UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-CHAMPAIGN BOOKSTACKS H or -..v I V H V^^ P ^ ^ m H H m 1 1~* t /■.- Digitized by the Internet Archive in 2011 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/useoftransfersby850blau S'^K FACULTY WORKING PAPER NO. 850 The Use of Transfers by trnmigrants Francine D. Slau O (\l CO cr. ^^5 « -r o • 'C - -^ ~.' .5 ^ .'^ Cfc College of ComrrercG snd Business Adrninislrfction Bureau of Economic and Business Resea-'cn University of !!>ino!S. Urbana-ChaT-p-iign BEBR FACULTY WORKING PAPER NO. 850 College of Conmerce and Business Administration University of Illinois at Urbana-Champaign March 1982 Draft - For Comment only The Use of Transfers by Immigrants Francine D. Blau, Associate Professor Department of Economics Institute of Labor and Industrial Relations April 1981 Revised, February 1982 Acknowledgment: I am grateful for the helpful comments and suggestions of Julian Simon, Glen Cain, Barry Chiswick, Marianne Ferber, Daniel Hamennesh, Marjorie Honig, Lawrence Kahn, Cordelia Reimers, Jean Grossman and the members of the Princeton and University of Illinois Labor Economics Workshops . I am indebted to Jeremy Atack for preparing the data tapes for analysis and to Judi Catlett, Denise Dimon, Susan Schwochau, and Andy Jaske for research assistance. Abstract This paper uses data from the 1976 Survey of Income and Education to examine the receipt of transfers by immigrants in conqiarison to the native bom. The average level of transfers is found to be considerably higher among families headed by male immigrants. However, this is almost entirely due to the higher average age of family members among the immigrant group — a reflection of the large inflows of immigrants into the U.S. during the pre-World War I period. Holding other factors (including age) constant, immigrant families are found to be considerably less likely to rely on welfare than native families, while their receipts from social insurance programs are found to be only slightly higher. I. Introduction The recent large inflows of refugees from Southeast Asia, Cuba, and Haiti, and of illegal immigrants from Mexico, Latin America, and elsewhere have focused public attention on the consequences of these population movements for the United States economy. When considering the advisability of admitting immigrants to the U.S., their use of the transfer system is necessarily a major consideration. While this issue is one that excites much public interest, relatively little previous work has been done on the topic. Some of this concern over the economic impact of immigrants may stem from beliefs that recent arrivals differ in important respects from earlier immigrants. However, given the time dependent nature of the "Americanization" process — the process by which immigrants adapt to their surroundings — and the importance of age-related factors in determining transfer receipts, it is helpful to assess the experiences of the whole immigrant population, including earlier arrivals, in forming judgments as to the likely impact of newcomers. This is the strategy pursued in this paper. In the conclusion we specifically consider the implications of our findings for recent immigrants. Immigrant-native differences in receipts from major transfer pro- grams during 1975 are shown in Table 1. At a superficial level, these figures lend some support to fears that immigrants constitute a burden on the transfer payment system. On average male immigrants and their families (immigrant families) received more of both welfare (e.g., pub- lic assistance, AFDC, supplemental security income) and social insurance (e.g., social security, unemployment insurance, workman's compensation) -2- 2 pajnnents than native-bom males and their families (native families) . This amounted to a total differential of 53 percent or $571 per family per year. This differential reflects both immigrant families' greater likelihood of participating in each type of program (i.e., welfare or social insurance) and their higher average payment levels conditional on participation. The purpose of this paper is to develop a fuller understanding of the reasons for these immigrant-native differences in the utilization of transfers. Such an investigation may be useful in formulating immi- gration policy. For example, if policy makers wish to minimize the size of outlays on transfers, such information may be relevant to the selec- tion of criteria for admitting immigrants and/or for determining the level of immigration permitted. On the other hand, a deeper investiga- tion of the causes of these observed differentials may suggest that im- migrants do not, in fact, unduly burden the transfer payment system (in conqjarison to the native bom) and that a reduction of transfer pajnnent outlays on their account need not be a major policy concern. Economic theory suggests that "...migration in response to economic incentives is generally more profitable for the more able and more 3 highly motivated" (Chiswick, 1978 p. 900). If such traits are distrib- uted similarly across countries, immigrants may be more able or highly motivated than native-bom individuals with similar characteristics. This implication is weakened to the extent that migration is politically motivated or is induced by the availability of more generous welfare benefits in the place of destination (Chiswick 1978). Research on the earnings of immigrants tends to support the self-selection hypothesis. -3- Using 1970 Census data, Chiswick (1978) finds that while male immigrants initially earn less than natives with similar characteristics, they tend to catch up to and then surpass the native-born in earnings in a 10 to 15 year period. Similar findings are obtained by Blau (1980) for the early twentieth century, a period when welfare availability would not have been a consideration in the immigration decision. These findings suggest interesting questions regarding the utili- zation of transfers by immigrants. Does the increasing economic success of immigrants over time imply a decreasing reliance on transfers? As a more highly motivated group, are immigrants more reluctant to rely on transfer payments, all else equal, particularly welfare payments which have a negative social connotation? It is hoped that this paper may shed further light on the notion of the selectivity of immigration and its implications for social policy by examining the utilization of transfers by immigrants. To elucidate these issues, we seek to deteirmine whether immigrant families place greater or lesser demands on the transfer system than na- tive families with similar characteristics (i.e., whether immigrant fam- ilies are more or less "transfer prone" than native families). We also investigate the role of differences in characteristics across immigrant and native families in producing a higher utilization of the transfer system on the part of the former group. The policy implications of the specific characteristics identified as of primary importance in produc- ing the differential may then be evaluated. Of particular interest is the role played by age and age-related variables. The age distribution of the native population is determined primarily 4 by domestic birth and death rates. However, in the case of immigrants. -4- the major factor is the historical pattern of in-migration. Such in- flows peaked during the late nineteenth and early twentieth centuries (Blau, 1980). After World War I, restrictive legislation was adopted that sharply curtailed the entry of iiranigrants» As a result, the average age of immigrant male family heads, 51, is considerably higher than that of native male heads, 45; and 30 percent of male immigrant heads are 65 years of age or over, in comparison to 14 percent of the male native heads (Table 2). Viewing the matter somewhat differently, on average, there is one additional family member (excluding the male head) 65 years of age or over for every five immigrant families in comparison to one such older individual for every ten native families (Table 2). While immigrants have other characteristics that could raise their reliance on transfers, ceteris paribus (e.g., a lower educational at- tainment, a higher representation of minority groups), the role of age- related factors has a particular policy relevance. To the extent that the differences in the receipt of transfer payments shown in Table 1 are due to the impact of age and age-related variables, a fairly good case may be made for the view that the higher level of transfers among immi- grants does not represent a cause for concern from a policy point of view. First, as with any