertainty people are willing to take, resulting in risk-loving individuals earning lower incomes than their risk-averse counterparts in our case. Finally, as in previous immigration studies, (18) we find that longer stays in the United States are linked to greater incomes for immigrants. This finding is typically argued to be due to 'U.S.-specific' skills acquired by immigrants over time, such as language proficiency, educational credentials, and information on the U.S. job market in the immigrant earni ngs literature (Borjas 1985). (19) We use the predictions from the income regressions to proxy permanent income and construct the wealth-to-permanent income ratios to be used when examining wealth accumulation patterns according to the buffer-stock model of saving.

In addition, to gauge the extent of precautionary wealth accumulation, we compute the average sum of the squared residuals from the annual income regressions and normalize it using permanent income to proxy for income uncertainty, as previously explained in section 3. Since this measure of income variance may be affected by considerable measurement error, we estimate the wealth regressions using two-stage least-squares random-effects estimator for unbalanced panels to obtain consistent coefficient estimates (Carroll and Samwick 1998). We instrument income uncertainty with the local unemployment rate along with the characteristics already included in Z. The instrumental-variables specification is econometrically identified by the exclusion of the local unemployment rate from the wealth regressions. (20) The results from the instrumental-variables specification are displayed in Appendix A. Overall, income uncertainty is lower among males, more educated and married, and those with larger households. Similarly, i n both groups, divorced and risk-loving individuals endure higher income uncertainty than their single and risk-averse counterparts. Higher local unemployment rates also result in less unexplained income and, thus, less income uncertainty for both natives and immigrants. (21) Finally, while income uncertainty significantly declines with age among natives, age does not significantly reduce income uncertainty among immigrants. Instead, income uncertainty is lower among immigrants with longer residence in the United States.

6. Do Immigrants and Natives Differ in Their Wealth Accumulation Patterns?

The results from the wealth to permanent income and the financial wealth regressions for both natives and immigrants using the Baltagi and Chang (1994) two-stage least-squares random-effects estimator for unbalanced panels are displayed in Table 8. The specifications include personal characteristics and risk preferences, as well as our measure of income uncertainty, (22) which varies according to the respondents' personal characteristics, attitudes towards risk, and job-related characteristics, such as work limitations due to a health condition and previous unemployment experience. (23) Columns two and four of Table 8 contain the results for the wealth-to-permanent income ratio regressions, whereas columns three and five display the results from the financial wealth regressions.

Net wealth-to-permanent income ratios significantly differ by age, education, race, marital status, and household size, as previously suggested by some of the descriptive statistics in Table 6. For instance, more educated natives, as well as older and married natives and immigrants, accumulate more net wealth than their less educated, younger, and single counterparts. However, blacks, separated and divorced natives, and natives with large families display lower net wealth-to-income ratios than do whites, single natives, or natives with small families, respectively. Finally, precautionary savings are statistically significantly different from zero for natives and not for immigrants. (24) In particular, a 10% increase in income uncertainty raises natives' net wealth-to-permanent income ratios by 0.056, which amounts to approximately 5.5% of the average net wealth-to-permanent income ratio for natives. (25,26)

By contrast, while still responsive to similar personal characteristics, such as age, gender, education, race, and marital status, both natives' and immigrants' financial wealth-to-permanent income appears to be quite sensitive to income uncertainty. As expected due to its greater liquidity, natives' responsiveness to income uncertainty, although still statistically significant, is greater in magnitude for financial than net wealth. In particular, a 10% increase in income uncertainty raises natives' financial wealth-to-permanent income ratios by 0.084. This represents 19.6% of the average financial wealth-to-permanent income ratio for natives versus the 5.5% increase in the case of net wealth. For immigrants, precautionary savings are statistically significantly different from zero only in the case of financial wealth. The same 10% increase in income uncertainty results in a 0.075 increase in their financial wealth-to-permanent income ratios, which is equivalent to 27.8% of the average financial wealth-to-per manent income ratio for immigrants. Note also that immigrants' financial wealth significantly increases with their time in the United States.

Do immigrants accumulate less, more, or the same wealth as natives? Using the predictions from the net wealth- and the financial wealth-to-permanent income ratios' regressions, we find that natives accumulate slightly more net wealth and financial wealth relative to immigrants with their same characteristics. More specifically, natives' actual average net wealth-to-permanent income ratio (27) is 1.03, whereas that of comparable immigrants is 1. Similarly, natives' actual average financial wealth-to-permanent income ratio is 0.43, whereas that of comparable immigrants is 0.27. One possible explanation for immigrants' lower wealth accumulation patterns may be their money remittances to their home countries. These income payments are likely to be counted by immigrant respondents as consumption in the NLSY, contributing to their lower wealth-to-permanent income ratios.

7. Conclusions

This paper uses data from the NLSY79 to learn about wealth accumulation patterns of younger cohorts and how they differ across natives and immigrants. In particular, the paper examines whether immigrants accumulate more, less, or the same wealth as natives and the factors driving their wealth accumulation pattern, such as whether they carry out any precautionary saving to face income uncertainty. We use two measures of wealth: net wealth and financial wealth, both measured as ratios to permanent income. Additionally, we construct a time-varying, conditional measure of income uncertainty instead of relying on a constant, unconditional, summary measure of income uncertainty for the entire period under examination.

The results reveal that immigrants on average accumulate less wealth than do comparable natives as captured by immigrant and native net wealth and their financial wealth as proportions of their permanent income. Our findings partially support those by Dustmann (1997), who argues that immigrants may save less than natives because of their different risk preferences or opportunities to diversify away economic risk. Our results are also consistent with those from Carroll, Rhee, and Rhee (1994), who find that recent immigrants to Canada save less than do Canadian-born.

