Estimates for the preferred states only

5. Estimates for the preferred states only

The magnitude of the welfare motivation in both the full immigrant sample and with the female only sample raises some suspicion. According to the results for the full sample a 1 percent increase in the maximum combined AFDC and food stamp benefits on average will increase the inflow of immediate relative immigrant admissions by 8.8 percent, 31.69 percent for family- sponsored immigrants and 40.62 percent for employment-based immigrants. The largest elastic- ity is in the Refugee and asylee category. The same increase in potential welfare benefits will produce a 150.98 percent increase in these immigrants locating in the state. Females produce comparable results. The large elasticity for refugees is not completely surprising. Borjas 1995 maintains that refugees are relatively more likely to participate in means tested benefit programs. Furthermore, the large elasticities with respect to welfare only confirm the welfare motivation advanced and tested by Buckley 1996. On the other hand, the level of disaggregation in this data could be culpable in producing suspiciously large elasticities. Table 1 clearly indicates that some states are preferred by immigrants, indeed, and there are several states with very small immigrant inflows. Furthermore, these small inflows are further separated by admission class producing many zero observations as discussed before. For example, the fourteen family-sponsored immigrants who located in Wyoming only represent four countries of origin; four from China, one from India, three from Korea, and six from Vietnam. In this example there are seventeen zero observations for the state of Wyoming. Among the observations for refugees from Guyana there are forty-five zero observations out of forty-nine. The small numbers that result from this dissagregation may create an overestimate of the median elasticities for the determinants. These small numbers will, of course, indicate large percentage changes with small changes in the levels. Any elasticity will be overstated due to these small levels. In order to account for any small number problems in the data a further specification is considered. In this case the determinants of location choice for immigrants who intend to reside in the preferred states only are examined. The dependent variable is now the number of immigrants from each of the twenty-one countries of origin who locate in those states with a total immigrant inflow of 5,700 or more in 1991. These thirteen states are selected for three reasons 17 . First, these states have total immigrant inflows that are cut off by the seventy-fifth percentile of the distribution. Since any immigration policy analysis will be more important to states with large inflows of immigrants, this cutoff may describe results that are more policy oriented. Second, the states in this upper fourth of the distribution are those that have been preferred by immigrants in the recent past. Between the fiscal years of 1988 and 1996, 9.42 million legal immigrants entered the United States, of this total these states absorbed 8.02 million or 85 percent. Third, these states have significantly fewer zero observations. This reduction in zeros will reduce any possibility of overstated elasticities. Table 6 reports the elasticities for this estimation. Specification tests were also run on this model. In the same manner as before log-likelihood ratio tests indicate that the model is well 17 The states are California, New York, Texas, Florida, New Jersey, Illinois, Massachusetts, Virginia, Pennsylvania, Washington, Maryland, Michigan, and Connecticut. 63 M. E. Dodson III International Review of Law and Economics 21 2001 47– 67 specified. In the first test all the coefficients were restricted to zero, leaving only the constant term. In this test, the null hypothesis was rejected in each case. The x 2 test statistics with 67 degrees of freedom were 1070.23, 899.296, 1110.41, and 560.659 for the family sponsored, employment based, immediate relative and refugee class respectively. All are significant test statistics and reject the restricted model at the 1 percent level. In a second test, all coefficients except the country dummies were restricted to a value of zero. If the null cannot be rejected in this case then any location determinants are unique to each source country and the state factors play no statistical part in the location decisions. As with the previous models the null was rejected in each case at the 1 percent level. The chi-square test statistics, in the same order as above, were 991.278, 700.175, 1043.99, and 343.707. The elasticities reported in Table 6 provide support for the earlier results and indicate that immigrants who migrate to these preferred states do display welfare motivation. Further- more, the elasticities are more reasonable indicating that the states with smaller immigrant flows did inflate the original elasticities. Table 6 indicates that welfare generosity is still a significant determinant in the location decision even when controlling for nativity factors. Like the results above, female immigrants who locate in preferred states also display a strong attraction to welfare generosity. In both the full sample and the female sample, every class of admission displays the positive and significant correlation with welfare payments. The results for female immigrants in these preferred states are found in Table 7. Table 6 Elasticity estimates for the preferred states Family Sponsored Immediate Relatives Refugees Asylees Employment Based Per Capita Gross State Product 21.159V 20.605 20.182 20.755 0.614 0.437 2.446 0.382 Welfare 6.614 3.348 14.412 V 4.701 1.914 1.316 7.519 1.174 Percent of Total Population that is Similarly Born 0.082 0.019 0.231 0.003 0.009 0.006 0.033 0.006 Similarly Born Recent Arrivals 0.179 0.158 20.007 0.04 0.01 0.007 0.029 0.005 Agricultural Share of Employment 23.179 21.055 20.029 21.837 1.47 1.002 5.804 0.906 Goods Producing Share of Employment 28.686 22.241 22.845 20.704 3.221 2.237 12.566 1.967 Total Population 1.931 1.122 4.975 1.326 0.59 0.411 2.32 0.365 Employment Growth 1.714 1.004 0.041 1.522 1.21 0.829 4.813 0.755 Unemployment Rate 5.305 1.491 3.619 21.037V 4.147 2.826 16.769 2.62 Sales Tax 3.756 1.546 1.022 20.295 1.912 1.327 7.479 1.179 Metropolitan Population 24.681 22.109 21.635 3.316 4.267 2.934 16.782 2.683 States included are CA, NY, TX, FL, NJ, IL, MA, VA, PA, WA, MD, MI, CT. A significance level of 1 is denoted by , 5 by , and 10 by V. 64 M. E. Dodson III International Review of Law and Economics 21 2001 47– 67 These estimates also indicate that the stock of similarly born immigrants does impact the location decision. Like Zavodny’s 1997 conclusion, these results suggest that similarly born immigrants are an important factor in the location choice. Additionally, the recent inflow of similarly born immigrants influences the location decisions of immigrants in each class of admission except refugees. However, as before her contention that country of origin is the dominant determinant in this decision finds no support here. There is relatively little difference between the full sample estimates and the estimates for the preferred states. State population is still a positive and significant determinant. Only the employ- ment-based immigrants respond significantly to the unemployment rate. None of the admission categories display a relationship between location decisions and employment growth. Only immigrants in the family-sponsored category react to the industrial make-up of the state.

6. Conclusions