investment in human capital, immigration is more profitable the earlier in the life cycle it takes place (i.e., the longer the remaining work life period in which the returns may be col- lected) . Thus, it is likely that the vast majority of older immigrants have spent most of their working lives, including their most productive years, in the United States. Second, to the extent that the collection -5- of transfer pajments by older Individuals may be conceptualized as an intergenerational transfer, older immigrants have their own working age children who make positive contributions to this system in the form of income and social security taxes, etc. Third, in the decision of which and how many immigrants to admit, higher costs many years in the future in the form of transfer payments to older individuals would have rela- tively little weight in a present value calculation, at most discount rates (Simon, 1980). Further, the age distribution of immigrants can be manipulated by public policy in a beneficial manner, e.g., to even out population imbalances in age composition due to past fluctuations in domestic birth rates (Wachter, 1980). II. Conceptual Framework A. The Determinants of Transfers Given that virtually all transfer programs have work disincentives associated with them, the labor-leisure trade-off is at the heart of the transfer decision. Individuals may be viewed as comparing the value of market work with the benefits of participation in the transfer program. Thus, the demand for transfers may be viewed as being determined by the following factors. All else equal, the higher the individual's market wage, the lower his demand for transfers is likely to be. Further, dif- ficulty obtaining employment at the market wage corresponding to one's characteristics, is expected to increase the demand for transfers, other things equal. Factors which increase the demand for leisure at a given price (e.g., higher nonlabor income or assets, certain demographic char- acteristics) are expected to raise the demand for transfers, ceteris paribus . -6- The supply of transfers is determined by the eligibility require- ments of the program; the program parameters (i.e., the guarantee level- benefit level at zero work hours; the variables that determine the guar- antee level; and the marginal tax rate on labor income); and the stringency with which the program is administered. Given the individ- ual's demand for and the potential supply of transfers, individual out- comes may differ depending on the skillfulness of the individual in understanding and navigating his way through the various bureaucratic requirements (i.e., his "program skills"). As we have seen in Table 1, differences between two groups (e.g., immigrants and natives) in the average level of transfer receipts may arise from group differences in the probability of participating in a transfer program [P(T > 0)] or the level of payments received, given participation [T|(T > 0)], or a combination of both. As part of our ef- fort to explore a variety of potential sources of the observed differ- ences between the two groups, we examine each separately. With respect to determinants, the analysis is similar in each case, except that eli- gibility would not influence the level of receipts of program partici- pants [T| (T > 0)]. However, since the empirical analysis is conducted on aggregates of programs, eligibility for additional programs within categories could have an impact on the level of receipts, given partici- pation. In analyzing the utilization of transfers, it is helpful to distin- guish between welfare (W) and social insurance (S) programs since the two types of programs differ in eligibility requirements and the vari- ables which determine guarantee levels (Lampman, 1976). While specific -7- welfare programs are targeted on particular groups (e.g., disabled and aged persons, broken families), the basic requirement for eligibility is economic need as determined on the basis of current sources of income and also of assets. Guarantee levels are designed to bring actual fam- ily Income up to a stated level of need. Social insurance programs, on the other hand may be viewed as being targeted on individuals who have had a "firm attachment to the labor force" (Lampman, p. 163). Eligibil- ity is determined by employment of a specified duration in a covered sector. The determination of guarantee levels is guided primarily by the replacement ratio principle where a specified portion of the income lost (through disability, unemployment, retirement, etc.) is replaced by the program. The level of the individual's nonwage income or assets is typically not taken into account in determining pajnnents under social insurance programs, however, other indicators of need, such as number of dependents, do play a role in determining benefit levels in some so- cial insurance programs. In light of these distinctions between the two types of programs (W and S) , they will be analyzed separately in the empirical work. B. Immigrants and Transfer Payments The analysis of the receipt of transfer payments by immigrants raises some interesting and unique issues. First, is the complex role played by the amount of time in the U.S. All else equal, the earnings of immigrants have been found to be positively related to time spent in the U.S. (Chiswick, 1978; Blau, 1980). While newcomers may initially be at a disadvantage due to lack of knowledge of customs, job opportuni- ties, etc., their earnings are expected to increase over time as they -8- accumulate country-specific experience and seek out the best opportuni- ties to utilize their skills and abilities. To the extent that higher wages are associated with lower participation in transfer programs, the indirect effect of time in the U.S. on the receipt of transfer payments (through the wage) is likely to be negative. However, controlling for the wage, the direct effect of time spent in the U.S. on transfers is unclear, a priori . On the one hand, although we do not have evidence on this point, it seems likely that the increased labor market informa- tion accumxilated over time should reduce the incidence and/or duration of unemplojment , ceteris paribus . This is expected to reduce the re- ceipt of transfer payments with time spent in the country. On the other hand, over time, individuals may accumulate more information about transfer programs; their greater program skills may work to increase their receipt of transfer payments, all else equal. Further, their eli- gibility for and, in some cases, benefit levels in social insurance pro- grams are expected to increase as immigrants accumulate emplojrment ex- perience in covered sectors. The total effect of time spent in the U.S. on the receipt of transfer payments is thus ambiguous a^ priori and de- pends on the sign and level of the direct and indirect effects. Second, like time in the U.S., the impact of English (speaking and understanding) ability on transfer payment use is a consideration which is primarily of relevance to the foreign bom (see Table 2), As in the former case, the ability to speak and understand English well has an am- biguous total effect on transfer payment use. All else equal, poor English ability is expected to increase the utilization of transfer pro- grams through its negative effect on wages. Controlling for wages, poor -9- English ability may be hypothesized to increase the incidence and/or duration of unemployment through its impact on access to labor market information, resulting in a positive effect on transfer payment utiliza- tion. However, the expected negative impact of poor English ability on program skills and possibly employment in a covered sector may be ex- pected to lower transfer payment use, ceteris paribus . III. The Data The data used in this study are from the 1976 Survey of Income and Education (SIE). The SIE data were gathered nationally on 158,500 house- holds, stratified so as to include more than proportional numbers of households with children living in poverty. The immigrant sample is comprised of all families in which the male head is a foreign-bom male, 18 years of age or over, who arrived in the U.S. during or before 1974, The income data in the SIE refer to calendar year 1975. Those arriving in 1976 were excluded from the analysis, since they did not earn income or receive transfer payments in the U.S. during 1975. In addition, since individuals who arrived during 1975 came after the start of the year, their income and transfer data do not represent a full year, and such individuals have therefore been excluded from the sample. To economize on data processing costs, the