Furthermore, immigrant and native wealth accumulation patterns are driven by different factors. In particular, we find that natives appear to carry out more precautionary savings than do comparable immigrants. The possibility exists, however, that immigrants' apparent lower precautionary savings really stems from the fact that immigrants, unlike natives, engage in precautionary saving by remitting sums to their home countries. Various motives for remitting sums to families in their home countries have been advanced in the literature, including altruism, portfolio diversification, insurance, to curry favor with family members, and in anticipation of future return migration (see Galor and Stark 1990; Faini 1994; Funkhouser 1995). In many cases, it may be that these remittances serve as wealth accumulation that can be drawn on in the future. We find it interesting that Osili (2000) reports that 51% of respondents interviewed for her survey of Nigerian immigrants in the Chicago area claim to have housing assets i n their Nigerian hometowns. Perhaps, our weaker results regarding immigrants' precautionary saving when a measure of net wealth (which includes housing) is used are partially due to the higher likelihood of mismeasurement of net wealth among immigrants given their housing assets abroad. For these reasons, future work on remittances constitutes an essential step for gaining a better understanding of immigrants' versus natives' precautionary savings and wealth accumulation patterns.

Nonetheless, insofar as many U.S. macroeconomic implications are concerned, the results of this paper remain relevant. Lower precautionary savings and lower wealth-to-income ratios suggest that immigrants have lower U.S. saving rates relative to native-born workers, limiting domestic saving available for investment.
Appendix A

Instrumental Variable Regressions

Group Natives
Dependent Variable Log (Income Uncertainty)
Statistics Coefficient
 (Robust Standard Error)

Age -0.0188 ***
Male -0.0882 ***
Education -0.1213 ***
Black 0.2572 ***
Other 0.2436 ***
Married -0.2215 ***
Separated 0.1204 ***
Divorced 0.0950 ***
Widowed -0.2557 **
Family size -0.0694 ***
Risk-loving 0.0111 **
Local unemployment rate -0.1568 ***
Time in the U.S. --

Number of observations 52,675
Number of groups 8004
Wald chi-square test Wald chi-square (12) =
 Prob > chi-square =

Group Immigrants
Dependent Variable Log (Income Uncertainty)
Statistics Coefficient
 (Robust Standard Error)

Age -0.0038
Male -0.1823 ***
Education -0.1266 ***
Black -0.0226
Other -0.0188
Married -0.1109 *
Separated 0.0279
Divorced 0.2076 *
Widowed 0.6248
Family size -0.0285 **
Risk-loving 0.0582 ***
Local unemployment rate -0.1321 ***
Time in the U.S. -0.0002 **
Number of observations 2648
Number of groups 409
Wald chi-square test Wald chi-square (13) =
 Prob > chi-square =

A constant term is included in the regression. Female, White, and Single
are used as reference categories.

* Significant at the 10% level.

** Significant at the 5% level.

*** Significant at the 1% level.

Appendix B: Data

Net family income is constructed using information on military income, farm, and nonfarm income (such as wages, salary, tips, and business income), unemployment compensation, child support, Aid for Families with Dependent Children payments, other welfare and supplemental security income, education benefits, disability and veterans benefits, interests, dividends, rents, rental subsidies, inheritances, gifts and any other income from family members. The variable calculates the net family income corresponding to the past calendar year. Net family income is deflated using the GNP deflator.

Net household worth is constructed taking assets minus liabilities. Within assets, we incorporate information on (i) monetary assets, such as savings and checking accounts, money market funds, credit unions, U.S. savings bonds or Treasury bills, individual retirement accounts (IRA or KEOGH), 401k or pretax annuities, certificates of deposit, personal loans to others, or any cash kept in a safe place at home or elsewhere; (ii) market value of main home; (iii) market value of farms, businesses, and other real estate, including a second home, land, and rental real estate; (iv) market value of vehicles owned, including cars, trucks, motor homes, trailers, and boats; and (v) other items worth over $500.

For liabilities, we used information on (i) home-related debts; (ii) other debts related to the permanent residence; (iii) farm/business/real estate-related debts; (iv) debts on owned vehicles; and (v) any amount of money over $500 owed to creditors, not including debts on 30-day charge cards. Net household worth is deflated using the GNP deflator.

Financial wealth equals the value of the aforementioned monetary assets deflated by the GNP deflator.

Risk loving: This variable is derived from the following survey questions: (i) For all respondents: "Suppose you are the only income earner in the family, and you have a good job guaranteed to give you your current (family) income every year for life. You are given the opportunity to take a new and equally good job with a 50-50 chance that it will double your (family) income and a 50-50 chance that it will cut your (family) income by a third. Would you take the new job?" (ii) For those who answered yes to the first question: "Suppose the chances were 50-50 that it would double your family income and 50-50 that it would cut it in half. Would you still take the new job?" (iii) For those who answered no to the first question: "Suppose the chances were 50-50 that it would double your (family) income and 50-50 that it would cut it by 20%. Would you take the new job?" If the respondent answers no to the first questions and no to the third question, a risk-loving variable is created that takes the value of zero (i.e ., risk averse). If the respondent answers no to the first question and yes to the third question, the risk-loving variable takes the value of 1 (i.e., low risk loving). If the respondent answers yes to the first question and no to the second question, the risk-loving variable takes the value of 2 (i.e., moderate risk loving). Finally, a person answering yes to the first question and yes to the second question has a risk-loving variable that equals 3 (i.e., high risk loving).

Sample Issues


Our sample refers to the period 1985-1990 and 1992-1993. No wealth data were collected in 1991 because of limited funding. Wealth information in the NLSY79 refers to the current year, whereas the income information corresponds to the past calendar year. Since the survey was not conducted in 1995, there are no income data available for 1994. Similarly, we lack household income information for 1996, since the survey was not fielded in 1997.