native sample is comprised of a random subsample of families from the SIE in which the male head is a native-bom male, 18 years of age or over (native men married to im- migrant women are not included in the native sample) . Individuals bom in Puerto Rico and other U.S. territories were excluded from the sample. No information is available regarding the legal-illegal status of the immigrants in the SIE sample. However, it seems likely that illegal -10- innnigrants are underrep resented, since no special efforts were made to include them. Other evidence suggests that the utilization of govem- ment transfers by illegal immigrants is less than that of legal immi- grants, since the former may be ineligible if their illegal status is discovered (U.S. House of Representatives, 1978) „ Thus, a focus on le- gal Immigrants is likely to give an upper bound on our estimates of the utilization of transfers by immigrants. IV. Empirical Procedures and Results A. Estimation Procedures The probability of participating in transfer programs [P(T > 0)] and the level of receipts, given participation [t|(T > 0)] were esti- mated separately for each type of transfer program (i.e., welfare and social insurance programs). The variables included in the estimating equations may be seen in Table 4 (see Table 3 for variable definitions). The results presented here are for pooled immigrant-native regressions. However, similar findings were obtained when separate regression equa- tions were estimated for each group. As noted earlier, all else equal, the wage is expected to nega- tively affect P(T > 0) and t|(T > 0). However, in the empirical work, we must qualify this prediction somewhat in the case of social insurance programs where benefits are determined on the basis of the replacement principle (i.e., are a positive function of earnings). Since we do not explicitly control for the program parameters in the estimating equa- tions, the sign of the wage variable becones uncertain a priori . The income and asset variables (OTHERY and HOUSEQ) are proxies for the demand for leisure as well as eligibility and benefit levels under -11- welfare programs. In the case of welfare programs, the latter consideration is likely to outweigh the former and to result in a negative sign for these variables. While positive signs on OTHERY and HOUSEQ are expected in the social insurance equations (due to the effects of these variables on the demand for leisure), we may note that their levels are measured at the end of the year. Participants in social insurance programs may have depleted their assets over the period, resulting in spurious negative correlations between the utilization of social insurance transfers and both variables. Further, high levels of income and assets may reflect a taste for work and saving. Thus, due to greater past work experience, persons with higher other income and assets may be more likely to qualify for certain kinds of social insurance payments. The racial and ethnic variables (BLACK, OTHER, SPAN) may reflect the impact of labor market discrimination, controlling for wages and other factors. If, for example, individuals from these groups expect slower future wage growth as a result of labor market experience or en- counter more difficulty locating a job (i.e., have higher unemployment rates), they may be more likely to utilize transfers than comparable white nonhispanics . There may be other racial and ethnic differences in the demand for leisure, all else equal, due to noneconomic factors, but one can only speculate on the signs of such possible effects. The other demographic variables (MAR through HD68PLUS) are proxies for the demand for leisure, as well as eligibility and benefit levels under a variety of transfer programs. Given the various factors repre- sented by these variables, it is not possible to predict all their signs a_ priori . However, a positive association between the age of the head and other family members and the collection of transfers is expected due to both an increased demand for leisure and an increased eligibility for -12- welfare (e.g., supplemental security) and social insurance (e.g., social security retirement and disability) programs. VET is expected to posi- tively affect the utilization of social insurance programs through its impact on eligibility for veterans benefits. The locational variables (SMSA through CENTER) are intended to control for area differences in the eligibility requirements and benefit formulas of transfer programs, as well as the stringency xd.th which the programs are administered at the local level. The inclusion of these variables also adjusts for lo- cational differences in economic conditions and the cost of living. Finally, the variables ENGPOOR through (FOR) (BEFR20) are included to capture the impact of foreign birth on the receipt of transfers. The dumny variable specification of the impact of time spent in the U.S. was dictated by the availability of this information only in categorical form on the SIE tape. As noted earlier, the direct effects of these variables on transfers, controlling for wages and other factors, are un- clear, a^ priori . However, a positive association between the utiliza- tion of social insurance transfers and time spent in the U.S. is likely due to the accumulation of experience in covered sectors. One problem in estimating these equations is that hourly wages can- not be computed for those who were not employed at the time of the sur- vey. It would be possible to use the estimated coefficients from an OLS wage regression based on labor force participants to predict wages for non-participants. However, a possible censoring bias arises from such a procedure, if the probability of being employed is correlated with the wage, given employment. This bias is potentially important in our sam- ple since wages were not observed for 30 percent of the male heads. The possible censoring bias in estimating wages can be eliminated using a -13- technique developed by Heckman (1980) . For each individual i, one in- cludes as an additional explanatory variable in a wage regression: f(B'X.) ^i = ■ F(B'X.)' ''^^^^ f(-) and F(-) are, respectively, the standard normal density and cumxila- tive distribution functions, B is a vector of probit coefficients from an equation estimating the probability of being employed at the time of the survey and X includes all the exogenous explanatory variables pre- sent in the wage and/or transfer equations. The other explanatory variables in the wage equation include the traditional human capital variables, marital status, race and ethnic group, veteran status and the locational variables, as well as dummy variables for English ability, foreign birth and time spent in the U.S. (see Table A-1 for the exact specification of the wage equation) . It should also be noted that a possible censoring bias may ex- ist in the case of the transfer receipt regression, as well, since it is estimated on a censored sample of transfer program participants. However, we have not attempted to correct for the problem in this case, since the explanatory variables included in the transfer probability and re- ceipt equations are identical. Although technically it would be possi- ble to include an estimate of X based on that transfer probability equa- tion in the transfer receipt equation (due to the nonlinear estimation technique), the collinearity between \ and the other explanatory vari- ables would be too severe to allow any meaningful conclusions. Censor- ing bias is essentially an omitted variable problem. To the extent that we have included proxies for the major determinants of the level of transfer payments, it may not be important in this case. Further, even -14- if censoring bias exists, some evidence suggest that its impact is con- siderably larger when values of the dependent variable are inputed for nonparticipants (Smith, 1980). Our use of the transfer receipt equation is primarily limited to the participant group. Finally, there is no reason to suppose that, to the extent censoring bias is present, immi- grants and natives are differentially affected by it in such a way as to bias our findings regarding group differences in transfer receipts in one direction or another. B, Empirical Results Our findings for the estimation of the probability of participating in welfare and social insurance programs [P(T > 0)] and the level of re- ceipts from each type of program, given participation [t|(T > 0)] are given in Table 4. We first briefly consider our findings for the con- trol variables and then proceed to a detailed examination of the impact of foreign-birth on the