Additionally, to work with the maximum number of observations, particularly given the limited number of immigrants in the sample, we use an unbalanced panel. Nonetheless, as noted by Wooldridge (2000), by using random effects we can allow attrition to be correlated with the unobserved effect. That is, some units are more likely to drop out of the survey, but this is captured by the unobserved effect. Additionally, we use the Baltagi and Chang (1994) consistent variance-component estimators for unbalanced panels to correct for possible heteroscedasticity due to the varying size of the individual cluster.

Top Coding

To account for the wealth and income top-coding, we multiply top-coded values by 1.45. The latter assumes that the distribution of income and wealth above the censoring point have a Pareto tail. The value of 1.45 is determined from calculations performed in years with practically no censoring. Up to 1988, the values of income, assets, and debts exceeding particular limits were truncated and the upper limits converted to a set maximum value. However, beginning in 1989, the amounts exceeding the upper limits were assigned the average value of all values exceeding the limits in an effort to more accurately reflect the true range of income, assets, and debts. The new algorithm does not affect the estimates of mean and median holdings. With that same purpose, we use the adjustment method described above to more appropriately reflect the range of values for those top-coded incomes, assets, and debts from 1985 through 1988. In any event, because of the wider distribution of wealth relative to income, this procedure might exclude the very top values of wealth. Therefore, the results concerning the wealth distribution in Table 4 should be interpreted with caution.

In any event, it is important to emphasize that, in most instances, the percentage of values being top coded only exceeds the upper 2% of the distribution for the market value of the main residence in 1988--when 2.4% of main residence market values are top coded. Therefore, top coding results in a typical trimming of the upper end of the distribution as in previous studies (Lusardi 1998).

Non-Response and Imputation Rates

Although the NLSY79 has very high response rates on the ownership questions--generally exceeding 99%, the percentage of those individuals providing an amount is sometimes smaller. Instead of deleting cases with any missing data, we impute the missing data using the predictions from the regression of the variable with the missing value on the list of variables included in the income and wealth regressions for which the values are not missing (Little and Rubin 1990). The following percentages of observations were imputed for each of the following assets/liabilities in our final sample: (i) an average of 3% for monetary assets, such as savings and checking accounts, money market funds, credit unions, U.S. savings bonds or Treasury bills, individual retirement accounts (IRA or KEOGH), 401k or pretax annuities, certificates of deposit, personal loans to others, or any cash kept in a safe place at home or elsewhere; (ii) 0.69% for market value of main home; (iii) 0.75% for market value of farms, businesses, and oth er real estate, including a second home, land, and rental real estate; (iv) 3.26% for market value of vehicles owned, including cars, trucks, motor homes, trailers, and boats; (v) 1.57% for other items worth over $500; (vi) 1.02% for home-related debts; (vii) 0.63% for other debts related to the permanent residence; (viii) 0.67% for farm/business/real estate-related debts; (ix) 0.75% for debts on owned vehicles; and (x) 0.74% for any amount of money over $500 owed to creditors, not including debts on 30-day charge cards.
Table 1

Distribution of Natives and Immigrants by Risk Preference

Variable: Risk Attitude
 (= 0, 1, 2, 3,) Nativdes Immigrants

Risk Averse (=0) 45.72 46.30
Low risk loving (=1) 12.76 11.88
Moderate risk loving (=2) 17.69 15.14
High risk loving (=3) 23.82 26.68
Chi-square test for significantly Pearson chi-square (3) =
 different risk attitudes: -642032973.81
 Prob = 0.00
Table 2

Description and Measurement of Variables Used in Analysis

Variables Description and Measurement

Income Deflated household income
Age Age of respondent in years
Male Gender dummy
Education Number of years of education
White, black, other Race dummies
Not married, married, separated, Civil status dummies
Family size Number of people in the household
Professionals and technical Occupations dummies
 workers, managers, clerical
 workers, operatives and laborers,
 service workers, and home workers
Agriculture, mining, construction, Industry dummies
 manufacturing, transportation;
 communications--public utilities,
 trade, finance--insurance--real
 estate, services (business,
 personal, entertainment, and
 professional), and public
Tenure Tenure accumulated at their jobs in
 number of weeks
Tenure (2) Tenure squared
Unemployment experience Number of unemployed weeks over the
 total number of active weeks
Health condition Dummy variable equal to 1 when the
 person faces work limitations
 because of health conditions
Risk attitude See Appendix B: Data appendix
Time in the U.S. Time in the U.S. in weeks
Permanent income Predictions from the regression of
 deflated family income on the
 respondent's personal and work
(Wealth/permanent income) Ratio of wealth to permanent income
(Financial wealth/ Ratio of financial wealth to
permanent income) permanent income
Unemployment rate The unemployment rate is measured
 slightly differently from state
 to state. The finest level is the
 standard metropolitan statistical
 area (SMSA). Locations outside
 SMSAs are typically lumped
 together for the balance of the
 state. It never gets finer than
 county level.
Income uncertainty Average sum, up to the year in
 question, of the squared
 residuals of annual regressions
 of the logarithm of income on the
 respondent's personal and job-
 related characteristics, health
 condition, and attitude towards
 risk, normalized by permanent
Table 3

Variables' Means and Standard Deviations for Natives and Immigrants

Variables Mean Standard Deviation

Income 51,611.36 112,193.5
Age 27.465 3.512
Male 0.491 0.500
Education 12.770 2.300
Black 0.280 0.449
Other 0.043 0.202
Married 0.467 0.499
Separated 0.046 0.210
Divorced 0.083 0.276
Widowed 0.003 0.056
Family size 3.095 1.710
Professionals & technical 0.131 0.337
Manager 0.085 0.278
Clerical 0.166 0.372
Operatives and laborers 0.286 0.452
Home workers 0.010 0.100
Agriculture 0.025 0.157
Mining 0.007 0.085
Construction 0.074 0.262
Manufacturing 0.187 0.390
Public utilities 0.058 0.234
Finance, insurance, estate 0.060 0.237
Trade 0.206 0.404
Public administration 0.052 0.221
Tenure 121.723 153.369
Tenure (2) 38,338.23 84,470.69
Unemployment experience 0.106 0.602
Health condition 0.069 0.253
Risk attitude 1.210 1.262
Time in the U.S. -- --
Permanent income 41,145.26 16,754.78
Wealth/permanent income 1.022 8.365
Financial wealth/permanent income 0.430 6.889
Local unemployment rate 2.837 0.931
Income uncertainty 2.8e-04 8.9e-03

Variables Mean Standard Deviation

Income 52,098.99 103,736.8
Age 27.756 3.601
Male 0.514 0.500
Education 11.790 3.316
Black 0.091 0.288
Other 0.255 0.436
Married 0.544 0.498
Separated 0.052 0.222
Divorced 0.058 0.234
Widowed 0.006 0.080
Family size 3.799 1.933
Professionals & technical 0.111 0.314
Manager 0.083 0.276
Clerical 0.170 0.375
Operatives and laborers 0.317 0.465
Home workers 0.006 0.080
Agriculture 0.052 0.221
Mining 0.004 0.062
Construction 0.058 0.233
Manufacturing 0.209 0.407
Public utilities 0.061 0.240
Finance, insurance, estate 0.081 0.274
Trade 0.207 0.405
Public administration 0.033 0.178
Tenure 135.142 162.978
Tenure (2) 44,819.96 91,693.22
Unemployment experience 0.083 0.358
Health condition 0.050 0.217
Risk attitude 1.194 1.295
Time in the U.S. 1384.986 269.458
Permanent income 42,002.31 15,353.2
Wealth/permanent income 1.082 6.419
Financial wealth/permanent income 0.271 1.857
Local unemployment rate 2.872 0.992
Income uncertainty 9.0e-05 1.4e-03
Table 4

Mean Wealth, Median Wealth, and Distribution of Wealth by Gross
Household Income Quartiles (in 1997 Dollars)

Group Natives Immigrants
Gross Household Mean Median Distribution Mean
Income Quartiles Wealth Wealth of Wealth Wealth

Year 1985
Lowest 9568.14 1941.75 10% 9412.54
Second 17,101.02 5836.89 23% 16,962.13
Third 29,635.54 11,650.49 27% 42,438.54
Highest 50,102.11 15,533.98 41% 46,291.23

Year 1993
Lowest 20,515.73 2030.34 7% 71,371.75
Second 53,067.70 15,869.31 14% 84,983.34
Third 181,306.80 38,389.73 23% 204,247.70
Highest 235,476.50 91,015.17 56% 305,659.50

Group Immigrants
Gross Household Median Distribution
Income Quartiles Wealth of Wealth

Year 1985
Lowest 1841.25 11%
Second 4077.67 23%
Third 14,951.46 39%
Highest 10,737.86 30%

Year 1993
Lowest 4200.7 9%
Second 23,337.22 21%
Third 34,472.15 21%
Highest 117,977.80 52%
Table 5

Ownership Rates and Median Value of Asset and Debt Holdings (in 1997

Assets/Debts Percentage that Own Assets/Debt Type
Group Natives Immigrants
Year 1985 1993 1985

Home 19.47 51.21 13.55
Farm, business, real estate 6.71 11.41 5.73
Financial assets 69.85 77.00 65.00
Vehicles 77.33 87.16 68.16
Other assets 57.15 66.42 52.32
Property debts 19.77 48.04 14.77
Car debts 39.19 42.70 28.66
Other debts 32.26 39.71 28.40

Assets/Debts Percentage Median Values of
 that Own Holdings for
 Assets/Deb Assets/Debt Holders
 t Type
Group Immigrants Natives
Year 1993 1985 1993

Home 50.02 87,378.64 93,348.90
Farm, business, real estate 12.87 38,834.95 43,173.86
Financial assets 73.25 1941.75 7491.25
Vehicles 85.62 7766.99 8955.19
Other assets 67.40 3883.50 5834.31
Property debts 45.93 58,252.43 65,344.23
Car debts 29.92 6213.59 7001.17
Other debts 31.14 3833.50 3733.96

Assets/Debts Median Values of Holdings for
 Assets/Debt Holders
Group Immigrants
Year 1985 1993

Home 116,504.90 140,023.30
Farm, business, real estate 71,916.51 49,203.04
Financial assets 1941.75 8160.00
Vehicles 7766.99 8168.03
Other assets 4854.37 7001.17
Property debts 79,611.65 77,012.84
Car debts 7378.64 5834.31
Other debts 3883.50 4084.01
Table 6

Descriptive Statistics of Wealth by Age, Family Size, and Civil Status
(in 1997 Dollars)

Group Natives
Year 1985 1993
Statistic Mean Median Mean Median

Total 27,102.75 6824.02 100,971.10 25,320.89
Married 45,829.71 19,392.37 137,036.30 46,441.07
Unmarried 15,612.95 3883.50 45,343.89 4900.82
Family Size <4 30,748.50 8737.86 88,284.11 21,003.50
Family Size [greater 19,773.99 4077.67 118,599.30 34,364.06
 than or equal to]4

Group Immigrants
Year 1985 1993
Statistic Mean Median Mean Median

Total 28,667.05 5825.24 167,143.00 29,171.53
Married 52,877.50 17,184.47 213,670.90 49,008.17
Unmarried 12,421.94 2621.36 71,773.78 5250.88
Family Size <4 24,290.23 8566.99 135,451.60 21,586.93
Family Size [greater 33,029.21 3883.50 193,483.70 40,268.27
 than or equal to]4
Table 7

Random-Effects Regression of the Log of Deflated Family Income

Group Natives

Variables Coefficient

Male 0.0594 ***
Education 0.1110 ***
Black -0.2747 ***
Other -0.2199 ***
Married 0.2535 ***
Separated -0.1205 ***
Divorced -0.1144 ***
Widowed 0.0708
Family size 0.0586 ***
Tenure 0.0012 ***
Tenure (2) -1.35E-06 ***
Unemployment experience -0.0069
Health condition -0.1614 ***
Risk loving -0.0099 ***
Time in U.S. --
Number of observations 54,214
Number of groups 8031
Wald Chi-square test Wald chi-square (27)
 = 8426.50 Prob >
 chi-square = 0.0000