receipt of transfers. Wages are found to have a significantly negative effect on the probability of participating in both welfare and social insurance pro- grams (despite the ambiguity of their expected effect in the latter case). Specifically, computations based on the estimated coefficients in Table 4 indicate that a 10 percent increase in wages is associated with a 5.5 percent reduction in the probability of participating in wel- fare programs and a 6 percent reduction in the probability of partici- Q pating in social insurance programs. Among participants in social in- surance programs, the estimated wage coefficient is significantly positively related to the level of receipts, possibly due to the impact of the replacement ratio principle. Among participants in welfare pro- grams, the coefficient on LNWAGE is positive but not significant. This -15- may indicate little labor supply response to wage changes among welfare recipients or in part reflect the truncated nature of the sample. The level of nonlabor income and assets is found to negatively in- fluence the probability of participating in both types of transfer pro- grams. It may be recalled that the negative effect of OTHERY and HOUSEQ in the social insurance probit analyses may reflect asset depletion. In this regard, it is interesting to note that, among participants, OTHERY is significantly positively related to the level of social insur- ance payments. (HOUSEQ is negative, but insignificant). Other things equal, members of minority groups are more likely to participate in welfare programs and less likely to participate in social insurance programs, although these effects are not always statistically significant. Among participants in transfer programs, there is some tendency for members of minority groups to receive higher welfare pay- ments, ceteris paribus . The coefficients on the racial and ethnic vari- ables in the social insurance payment regression are not significant and are generally small relative to their standard errors. As noted earlier, the greater reliance of minority individuals on welfare may re- flect the impact of labor market discrimination with respect to wage growth and/or the probability of obtaining employment (as well as group differences in the demand for leisure due to noneconomic factors). The findings in Table 4 suggest that, all else equal, minority males (and/or their spouses) have greater difficulty in qualifying for social insur- ance programs, perhaps due in part to discrimination in gaining access to covered employment. This would also work to increase their utiliza- tion of welfare programs, ceteris paribus . With respect to the impact of the demographic variables, it is par- ticularly interesting, in light of the age composition differences between -16- innnigrant and native households, to note the sizable effect of the age- related variables on the probability of participating in both t3rpes of transfer programs. Considering first the impact of the age of the male head, we see, for example, that having a male head aged 65 to 67 or 68 and over substantially raises the probability of the family receiving welfare payments (by 2.3 and 1.3 percentage points, respectively) rela- 9 tive to a family headed by a male aged 18 to 44, all else equal. Similarly, in comparison to a family headed by a male aged 18 to 44, the family's probability of participating in social insurance programs is 49.3 percentage points higher when the male head is aged 65 to 67 and 67.3 percentage points higher when the head is 68 or over, ceteris paribus . Among social insurance recipients, families headed by a male between the ages of 65 and 67 receive $1706 more, and those headed by a male aged 68 or over receive $1978 more than families headed by an 18 to 44 year old male, other things equal. All these effects are large relative to the means of the dependent variables given in Table 1. (The effects of the age of head variables on welfare receipts, given participation, are posi- tive, but considerably smaller and are not statistically significant). The presence of additional family members over 65 (excluding the »male head) also has a strong positive effect on the collection of trans- fers. An additional older person in the family raises the probability of participation by 3.8 percentage points, in the case of welfare pro- grams, and by 41.9 percentage points, in the case of social insurance programs, ceteris paribus . All else equal, among participants in social insurance programs, an additional older person raises payments received by about $1092. (Again, the effect of an additional older person on welfare payments received by participants is positive, but much smaller and not statistically significant.) -17- As expected, veterans are significantly more likely to participate in social insurance programs which include veterans programs for which only they are eligible. They also receive higher social insurance pay- ments, all else equal, suggesting that veteran's progams are either more generous than other social insurance programs or have a larger work dis- incentive effect or both. We now turn to a consideration of the impact of English (speaking and understanding) ability and the dummy variables for time spent in the U.S. on the utilization of transfers by immigrants. It may be recalled that, all else equal, these variables influence the receipt of transfers both indirectly, through their impact on wages, and directly, holding wages constant. As expected, poor English ability was found to be sig- nificantly negatively related to wages, while length of time in the U.S. had a significant positive effect on wages, all else equal (see Table A-1) . This implies that the indirect effect of poor English ability on the probability of participating in both types of transfer programs is positive, while the indirect effect of time in the U.S. on transfer program participation is negative. It may be recalled that the coeffi- cient on wages in the receipt of transfer payments, given participation regression was positive, although small and insignificant in the case of welfare programs. The estimated direct effects of English ability and time spent in the U.S. on the probability of receiving transfers and on the level of transfers, given participation, are given by the coefficients on these variables in Table 4. Poor English ability of the head significantly raises the probability of the family being on welfare by 1.2 percentage -18- points, all else equal, possibly through its hypothesized positive iia- pact on unemployment of the head. On the other hand, families headed by a male who does not speak and understand English well have a signifi- cantly lower probability (9.4 percentage points less) of collecting so- cial insurance transfers, ceteris paribus , possibly due to the head's difficulty in obtaining employment in the covered sector. Further, among both types of transfer recipients, poor English ability of the head is negatively associated with the level of receipts, although the coefficients on ENGPOOR are not significant. This is consistent with the notion that those with poor English ability have weaker program skills which results in a lower level of receipts. In the case of so- cial insurance programs this finding may also reflect a shorter duration of covered employment among this group. We now consider the direct impact of length of time in the U.S. of the male head on the family's receipt of transfers (Table 4). With re- spect to welfare programs, there is some tendency, albeit not a com- pletely consistent or significant one, for earlier cohorts to be more likely to participate than more recent arrivals, all else equal (includ- ing wages) . This may be due to the greater accumulation over time of 9 information about transfer programs discussed earlier. Among welfare recipients, those who have been in the U.S. a longer period of time ap- pear to have lower receipts, all else equal, although the effects are not significant. Length of time in the U.S. appears to have a stronger direct effect on the collection of social insurance transfers. Controlling for the wage and other factors, the length of time in the U.S. dummy variables -19- are significantly positively related to the probability of participating in social insurance programs. The magnitude of the coefficients tends to rise mono tonically from the more recent entrants to the earlier ar- rivals (until the group that arrived before 1920). Other things equal, families headed by a male who arrived in the U.S. before 1960 are esti- mated to have a 15.4 to 20.5 percentage point higher incidence of par- ticipation in social insurance programs than those headed by more recent (1970-74) arrivals. The coefficients on the year dummies in the social insurance payment regression are also positive, significantly so among those arriving between 1920 and 1949. Taken as a whole, these findings suggest increased utilization of social insurance transfers by immi- grants over time, ceteris paribus , as they gain access to and accumulate experience in covered sectors (with a corresponding increase in eligi- bility and, in some cases, benefit levels). The total effects of these explanatory variables on the utilization of transfers have been estimated on the basis of reduced form equations (Table 5). Taking into account both direct and indirect effects, those who do not speak and understand English well are more likely (than those who do) to receive welfare payments, but less likely to partici- pate in social insurance progrsuns. Those with poor English ability re- ceive a lower level of each type of payment, given that they participate in a transfer program. On net, those who do not speak and understand English well have higher expected payments from welfare programs but lower expected payments from social insurance programs. Their total ex- pected payments are $202 lower than those whose command of Engish is su- perior. This somewhat surprising result is dictated by their lower uti- lization of social insurance programs. -20- The cohort effects in Table 5 have been expressed relative to the native-born to give an indication of the relative utilization of trans- fers of immigrants in comparison to this group, ceteris paribus . Other things equal, immigrant families are estimated to have a lower incidence of welfare dependency and lower welfare receipts, given participation, than native families, regardless of length of time in the U.S. of the male head. As a result, the expected welfare receipts of immigrant fam- ilies range from $17 to $34 less, ceteris paribus , depending on length of residence. On the other hand, while participation in social insur- ance is initially less for immigrant than native families, after about 15 years of residence of the male head, immigrant families begin to have a higher incidence of participation, all else equal. The social insur- ance receipts of immigrant families given participation, are about the same or less than native families, all else equal, when the male head has arrived after 1949. However, among the families of longer-term res- idents, receipts are higher than for native families, ceteris paribus . To obtain an estimate of the relative importance of differences in the response to given characteristics and differences in characteristics in generating a higher utilization of transfers by immigrant families, we turn to the decomposition in Table 6. We first consider immigrant- native differences in the response to a given set of characteristics — or the overall effect of foreign-birth, all else equal, when the length of time in the U.S. variables are evaluated at their mean levels for im- migrants (row 7) . Again, reduced form transfer equations (omitting wages as an explanatory variable) are employed to obtain total effects. But it is interesting to note that the wages of immigrant male heads are -21- found to be about 14 percent higher than those of comparable native's (column 1, row 7). This finding is supportive of the notion of the selectivity of inmigration in terms of ability and/or motivation of which Chiswick (1978) and Blau (1980) have found evidence. Higher immigrant wages, all else equal, work to lower the probability of participating in welfare and social insurance programs of immigrants in comparison to natives, while raising the level of receipts in these programs, given participation. Turning to transfers, we see that in fact immigrant families have a lower utilization of welfare than natives with similar characteristics. Their probability of participating in welfare programs is 1,2 percentage points lower than similar native families (column 2, row 7). This is 52 percent less than the predicted welfare probability for a native fam- ily with native mean characteristics (2.3 percent). Similarly, their level of welfare payments, given participation, is estimated to be 12 percent ($174) lower than the predicted level for a native fa mi ly with mean native characteristics ($1472) . This resvilts in an expected wel- fare payment among immigrant families that is estimated to be 37 percent ($19) lower than the predicted level for a native family with mean na- tive characteristics ($34), On the other hand, immigrant families are as likely to participate in social insurance programs as comparable na- tive families (column 3, row 7). But their receipt of social insurance payments, given participation, is 4 percent ($98) higher than a native family with native mean characteristics ($2611) , Thus, the expected social insurance payment of an immigrant family is estimated to be 4 percent ($37) higher than the predicted value for a typical native family ($998), -22- On net, all else equal, the total expected transfer payment of an innnigrant family is only 2 percent ($18) higher than the predicted level for a native family with mean native characteristics. This reflects a considerably lower (in relative terms) utilization of welfare and a slightly higher (in relative terms) utilization of social insurance by immigrant families than by native families, ceteris paribus . The higher utilization of social insurance programs by immigrant families, ceteris paribus , was solely due to their higher level of receipts, given partic- ipation. This, in turn, is partly a reflection of their greater labor market success (higher wages), all else equal, since wages are posi- tively related to the level of social insurance receipts, given partic- ipation, due to the replacement ratio principle. These findings are con- sistent with the notion that immigrants are a more highly motivated or able group than the native bom. But it is interesting to see that the impact of economic success is not unambiguously to lower transfers. Overall, differences in immigrant-native responses to the same set of variables do not account for a substantial portion of the observed differences in transfer receipts between the two groups. Thus, virtually the entire difference must be due to differences in characteristics. As noted earlier, a variety of characteristics may contribute to the observed differences. Table 6 suggests that, as expected, age and age-related variables play a major role. Age-related factors account for 44 percent of the predicted immigrant-native differential in expected welfare receipts ($7.42/$16.84) ; 96 percent of the predicted differential in expected so- cial insurance receipts ($750.43/$778.72) ; and 95 percent of the pre- dicted differential in total expected transfer receipts ($757.85/ $795.56). As noted earlier, to the extent that the higher utilization -23- of transfers by inmigrants is due to age-related factors, it may not constitute a cause for concern from a policy point of view. Immigrants ' lower levels of education and greater concentration in SMSA's and in the Northeast tend to raise their expected receipts from both welfare and social insurance programs. The higher representation of minority groups among immigrants increases their expected welfare payments but lowers their expected social insurance payments, resulting in lower total transfer receipts. On net, the other personal character- istics of immigrants work to lower their total transfer receipts, with a major factor being the lower proportion of veterans among immigrant male heads (which lowers their utilization of social insurance trans- fers). The higher representation of individuals who have poor English ability reduces their total transfer receipts on balance. This reflects their lower probability of participating in social insurance programs, as well as their lower receipts, given participation, in both types of programs. V. Conclusion Considerable social concern has been expressed over the possibility that immigrants may constitute a burden on the U.S. transfer payment system. The total expected transfer payment to an immigrant family is indeed found to be $571 higher on average than for a native family. However, when these differences are considered more closely, consider- able doubt is cast on the validity of the above concern. We have found that immigrant families are, in fact, considerably less likely to rely on welfare than native families with similar charac- teristics. As a result, their expected receipts from such programs are 57 percent lower than