Group Natives Immigrants
Variables Standard Error Coefficient

Male 0.0117 -0.0069
Education 0.0025 0.0969 ***
Black 0.0123 -0.0357
Other 0.0271 -0.0277
Married 0.0085 0.1326 ***
Separated 0.0166 -0.2060 ***
Divorced 0.0142 -0.1217 *
Widowed 0.0692 -0.9933 ***
Family size 0.0021 0.0465 ***
Tenure 514E-05 0.0011 ***
Tenure (2) 8.69E-08 -1.37E-06 ***
Unemployment experience 0.0044 -0.0982
Health condition 0.0138 -0.2970 **
Risk loving 0.0044 -0.0479 ***
Time in U.S. -- 0.0002 *
Number of observations
Number of groups
Wald Chi-square test

Group Immigrants
Variables Standard Error

Male 0.0486
Education 0.0082
Black 0.0866
Other 0.0535
Married 0.0358
Separated 0.0689
Divorced 0.0671
Widowed 0.2728
Family size 0.0075
Tenure 0.0002
Tenure (2) 3.64E-07
Unemployment experience 0.0403
Health condition 0.0624
Risk loving 0.0179
Time in U.S. 9.26E-05
Number of observations 2740
Number of groups 410
Wald Chi-square test Wald Chi-square (28)
 = 407.51 Prob > Chi
 -square = 0.0000

A constant term, dummy variables for industry and occupation are
included in the regression. Female, White, and Single are used as
reference categories.

* Significant at the 10% level.

** Significant at the 5% level.

*** Significant at the 1% level.
Table 8

Two-Stage Least-Square Random Effects Regressions of the Ratio of Wealth
to Permanent Income

Group Natives
Dependent Net Wealth Financial Wealth
Variable Coefficient Coefficient
Statistics (Robust Standard Error) (Robust Standard Error)

Age 0.1182 *** 0.0529 ***
 (0.0121) (0.0097)
Male 0.0427 0.0595
 (0.0833) (0.0656)
Education 0.0975 *** 0.1240 ***
 (0.0365) (0.0289)
Black -0.6336 *** -0.4279 ***
 (0.1194) (0.0946)
Other -0.0967 -0.2191
 (0.2057) (0.1623)
Married 0.5062 *** 0.1216
 (0.1084) (0.0859)
Separated -0.5144 ** -0.3796 **
 (0.1985) (0.1618)
Divorced -0.3945 ** -0.3358 ***
 (0.1546) (0.1246)
Widowed 1.1434 0.7302
 (0.7813) (0.6356)
Family size -0.0549 * -0.0174
 (0.0316) (0.0252)
Risk-loving 0.0263 0.0206
 (0.0317) (0.0250)
Log (income 0.5617 ** 0.8443 ***
uncertainty) (0.2658) (0.2085)
Time in the U.S. -- --

Number of observations 52,675 52,675
Number of groups 8004 8004

Wald Chi-square test Wald chi-square (12) = Wald chi-square (12) =
 234.06 77.80
 Prob > Chi-square = Prob > chi-square =
 0.0000 0.0000

Group Immigrants
Dependent Net Wealth Financial Wealth
Variable Coefficient Coefficient
Statistics (Robust Standard Error) (Robust Standard Error)

Age 0.1612 *** 0.0311 ***
 (0.0267) (0.0123)
Male 0.0585 0.3 142 ***
 (0.2137) (0.0986)
Education 0.0137 0.1353 ***
 (0.0894) (0.0412)
Black -0.6685 ** -0.2425
 (0.3321) (0.1532)
Other -0.2602 -0.1260
 (0.2090) (0.0964)
Married 0.4917 ** 0.3385 ***
 (0.2241) (0.1034)
Separated -0.4198 -0.0319
 (0.4278) (0.1973)
Divorced 0.1148 0.3880 **
 (0.4117) (0.1899)
Widowed 0.0459 0.4747
 (1.4477) (0.6677)
Family size 0.0215 0.0288
 (0.0529) (0.0244)
Risk-loving -0.0267 0.0321
 (0.0806) (0.0372)
Log (income 0.2134 0.7521 **
uncertainty) (0.6648) (0.3066)
Time in the U.S. 3.57e-05 4.60e-04 ***
 (3.80e-04) (1.75e-04)
Number of observations 2648 2648
Number of groups 409 409

Wald Chi-square test Wald chi-square (13) = Wald chi-square (13) =
 66.62 59.60
 Prob > chi-square = Prob > chi-square =
 0.0000 0.0000

A constant term is included in the regression. Female, White, and Single
are used as reference categories.

* Significant at the 10% level.

** Significant at the 5% level.

*** Significant at the 1% level.

Received June 2000; accepted October 2001.

(1.) There is a voluminous literature on savings. Much of the literature on saving patterns has primarily relied on wealth accumulation measures as a proxy for savings given the disagreement on what constitutes savings as well as the lack of reliable microlevel data on savings. Therefore, most analyses (including this one) resort to measures of wealth accumulation that combine respondents' income, assets, and liabilities. In any event, some papers examine savings using U.S. aggregate data; others analyze U.S. saving patterns using microlevel data, and a third group of papers addresses savings in foreign countries. This study varies from the studies above in that it distinguishes immigrants' from natives' saving patterns using individual-level data from the NSLY79. Studies using longitudinal data tracking the evolution of savings through the individual's life cycle have been relatively recent (Jianakoplos, Irvine, and Menchik 1986; Carroll and Samwick 1998; and Lusardi 1998, among others). A common trend acros s the mentioned studies is their primary focus on precautionary savings using different theoretical frameworks, such as the life-cycle permanent-income model with and without liquidity constraints or the buffer stock model (Blanchard and Mankiw 1988; Skinner 1988; Kimball and Mankiw 1989; Zeldes 1989; Caballero 1990).