those of comparable native families. Immigrant -24- families are about as likely to participate in social insurance programs as native families with the same characteristics, but their receipt of payments from such programs, given participation, is somewhat higher. As a result, the expected social insurance payment of an immigrant fam- ily is found to be slightly (4 percent) higher than a comparable native family. Overall, we estimate that, on average, the total expected transfer payment to immigrant families is only 2 percent higher than their native counterparts. Thus, behavioral differences in the response of imm igrant and native fa m ilies to the same characteristics account for a negligible portion of the observed higher receipts of transfer payments by immigrant families. This implies that the observed differ- ence between the two groups is almost entirely due to group differences in characteristics. With respect to specific characteristics, we have found that immigrant-native differences in age-related variables are the primary cause of the higher utilization of transfers by immigrant families, ac- counting for 95 percent of the total immigrant-native difference. Since the age distribution of immigrants is determined by historical trends in immigration (particularly the adoption of a more restrictive immigra- tion law after World War I) rather than by domestic birth and death rates, immigrants tend to be older, on average, than native-born indi- viduals. This factor tends to increase their utilization of transfers, all else equal. However, since it is likely that the vast majority of older immigrants have spent most of their working lives in the U.S. and have working age children who contribute to tax revenues, a higher uti- lization of transfers of immigrants due to age-related factors does not appear to unduly burden the transfer payment system. -25- In considering the policy implications of these findings, it is im- portant to point out that they are in part dependent on the past compo- sition of the immigrant group. If, in the future, we were to see, for example, a widening of the immigrant-native disparity in educational attainment or an increase in the proportion of refugees among immigrants, the relative utilization of transfers by immigrants might begin to pose a social problem. However, for the most part, these findings suggest grounds for cautious optimism. For example, one issue regarding recent immigrants is the high representation of minority group individuals among them. But, as we have seen, the higher representation of minority groups among immigrants in the past has actually worked to lower their total transfer receipts, on net: increasing their expected welfare pay- ments but lowering their expected social insurance pajnnents. Similarly, the higher representation among immigrants of individuals who have poor English ability has also reduced their total transfer receipts, on balance due to their lower probability of participating in social in- surance programs, as well as their lower receipts, given participation, in both types of programs. Finally, it is possible that the immigra- tion decisions of the post-1949 cohorts of immigrants were influenced to a greater extent than earlier cohorts by the availability of more generous transfers in the U.S. than in other countries. However, when the analysis was replicated including only immigrants who had arrived after 1949, essentially the same results were obtained. -26- Footnotes For an attempt to balance out the costs and benefits of Immigra- tion at an aggregate level, see Simon (1980). For a study dealing with illegal immigrants, see North and Houston (1976). The U.S. House of Representatives, Report of the Special Committee on Population (1978) states: "Tmmi grants undoubtedly have an effect on the cost of provid- ing a whole range of public services including welfare, medical care, education, public housing, fire and police protection, sani- tation, transportation, and recreation. Few of these effects, how- ever, have received more than cursory attention by immigration re- searchers or by the administrators of these public services." 2 The nature and significance of the distinction between welfare and social insurance type programs is explained in detail below. 3 Greater ability or motivation raises both the opportunity cost of migrating (foregone earnings in the place of origin) and the return to migration (expected earnings tn the place of destination). However, the direct costs of migrating wotild be the same for all individuals regard- less of ability or motivation. Thus greater ability or motivation is expected to raise the returns to migration more than the costs. 4 Outmigration is a relatively minor factor in the case of the United States. Using 1970 census data, Chiswick (1977) has found that, other things equal, the native-born sons of immigrants have higher earnings than the native-bom sons of native-born parents. For a similar finding using data from the early 1900 's, see Blau (1980). See, for example, Lyon (1977), Abrahamse, et al. (1976), Hamermesh (1979), Parsons (1980). In 1973, the unenployment rate of males 20 years of age and over was 11.7 percent among Blacks and other nonwhites, 9.7 among those of Spanish origin, but only 6.2 percent among whites (Employment and Train- ing Report of the President , 1976, p. 223). Q The computation of these wage effects is illustrated in the case of social insurance programs (S) where the percentage change in the par- ticipation probability for a 10 percent increase in wages is given by: In 1.1 • B^g f(B^X)/ SOC , where By„ is the estimated probit wage coefficient for participating in social insurance programs; Bg is the estimated vector of probit coefficients for participating in social insurance programs; f (-) is the density -27- f unction for a standard normal random variable; SOC is a dummy variable equaling 1 if the individual particpates in social insurance programs, and otherwise (i.e., the dependent variable in the social insurance probit analysis); and a line over the expression indicates that the weighted sample mean value is employed. 9 These partial derivatives are computed analogously to the wage effects explained above (footnote 8), i.e., in the case_of social insur- ance programs, for an exogenous variable (X, ) ; B, f(B'X). As may be seen in Table A-1, the coefficient on X is signifi- cantly negative. The use of coefficients from wage regressions includ- ing A to estimate predicted wages for the full sample .(including nonla- bor force participants) resulted in lower mean estimates than the use of coefficients from wage regressions that did not include an adjustment for selectivity bias. This suggests that employed individuals have unob- servables that are positively related to wages. "'■■'"See Table A-2. -28- References Abrahamse, Allan; David Deferranti; Patricia Fleischauer; and Albert Lipson, Welfare Caseload Estimating Techniques; A Survey and Evaluation . The Rand Corporation, R-1916-CIX)BP, January 1976. Blau, Francine D., "Immigration and Labor Earnings in Early Twentieth Century America." Research in Population Economics, Volume 2 . Ed. by Julian L. Simon and Julie Da Vanzo (Greenwich, Conn.: JAI Press, Inc., 1980). Chiswick, Barry R. , "Sons of Immigrants : Are They at an Earnings Dis- advantage?" American Economic Review Papers and Proceedings , Vol. 67, No. 1 (February 1977), pp. 376-80. Chiswick, Barry R. , "The Effects of Americanization on the Earnings of Foreign-born Men." Journal of Political Economy , Vol. 86, No. 5 (October 1978), pp. 897-921. Hamermesh, Daniel S., "Entitlement Effects, Unemployment Insurance and Employment Decisions." Economic Inquiry , Vol, 17, No, 3 (Jxxly 1979), pp. 317-332. Heckman, James, "Sample Selection Bias as a Specification Error." Female Labor Supply . Ed, by James P. Smith (Princeton: Princeton Univer- sity Press, 1980). Lampman, Robert J., "Employment versus Income Maintenance." Jobs for Americans . Ed. by Eli Ginzberg (Englewood Cliffs: Prentice-Hall, Inc., 1976). Lyon, David W., "The Dynamics of Welfare Dependency: A Survey," mimeo- graph. Duke University, Institute of Policy Sciences and Public Affairs, and the Ford Foundation: Welfare Policy Project, Spring, 1977. North, David S. and Marian F. Houston, The Characteristics and Role of Illegal Aliens in the U.S. Labor Market; An Exploratory Study (Washington: Linton and Company, 1976) . Parsons, Donald 0., "Racial Trends in Male Labor Force Participation." American Economic Review , Vol. 70, No. 5 (December 1980), pp. 911-920. Simon, Julian, "What Immigrants Take From, and Give to, the Public