(2.) For instance, Djajic and Milbourne (1988) construct an intertemporal utility maximization problem and model the links among the migration decision, rate of saving abroad, and length of stay in the host country. Berninghaus and SeifertVogt (1993) posit that immigrants who do not reach their target savings levels become permanent migrants. Finally, using a West German data set, Merkle and Zimmermann (1992) relate saving and remittances to demographic variables and to the anticipated length of stay in Germany.

(3.) Return and repeat migration may occur for a variety of reasons. For instance, illegal immigrants may be forced to exit the country if apprehended, economic downturns may place pressures on immigrants to return to their homelands or migrate to a third country, and psychological/cultural factors may prompt immigrants to return home.

(4.) Lifetime expected utility is given by:

U([c.sup.I], [c.sup.E], t) = [tu.sup.I]([c.sup.I]) + (I - t][u.sup.E]([c.sup.E])

where [u.sup.I]([c.sup.I]) and [u.sup.E]([c.sup.E]) represent utility from consumption in the host and home countries, respectively. The subutility functions are weighted by t and (I - t) the proportion of the lifetime the individual chooses to spend abroad and at home.

The budget constraint facing the immigrant is:

[tpc.sup.I] + (I - t)[c.sup.E] + [eta] = [y.sup.I](t, x) + [y.sup.E](t, z)

where [eta] represents the migration cost and p is the price level in the host country relative to the home country. Income earnings in the host country, [y.sup.I](t, x), and in the home country, [y.sup.E](t, z), are related to the length of time working in each of the locations and to the "state of the labor markets" abroad (x) and at home (z). Labor market conditions are uncertain, making future income uncertain. This uncertainty is represented by the variances of the state of the labor market variables, Var([y.sup.I]) = [[sigma].sup.2.sub.x] and Var([y.sup.E]) = [[sigma].sup.2.sub.z] and the covariance between them.

(5.) In particular, if any of the following apply: (i) The index of the state of the labor market at home is lower than abroad (z < x); (ii) migrants prefer home country consumption [[u.sup.I](k) > [u.sup.E](k)]; or (iii) prices in the host country are greater than in the home country (p > I). In addition, economic agents will engage in precautionary saving as long as the utility function allows for decreasing absolute risk aversion. That is, both immigrants and natives will be influenced by the variance of lifetime income.

(6.) That is, Var([y.sup.I]) is higher for an immigrant than for a native. It is reasonable to assume that immigrant workers experience greater income uncertainty. For example, immigrants are less likely to be covered by social insurance programs. Hence, unanticipated negative shocks to income will not be cushioned as they would for individuals who are covered by social insurance programs. Also, immigrants tend to have less information and less human capital specific to host country labor markets, making it more difficult to maintain steady income streams.

(7.) A detailed description of how net family income is constructed and used in this analysis is contained in Appendix B, the data Appendix.

(8.) See Engelhardt (1998) for a discussion of the advantages and weaknesses of using other data sets containing detailed information on wealth and income, such as the Surveys of Consumer Finances, the Consumer Expenditure Survey, the Panel Study of Income Dynamics, the HRS, the Asset and Health Dynamics Survey, the Longitudinal Retirement History Surveys, and the Survey of Income and Program Participation.

(9.) Detailed information on which assets and liabilities are included in our wealth measure is contained in the data Appendix.

(10.) Refer to Appendix B for greater detail on this transformation.

(11.) No wealth data were collected in 1991 because of limited funding. Wealth information in the NLSY79 refers to the current year, whereas the income information corresponds to the past calendar year. Since the survey was not conducted in 1995, there are no income data available for 1994. Similarly, we lack household income information for 1996, since the survey was not fielded in 1997.

(12.) Refer to Appendix B for more detailed information on our sample.

(13.) See Appendix B for greater details on the risk-loving variable. Using data from the HRS, Barsky et al. (1997) show that the risk variable may, according to who is being sampled, reflect other respondents' preferences aside from their risk attitudes. Respondents, for example, may value their jobs for reasons different from the income flow associated with them, resulting in a "status quo bias" that overstates risk aversion. Similarly, respondents may be willing to switch their jobs for reasons different from income, such as a better fit or more flexible schedules being offered by the new job. In that case, risk tolerance would be overestimated.

(14.) A variety of income uncertainty measures have been previously used in the literature. Skinner (1988) used the household head's occupation. The use of occupation to proxy income variance has been widely criticized because of the self-selection of individuals into more or less risky occupations; therefore occupations do not provide an exogenous source of risk. Guiso, Jappelli, and Terlizzese (1992) relied on the subjective probability distribution of future income. The measure used by Guiso, Jappelli, and Terlizzese (1992) has also been referred to as problematic provided the difficult understanding of the questions being asked (Lusardi 1998). Lusardi (1998) uses a measure of income variance derived from data on subjective probabilities of job loss normalized by permanent income. Finally, Carroll and Samwick (1998) opt for the variance of the logarithm of income. Although using subjective probabilities of job loss as Lusardi (1998) does would be appealing, particularly in our case, since we are dealing wi th a sample not close to retirement, the NLSY79 does not question respondents on their subjective probabilities of future job loss or income. On the other hand, although suitable for their paper because of their use of one-year of data on wealth, Carrol and Samwick's (1998) variance of the logarithm of income results in a constant, summary measure of the variability of income, whether predicted (explained) or unpredicted (unexplained), for the entire period under examination. Therefore, we opt to use a measure that captures the average unexplained income (as captured by the residuals from annual log income regressions) relative to explained or predicted income over time. The greater the ratio, the greater is income uncertainty.