Coffers." Final Report to the Select Commission on Immigration and Refugee Policy, August 15, 1980. Mimeograph. -29- Smith, James P., "Introduction." Female Labor Supply . Ed. by James A. Smith (Princeton: Princeton University Press, 1980). U.S. Department of Labor, Employment and Training Administration. Employ- ment and Training Report of the President (Washington, D.C.: Government Printing Office, 1976). U.S. Hoiise of Representatives, Special Committee on Population, Legal and Illegal Immigration for the United States . US//962-17 (1978). Wachter, Michael L., "The Labor Market Outlook for the 1980's." Industrial and Labor Relation Review . Vol. 33, No. 3 (April 1980), pp. 342-354. M/D/340 TABLE 1 RECEIPT OF TRANSFER PAYMENTS BY IMMIGRANT AND NATIVE FAMILIES, 1975*' Natives Iimnl grants Welfare [W] $ 71.59 $ 98.09 Social Insurance [S] $ 997.35 $1541.64 Total [W + S] $1068.94 $1639.73 Participation in Welfare [P(W > 0)] .045 .057 Participation in Social Insurance [P(S > 0)] .369 .464 Welfare, Given Participation [W|(W>0)3 $1584.53 $1708.3^ Social Insurance, Given Participation [S|(S > 0)] $2700.70 $3324.40 Number of observations 7205 5730 welfare payments include income received by the family from public as- sistance, welfare (including Aid to Families with Dependent Children) or supplemental security income. Social insurance payments include income received by the family from social security, railroad retire- ment, veterans payments, unemployment compensation or workman's compen- sation. b Observations are weighted by sampling weights reported in the SIE. Source: Survey of Income and Education, 1976. TABLE 2 MEANS AND STANDARD DEVIATIONS a,b Variable Natives Mean S.D. Mean Immigrants S.D. EDUC EXP EXP'' BLACK OTHER SPAN MAR VET OTHERY HOUSEQ SPEDUC PER18T64 PERGE65 KIDSLT18 HD45T59 HD60T64 HD65T67 HD68PLUS ENGPOOR ENTRY COHORT 1965-69 1960-65 1950-59 1920-49 BEFR20 SMSA SOUTH WEST CENTER predicted lnwage'^ 12.045 27.799 1086.674 .087 .006 .019 .851 .480 863.433 15646.647 10.265 1.021 .096 .945 .274 .071 .040 .097 .001 .908 .331 .180 .281 1.474 3.324 17.717 1161.369 .282 .076 .136 .356 .500 3204.229 18720.127 4.893 .748 .304 1.326 .446 .258 .196 .296 .032 * .289 .470 .384 .450 .349 10.783 35.676 1691.963 .035 .095 .231 .825 .184 1162.990 15359.953 8.843 .923 .209 .825 .221 .067 .048 .256 .114 .132 .105 .195 .358 .070 .967 .178 .272 .168 1.379 4.695 20.474 1604.384 .185 .293 .421 .380 .387 3903.915 19561.713 5.442 .823 .419 1.281 .415 .250 .213 .436 .317 .339 .306 .396 .480 .256 .180 .382 .445 .373 .458 a. See Table 3 for variable definitions. Observations are weighted by sampling weights reported in the SIE. 'Predicted on the basis of the wage regression with selectivity bias correction presented in Table A-1. *Not applicable. TABLE 3 VARIABLE DEFINITIONS EDUC Highest grade completed. EXP Potential experience = age - education - 5 (constrained to be greater than or equal to 0). EXP2 Potential experience squared. BLAOC Equals 1 if the male head is black, and otherwise. OTHER Equals 1 if the male head is other nonwhite, and otherwise. SPAN Equals 1 if the male head is of Spanish origin, and other- wise. MAR Equals 1 if the male head is married, spouse present, and otherwise. VET Equals 1 if the male head is a veteran, and otherwise, OTHERY Amount of the family's nonlabor, nontransfer income. HOUSEQ Market value of the house minus value of the mortgage out- standing. SPEDUC Equals highest grade completed of spouse, if married, and otherwise. PER18T64 Number of family members, excluding male head, aged 18-64. PERGE65 Number of family members, excluding male head, aged 65 and over. KIDSLT18 Nvimber of children less than 18 years of age. HD45T59 Equals 1 if the male head is aged 45-59, and otherwise. HD60T64 Equals 1 if the male head is aged 60-64, and otherwise. HD65T67 Equals 1 if the male head is aged 65-67, and otherwise. HD68PLUS Equals 1 if the male head is aged 68 or more, and otherwise. ENGPOOR Equals 1 if the male head does not speak or understand English well, and otherwise. FOR Equals 1 if the male head is foreign born, and otherwise. (FOR) (EDUC) Equals highest grade completed by the male head if foreign- born, and otherwise. TABLE 3 (cont.) VARIABLE DEFINITIONS (FOR) (1970-74) Equals 1 if the male head is foreign born and entered the U.S. between 1970 and 1974, and otherwise (refer- ence category in the regressions). (FOR) (1965-69) Equals 1 if the male head is foreign born and entered the U.S. between 1965 and 1969, and otherwise. (FOR) (1960-64) Equals 1 if the male head is foreign born and entered the U.S. between 1960 and 1964, and otherwise. (FOR) (1950-59) Equals 1 if the male head is foreign born and entered the U.S. between 1950 and 1959, and otherwise. (FOR) (1920-49) Equals 1 if the male head is foreign born and entered the U.S. between 1920 and 1949, and otherwise. (FOR)(BEFR20) Equals 1 if the male head is foreign born and entered the U.S. before 1920, and otherwise. LNWAGE Natural log of the hourly wage. WELFARE Amount of income received by the family from public assis- (W) tance, welfare (including Aid to Families with Dependent Children) or supplemental security income. SOCIAL Amount of income received by the family from social security, INSURANCE railroad retirement, veterans payments, unemployment compen- (S) sation or workman's compensation. SMSA Equals 1 if the family resides in an SMSA, and otherwise. SOUTH Equals 1 if the family resides in the South, and otherwise. WEST Equals 1 if the family resides in the West, and otherwise. CENTER Equals 1 if the family resides in the North Central U.S., and otherwise. TABLE 4 REGRESSION RESULTS Variables Participation Probability (Probit)' Level of Receipts Among Participants (OLS) Welfare Social Social Insurance Welfare Insurance -.104x10"^** (.435x10'"^) -.437x10"^*** (.812x10"^) -.008 .022*** (.043) (.007) -.008 -.001 (.005) (.001) -.036 338.23** 77.39 (.062) (168.26) (126.48) -.165*** 254.63 -142.66 (.063) (205.57) (124.61) -.109** 437.19** 121.18 (.055) (182.44) (108.75) .432*** 145.38 393.40*** (.079) (237.91) (121.14) -.035*** -16.86 -7.14 (.006) (20.61) (9.32) .187*** 148.38** 124.34*** (.021) (69.35) (37.36) 1.054*** U8.22 1092.40*** (.060) (136.84) (67.89) -.036*** 97.97** 67.49** (.011) (37.97) (26.63) -.034 274.73* 325.28*** (.036) (151.54) (81.86) .327*** 254.91 791.76*** (.053) (261.65) (107.14) 1.240*** 79.15 1705.50*** (.075) (266.60) (109.21) 1.691*** 200.39 1978.10*** (.076) (234.06) (102.02) .417*** -68.71 140.24** (.031) (136.14) (56.78) -.237*** -115.46 -124.45 (.078) (201.95) (135.35) -.289*** -109.16 -157.59 (.070) (253.48) (174.18) .283*** 80.63 129.64 (.081) (302.19) (203.89) .334*** 105.74 138.84 (.088) (323.62) (215.94) .403*** -60.41 118.65 (.077) (304.10) (187.92) OTHERY HOUSEQ BLACK OTHER SPAN MAR SPEDUC PER18T64 PERGE65 KIDSLT18 HD45T59 HD60T64 HD65T67 HD68PLUS VET ENGPOOR FOR (FOR) (1965-69) (FOR) (1960-64) (FOR) (1950-59) -.412x10" J** (.163x10':) -.186x10"^*** (.185x10"^) .494*** (.078) .268*** (.089) .286*** (.077) -.106 (.104) -.032*** (.008) .293*** (.030) .637*** (.065) .136*** (.017) .168*** (.062) .079 (.103) .392*** (.111) .223** (.100) -.021 (.053) .195** (.094) -.425*** (.108) .083 (.126) .250* (.136) .191 (.125) TABLE 4 (cont.) REGRESSION RESULTS Variables Participation Probability (Probit)' Level of Receipts Among Participants (OLS)' Social Social Welfare Insurance Welfare Insurance (FOR) (1920-49) .218* .515*** -295.75 294.61* (oll7) (.077) (274.83) (177.46) (FOR) (BEFR20) .074 .388** -98.17 152.82 (.149) (.162) (341,43) (196.34) PPFniCTED LNWAGE -.491*** -.660*** 21.77 249.93*** (.082) (.069) (180.90) (83.55) SMSA .152 .020 202.40 34.36 (.104) (.061) (255.77) (107.29) SOUTH -.068 -.279*** -550.41*** -132.45* (.067) (.041) (163.40) (77.23) WEST .019 -.177*** ■ -297,58** -137.71** (.060) (.036) (151.56) (65.28) CENTER -.148** -.176*** -374.48** -44.31 (.066) (.038) (169.98) (66.61) CONSTANT -1.167*** .226** 1135.00*** 773.78*** (.157) (.112) (374,93) (167.08) -2xL0G LIKELI- HOOD RATIO ADJUSTED R 837.260 5517.138 ,051 .293 N OF OBSER- VATIONS 12935 12935 595 5460 (a) Asymptotic standard errors in parenthesis (b) Standard errors in parenthesis *Significant at the 10% level on a two-tailed test **Signif icant at the 5% level on a two-tailed test ***Signif icant at the 1% level on a two-tailed test TABLE 5 TOTAL EFFECTS OF FOREIGN BIRTH, LENGTH OF TIME IN THE U.S. AND ENGLISH ABILITY ON THE RECEIPT OF TRANSFER PAYMENTS^ Participation Receipts, Given Expected Transfer Variable Probability Participation*^ Receipt" Social Social Social Welfare Insurance Welfare Insurance Welfare Insurance Total Foreign^ and Arrived: 1970-74 -.024 -.120 -163.07 -162.68 -33.53 -353.48 -387.01 1965-69 -.022 -.039 -95.49 -1.27 -28.96 -90.19 -119.15 1960-64 -.012 -.022 -66.49 14.86 -16.84 -44.74 -61.58 1930-59 -.015 .016 -230.89 -12.24 -23.49 31.14 7.65 1920-49 -.010 .056 -553.10 177.38 -25.45 2U.01 185.56 Before 1920 -.011 .051 -415.65 109.70 -23.32 168.66 145.34 ENGPOOR .019 -.066 -227.22 -143.28 17.55 -219.90 -202.36 Computed on the basis of reduced form equations, see Table A-2. The partial derivative of the probability of participating in a transfer program (PT) with respect to an exogenous variable (X. ) is computed as: where 6 is the estimated probit coefficient on X, and f (3'X) is the value of the normal density function evaluated at the weighted sample means of the exogenous variables. The partial derivative of the level of receipts, given participation (T) with respect to an exogenous variable (X, ) is the estimated OLS regression coefficient. The expected transfer receipt (ET) is defined as: ET = PT X T where PT and T have been defined above. The partial derivative of the expected transfer receipt with respect to an exogenous variable (X. ) is computed as: 9ET _ 3PT 9T 3Xj^ ~ X 3X^ X 3Xj^ where T— and PT— are evaluted at the weighted sample means of the exogenous variables, %ote that all length of time in the U.S. effects are expressed relative to native- born individuals. The (FOR) (EDUC) interaction term is evaluated at the weighted sample mean level of EDUC. DECOMPOSITION OF THE IMMIGRANT -NATIVE WAGE AND TRANSFER DIFFERENTIALS^ Characteristic Wage (1) Participation Probability^ Receipts, Given , Participation Expected Trans Receipts fer Welfare (2) Social Insurance (3) Welfare (4) Social Insurance (5) Welfare (6) Social Insurance Total (7) (8) Race or Ethnic Group -.0281 .0034 -.0076 133.53 1.78 8.51 -19.16 -10.65 Age-related Factors -.0960 .0040 .1657 57.06 580.02 7.42 750.43 757.85 Education^ -.0732 .0066 ,0308 -21.93 -10.70 9.07 76.03 85.10 Other Personal , Characteristics -.0254 .0028 -.0301 0.71 -35.40 4.14 -91.06 -86.92 Location .0105 .0022 .0153 85 . 20 25.60 5.37 50.14 55.51 Poor English Ability -.0127 .0022 -.0076 -65.32 -13.74 1.61 -24.97 -23.36 Foreign-birth-" - .1306 -.0118 .0 -173.98 97.59 -19.28 37.31 18.03 Total Predicted Differential*^ -.0943 .0094 .1665 15.27 645.15 16.84 778.72 795.56 insfer decompositions based on reduced form equations, see Table A-2. the case of the OLS regressions, e.g. for the level of receipts, given participation (T) , the ierential associated with the 1. . .m variables included in characteristic C (DC_) is computed m m I DC„ = Z B.CX.^. - X,.-) = E B.X._ + Z B.X,., - Z B.X,., T . , 1 il IN . T i il . . 1 i iN , , i iN 1=1 1=1 i=m+l i=l m = Z B.X.^ - ( E B.X.-- + E B.X.t) . T 1 il . T i iN . _.T i il i=l i=l i=m+l :e B. is the estimated regression coefficient on X,; X , ^ and X. are the weighted means of X. immigrants and natives respectively; and there are n variables. Note that the weighted means computed for the relevant subgroup, i.e., welfare or social insurance program participants. 1 the case of the probit analyses of transfer participation probabilities (PT), the differential liciated with the l...m variables included in characteristic C (DCp„) is computed as: DCp^ = F( E B^.j + /_/?iN^ - ^V/^iN^ i=l i=i!rH i=l I re B is the estimated probit coefficient on X ; F( - ) is the cumulative distribution function a standard normal variable; and the other symbols have been defined above. Note that in the rical work DCp_ as con5>uted above was found to be approximately equal to Continue TABLE 6 m n [F(_ZB.X.^) -F(ZB.X.^+ .' /Al^^- x=l i=l i=m+-l In the case of the expected transfer payment (ET) the differential associated with the l...m variables included in characteristic C (DC_„) is computed as : 1=1 i=m+l 1=1 i=nt+-l 1=1 i=l ?here the symbols have been defined above. 'Includes BLACK, OTHER, SPAN. "Includes EXP, EXP2, PER18T64, PERGE65, KIDSLT18, HD45T59, HD60T64, HD65T67, HD68PLUS. 'Includes EDUC. ^Includes OTHERY, HOUSEQ, MAR, SPEDUC, VET. "Includes SMSA, SOUTH, WEST, CENTER. Includes FOR, (FOR) (EDUC), (FOR) (1965-69), (FOR) (1960-64) , (FOR) (1950-59) , :fOR) (1920-49), (FOR) (BEFR20) . 'Sum of rows (1) through (7). TABLE A-1 WAGE REGRESESSIONS (STANDARD ERRORS) Without Selectivity With Selectivity- Variables Bias Correction Bias Correction EDUC .058*** .058*** (.003) (.003) EXP o035*** .038*** 2 EXP (.002) -.552x10"^*** (.293x10 ) (.002) - -.658x10"^*** (.479x10"^) BLACK -.101*** -.104*** (.032) (.032) OTHER -.007 -.013 (.031) (.031) SPAN -.159*** -.153*** (.027) (.027) MAR .126*** .130*** (.019) (.020) VET .059*** .074*** (.016) (.016) ENGPOOR -.124*** -.113*** (.041) (.042) FOR .175*** .196*** (.061) (.061) (FOR) (EDUC) -.016*** -.018*** (.004) (.004) (FOR) (1965-69) .109*** .117*** (.038) (.038) (FOR) (1960-64) .151*** ,143*** (.042) (.042) (FOR) (1950-59) .105*** .108*** (.036) (.036) (FOR) (1920-49) .144*** .136*** (.038) (.038) (FOR)(BEFR20) .398*** .334*** (.117) (.120) SMSA .085*** .091*** (.031) (.031) SOUTH -.040* -.046** (.021) (.021) WEST -.041** -.049*** (.018) (.019) CENTER -.021 -.023 (.019) (.019) TABLE A-1 (cont.) Variables Without Selectivity Bias Correction With Selecti-vity Bias CorrectidTt" CONSTANT ADJUSTED R .317*** C.060) .139 -.191*** G068I .237*** C.066) ,139 Number of Observations 9116 9116 *Signi£icant at the 10% level on a two-tailed test **Significant at the 5% level on a two-tailed test ***Significant at the 1% level on a two-tailed test TABLE A-2 REDUCED FORM REGRESSIONS Participation ^ Level of Receipts Among L pants (OLS) Probability (Probit)° Partic: Variables Welfare Social Insurance Welfare Social Insurance EDUC -.087*** -.062*** 27.89 6.44 (.012) (.007) (26.81) (13.48) EXP -.009 -.013*** 32.93** 10.83 (.006) (.004) (15.73) (8.40) EXP^ .101x10^? .804x10"^ .137x10^?* (.747x10"^)' -.209 -.192** (.185) (.092) BLACK .494*** .015 329.67** 43.29 (.079) (.062) (167.57) (126.44) OTHER .240*** -.149** 308.48 -148.66 (.089) (.063) (206.21) (124.71) SPAN .310*** -.030 453.22** 80.60 (.076) (.054) (180.09) (108.35) MAR -.262** .238*** 96.39 414.01*** (.112) (.085) (248.52) (129.29) VET -.045 .361*** -78.86 160.81*** (.052) (.030) (136.06) (56.24) OTHERY -.372x10^** (.160x10 ) -.100xlO"t** (.436x10"^) -.011 -.022*** (.043) (.007) HOUSEQ -.183x10"^*** (.186x10"^) -.445x10"^*** (.822x10"^) -.008* (.005) -.001 (.001) SPEDUC -.025*** -.027*** -7.88 -6.18 (.009) (.006) (21.39) (9.83) PER18T64 .291*** .185*** 136.15** 123.94*** (.030) (.021) (69.42) (37.40) PERGE65 .636*** 1.062*** 119.71 1093.30*** (.066) (.060) (137.59) (68.30) KIDSLT18 .125*** -.047*** 78.27** 65.40** (.018) (.012) (38.78) (27.52) TABLE A-2 (cont'd). REDUCED FORM REGRESSIONS Participation Probability (Probit)^ Level of Receipts Among Participants (OLS) Variables Welfare Social Insurance Welfare Social Insurance HD45T59 .108 (.102) -.014 (.061) -144.56 (247,24) 316.31** (140.80) HD60T64 .084 (.161) .448*** (.097) -359.92 (391.97) 716.90*** (194.22) HD65T67 .426** (.179) 1„423*** (ol22) -563.66 (421.64) 1707.2*** (208.22) HD68PLUS .379* (.199) 2.014*** (.150) -503.54 (463.96) 1993.80*** (230.03) ENGPOOR .327*** (.097) -.167** (.080) -227.22 (211.32) -143.28 (136.99) FOR -1.128*** (.175) -.503*** (.119) 251.47 (388.02) -192.55 (237.84) (FOR) (EDUC) .063*** (.012) .018** (.008) -36.35 (29,53) 2.62 (13.91) (FOR) (1965-69) .041 (.125) .205** (.081) 67.58 (301.13) 161.41 (203.91) (FOR) (1960-64) .201 (.135) .246*** (.088) 96.58 (322.74) 177,54 (215.82) (FOR) (1950-59) .165 (.125) .343*** (.077) -67.82 (304.02) 150.44 (188.06) (FOR) (1920-49) .251** (.119) .443*** (.078) -390.03 (283.51) 340.06* (179.02) (FOR) (BEFRZO) .233 (.166) .431** (.178) -252.58 (378.47) 272.38 (209.75) SMSA .109 (.104) -.036 (.060) 205.61 (254.24) 59.65 (106.86) SOUTH -.059 (.067) -.241*** (.041) -546.05*** (163.16) -145,51* (77.46) Variables TABLE A-2 (cont'd.) REDUCED FORM REGRESSIONS Wei fare Participation Probability (Probit)' Social Insurance Level of Receipts Among Participants (OLS) Wei fare Social Insurance WEST CENTER CONSTANT -2XL0GLIKE- LIHOOD RATIO ADJUSTED R .063 (.060) -.135** (.066) -.657*** (.212) 874.790 -.135*** (.036) -.158*** (.038) .299** (.128) 5541.827 -325.23** (151.38) -404.70** (170.00) 436,66 (502.52) -147.98** (65,34) -50.32 (66.71) 921.50*** (248.95) .056 .293 N OF OB- SERVATIONS 12935 12935 595 5460 (a) Asymptotic standard errors in parenthesis (b) Standard errors in parenthesis *Significant at the 10% level on a two-tailed test **Significant at the 5% level on a two-tailed test ***Significant at the 1% level on a two-tailed test ffiSllS^SWIIi^^ism.it,^^ M'i'^';tifwi« HECKMAN BINDERY INC. JUN95 s N MANCHESTER.! B„„^.To-Pl">f INDIANA 46962