(15.) Permanent income is proxied by the predictions of the income regressions.

(16.) See Lusardi (1998) and Carroll and Samwick (1998).

(17.) The instrumenting procedure used in the estimation of the structural equation of the wealth-to-income ratio is explained in the next section.

(18.) See Chiswick (1978) and Borjas (1985).

(19.) A cohort effect could be at work here.

(20.) We use the local unemployment rate since, when included in the wealth-to-permanent income regressions, it always turned Out to be statistically insignificant in determining wealth and, yet, it is negatively correlated with income uncertainty, as shown by the instrumental regressions in Appendix A.

(21.) This result may appear, at first, counterintuitive. However, evidence exists that argues that real wages are procyclical (e.g., Bils 1985; Solon, Whatley, and Stevens 1997), growing by less and varying by less during periods of high unemployment. This is true even for workers who stay with the same employer (Solon, Barsky, and Parker 1994; Shin 1994). Hence, higher area unemployment rates result in less unexplained income.

(22.) Income uncertainty enters in logarithmic form. After estimating a variety of models, the actual level--log model was preferred since it more closely captured the relation between precautionary savings and income uncertainty than, for instance, the level--level model. The actual model allows for nonlinearities between precautionary saving and income uncertainty, In particular, it allows for the income uncertainty effect to diminish as income uncertainty increases.

(23.) Note that like tenure, occupation, and industry, the variables unemployment experience and health condition affect wealth accumulation indirectly through their effect on the respondent's earned job income. They do not enter directly as dependent variables in the wealth regressions since income risk is associated with those employment characteristics that, per se, could capture some precautionary motive. Therefore, to be able to infer the most precautionary saving from our income uncertainty measure, we exclude these variables from the wealth regressions. Risk preferences, however, are likely to affect job income through the individual's preference for certain jobs as well as wealth accumulation through the respondent's own saving or investment preferences. Therefore, they are included in the income and wealth regressions along with other personal characteristics likely to affect both job choices and wealth accumulation. In this manner, we are able to gauge the effect of income uncertainty on precautiona ry savings while controlling for the individual risk preferences. We also considered interacting risk preferences with income uncertainty to further capture the varying effect of income uncertainty by risk preferences. Nonetheless, the interaction term was never significantly different from zero; therefore, we excluded it from our final specification.

(24.) Instead, immigrants may be responding to income uncertainty through the remittance of income to their home countries to aid their family members, These payments could be considered an insurance payment for when they possibly retum home; in which case, immigrants' wealth accumulation does respond to income uncertainty, but its responsiveness is captured differently from that for natives. Unfortunately, the NLSY does not contain information regarding respondents' money remittances to allow us to capture this possibility regarding the responsiveness of immigrants' wealth accumulation to income uncertainty.

(25.) To obtain the impact of a 10% increase in income uncertainty, we computed the following: [DELTA]([W.sub.h]/[Y.sub.h.sup.p]) = (coefficient/l00)(%[DELTA]income uncertainty) = (0.56/100)(l0) = 0.056. Since average [W.sub.h]/[Y.sub.h.sup.P] for natives is 1.022, a 10% increase in income uncertainty raises natives' average wealth-to-permanent income ratio to 1.078 or approximately 5.5%.

(26.) Lusardi (1998), using a subjective measure of income uncertainty--equal to the variance of income constructed using respondents' subjective probability of losing their jobs and later on normalized by dividing by their permanent income-- and data from the HRS survey, finds that the contribution of precautionary savings to wealth accumulation ranges between I and 3.5%. Our result is slightly larger, which might be partially due to the younger age of our respondents, This is consistent with the work of Irvine and Wang (2001), who developed a model in which income uncertainty affects saving differently over the life cycle, with stronger impacts when individuals are young.

(27.) Immigrants' average net wealth- and financial wealth-to-permanent income ratios if they had natives' characteristics are predicted using the estimated coefficients from the wealth-to-permanent income regressions for both groups.


Baltagi, Badi H., and Young-Jae Chang. 1994. Incomplete panels: A comparative study of alternative estimators for the unbalanced one-way error component regression model. Journal of Econometrics 62:67-89.

Barsky, Robert B, Thomas F. Juster, Miles S. Kimball, and Matthew D. Shapiro. 1997. Preference parameters and behavioral heterogeneity: An experimental approach in the health and retirement study. Quarterly Journal of Economics 112:537-79.

Berninghaus, Siegfried, and Hans Gunther Seifert-Vogt. 1993. The role of target saving motive in guest worker migration. Journal of Economic Dynamics and control 17:181-205.

Bils, Mark J. 1985. Real wages over the business cycle: Evidence from panel data. Journal of Political Economy 93:666-89.

Blanchard, Oliver J., and N. Gregory Mankiw. 1988. Consumption: Beyond certainty equivalence. NBER Working Paper No. 2496.

Borjas, George J. 1985. Assimilation, changes in cohort quality, and the earnings of immigrants. Journal of Labor Economics 3:463-89.

Browning, Martin, and Annamaria Lusardi. 1996. Household saving: Micro theories and micro facts. Journal of Economic Literature 34:1797-1855.

Buchinsky, Moshe. 1994. Changes in the U.S. wage structure 1963-1987: Application of quantile regression. Econometrica 62:405-58.

Caballero, Ricardo J. 1990. Consumption puzzles and precautionary savings. Journal of Monetary Economics 25:113-36.

Carroll, Christopher D. 1992. The buffer-stock theory of savings: Some macroeconomic evidence. Brookings Papers on Economics Activity 2:61-135.

Carroll, Christopher D. 1997. Buffer stock saving and the life cycle/permanent income hypothesis. Quarterly Journal of Economics 112:1-56.

Carroll, Christopher D., Byung-kun Rhee, and Changyong Rhee. 1994. Are there cultural effects on saving? Some cross-sectional evidence. Quarterly Journal of Economics 109:685-99.

Carroll, Christopher D., and Andrew A. Samwick. 1998. How important is precautionary saving? The Review of Economics and Statistics 80:410-9.

Center for Human Resource Research. 1997. NLSY79 users' guide. Columbus, OH: The Ohio State University.

Center for Human Resource Research. 1997. NLSY79 update, volume 1, issue 1. Columbus, OH: The Ohio State University.

Chiswick, Barry R. 1978. The effect of Americanization on the earnings of foreign-born men. Journal of Political Economy 86:897-921.

Deaton, Angus. 1991. Saving and liquidity constraints. Econometrica 59:1121-42.

Djajic, Slobodan, and Ross Milbourne. 1988. A general equilibrium model of guest-worker migration: A source country perspective. Journal of International Economics 25:335-51.

Dustmann, Christian. 1997. Return migration, uncertainty and precautionary saving. Journal of Development Economics 52:295-316.

Engelhardt, Gary V. 1998. Income and wealth in the NLSY79. Unpublished paper, Dartmouth College.

Faini, Riccardo. 1994. Workers remittances and the real exchange rate: A quantitative framework. Journal of Population Economics 7:235-45.

Funkhouser, Edward. 1995. Remittances from international migration: A comparison of El Salvador and Nicaragua. Review of Economics and Statistics 77:137-46.

Galor, Oded, and Oded Stark. 1990. Migrants' savings, the probability of return migration and migrants' performance. International Economic Review 31:463-7.

Guiso, Luigi, Tullio Jappelli, and Daniele Terlizzese. 1992. Earnings uncertainty and precautionary saving. Journal of Monetary Economics 30:307-37.

Hurst, Erik, Ming Ching Luoh, and Frank P. Stafford. 1998. The wealth dynamics of American families, 1984-94. Brookings Papers on Economics Activity 1:267-332.

Irvine, Ian, and Susheng Wang. 2001. Saving behavior and wealth accumulation in a pure lifecycle model with income uncertainty. European Economic Review 45:233-58.

Jianakoplos, Nancy, F. Owen Irvine, and Paul Menchik. 1986. Household precautionary savings as an alternative to government income-maintenance programs. Unpublished paper, Michigan State University.

Juhn, C., Kevin Murphy, and Brooks Pierce. 1993. Wage inequality and the rise in returns to skill. Journal of Political Economy 101:410-42.

Kazarosian, Mark. 1997. Precautionary savings--A panel study. Review of Economics and Statistics 79:241-7.

Katz, L., and Ana Revenga. 1989. Changes in the structure of wages: The U.S. versus Japan. Journal of Japanese and International Economics 3:522-53.

Kimball, Miles S., and Gregory Mankiw. 1989. Precautionary saving and the timing of taxes. Journal of Political Economy 97:863-79.

Little, Roderick J. A., and Donald B. Rubin. 1990. Statistical analysis with missing data. New York: John Wiley & Sons.

Lusardi, Annamaria. 1998. On the importance of the precautionary saving motive. AEA Papers and Proceedings 88:449-54.

MaCurdy, Thomas, Thomas Nechyba, and Jay Bhattacharya. 1998. An economic framework for assessing the fiscal impacts of immigration. In The immigration debate: Studies on the economic, demographic, and fiscal effects of immigration, edited by James P. Smith and Barry Edmonstron. Washington, D.C.: National Academy Press, pp. 13-65.

Martin, John. 1995. Immigration contributes over half of U.S. population growth. Immigration Review 23:10-2.

Merkle, Lucie, and Klaus Zimmermann. 1992. Savings, remittances, and return migration. Economics Letters 38:77-81.

Murphy, Kevin, and Finis Welch. 1992. The structure of wages. The Quarterly Journal of Economics 107:285-326.

Osili, Una Okonkwo. 2000. Migrants and housing investments: Theory and evidence from Nigeria. Unpublished paper, Indiana University-Purdue University at Indianapolis.

Shin, Donggyun. 1994. Cyclicality of real wages among young men. Economics Letters 46:137-42.

Skinner, Jonathan. 1988. Risky income, life cycle consumption, and precautionary savings. Journal of Monetary Economics 22:237-55.

Solon, Gary, Robert Barsky, and Jonathan A. Parker. 1994. Measuring the cyclicality of real wages: How important is composition bias? Quarterly Journal of Economics 109:1-25.

Solon, Gary, Warren Whatley, and Ann Huff Stevens. 1997. Wages changes and intrafirm job mobility over the business cycle: Two case studies. Industrial and Labor Relations Review 50:402-15.

Wooldridge, Jeffrey. M. 2000. Introductory econometrics: A modern approach. Cincinnati, OH: South-Western Thomson Learning.

Zeldes, Stephen P. 1939. Optimal consumption with stochastic income: Deviations from certainty equivalence. Quarterly Journal of Economics 104:275-98.

Catalina Amuedo-Dorantes * and Susan Pozo +

* Department of Economics, San Diego State University, San Diego, CA 92182, USA; E-mail camuedod @; corresponding author.

+ Department of Economics, Westem Michigan University, Kalamazoo, MI 49008, USA; E-mail

We are indebted to Eskander Alvi, Randall Olsen, Una Okonkwo Osili, the editor, two anonymous referees, and participants at the Midwest Economics Association meetings for helpful comments and suggestions.
COPYRIGHT 2002 Southern Economic Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2002, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Pozo, Susan
Publication:Southern Economic Journal
Geographic Code:1USA
Date:Jul 1, 2002
Previous Article:The logit equilibrium: a perspective on intuitive behavioral anomalies.
Next Article:Purchasing power parity under the gold standard.

Related Articles
Report details immigration's impact on real estate.
Muddled Masses.
Estimating the Effects of Earnings Uncertainty on Families' Saving and Insurance Decisions.
Suicide less likely among immigrants, study reports.
The flip-flop crowd hits the corporate suites.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters