Conclusions
WEAKNESSES IN THE CASE STUDIES
Workshop participants noted that estimates of the illegal immigrant population have been made for only selected time points and generally only for states. As a result, the case studies usually needed to extrapolate from available point estimates and make assumptions about the number of illegal immigrants in smaller areas, such as counties or metropolitan areas. Moreover, data on the characteristics of the illegal alien population are generally available only for national aggregates, requiring strong assumptions in order to develop subnational estimates.
One of the most difficult challenges in calculating fiscal costs is first obtaining accurate estimates of the public service costs per illegal immigrant. To produce reliable estimates of the fiscal costs of illegal immigration, data are needed on the number of illegal immigrants, the actual use of services, and the cost of providing these services. The types of public services used and characteristics of the illegal immigrant are also important. For example, education cost estimates require information on the illegal alien population enrolled in school and per pupil cost data for all enrolled students.
The case studies estimated costs on either an average or a marginal basis. In most cases, costs were estimated on the basis of average costs; assuming that each person in the population had an equal basis—or an equal "share"—of total operating costs. But marginal costs, calculated as the cost of each additional public service recipient to an existing public service system, may differ significantly from average costs. Workshop participants noted that whether costs are
estimated on an average or marginal basis is a critical issue, and it is not adequately addressed in either the case studies or studies on immigration effects.
Workshop participants noted that there are instances in which marginal costs are below average costs (e.g., an additional car on an uncongested highway) as well as instances in which marginal costs are above average costs (e.g., an additional car on a congested highway). Some public service costs, such as police protection, increase on a per capita basis as city population increases, suggesting that additional people, beyond some limit, are more expensive to protect than the average. Workshop participants did not propose ways to estimate the public service costs of immigrants; however, they stressed that there is a need for a common conceptual basis in order to improve case studies and make comparable estimates.
Public service costs in the case studies were most frequently calculated as the product of the relevant undocumented population multiplied by per capita costs. Most of the case studies calculated per capita immigrant costs on the basis of average costs (dividing total program costs by the total number of recipients). Average costs assume that the per capita use of public services by illegal immigrants is the same as the use by the general population—an assumption that workshop participants noted has not yet been supported by empirical evidence. The assumption that average costs prevail for illegal immigrants is equivalent to taking a pro rata approach. A pro rata approach, participants observed, may distort the measurement of costs because it is unlikely that illegal immigrants use diverse public services in ways exactly like the native-born and legal immigrant populations.
A contrasting approach is to base estimates on marginal costs, where only incremental costs are attributed to each additional user. For quasi-public goods, marginal costs are likely less than average costs and may be quite low for small increments of additional population. For example, the marginal costs of one more car driving on an existing highway are typically relatively low—much lower than the average cost for constructing the road. However, like average costs, marginal costs attributed to the illegal immigrant population may also differ from the population at large.
A central weakness of the case studies is that they present estimates for costs and revenue sources, but neither the list of costs nor revenues is complete, making it impossible to assess the net fiscal impact of immigrants. More attention needs to be paid to the issues of net fiscal effects, which would involve a more complete accounting of relevant costs and revenues.
One general issue that receives scant attention in available case studies is the absence of statistical uncertainty calculations attached to point estimates and the lack of model sensitivity analysis associated with alternative assumptions. Both uncertainty and sensitivity measures are critical for statistical and policy interpretations. If illegal immigrant populations varied, how much difference would it make? Exploring alternative assumptions would also improve future case studies
by providing an assessment of the relative contribution of missing components. If property tax contributions from rental units are omitted, how much difference would it make if various proportions of rental property tax revenues were included? Overall assessments and bounds of uncertainty would benefit from studies that offered "high" and "low" assumptions for illegal immigration cost and revenues.
RESEARCH AND DATA NEEDS
After discussing the six studies of the fiscal impact of immigration on state and local governments, workshop participants briefly discussed three research and data priorities: improving data sources, expanding the range of analytical methods, and examining the need for better research frameworks.
Regarding data sources, it was noted that an earlier workshop offered a comprehensive overview of data improvements needed for immigration research (Edmonston, 1996). In response to some of the needs identified at the 1992 workshop, Robert Warren, then chief of INS's statistics division, described three new data sources on immigration that are now available or in development.1 First, the INS and the National Institutes of Health jointly requested that the Census Bureau add questions on nativity to the Current Population Survey. Questions on nativity were added beginning with the January 1994 Current Population Survey and will soon be available for analysis. Second, the Social Security Administration is in the process of adding place of birth to a 1 percent sample of its records. When that is completed, the INS will match the sample with data on immigration status to create a longitudinal database on the earnings of immigrants by immigration status. Third, the INS will soon initiate an annual survey of a cohort of immigrants; it will include questions on current occupation, internal migration, education level, and language ability. The survey will have a longitudinal component because the INS hopes to reinterview the immigrants in successive years, which will provide data on the adjustment of immigrants over time.
Regarding analytical methods, participants emphasized that the key questions about the fiscal impact of illegal immigration are largely hypothetical: What would the population of a particular local area be if there had been no illegal immigration in recent years? What would the current account balance for cities and states look like if there had been no illegal immigration? Workshop participants noted the limitation of using current account information to deal with the conditional situation of no illegal immigration because so many aspects of the economy would be altered.
Workshop participants suggested three areas for improvements in methods. First, case studies often neglect capital costs. It would be helpful for future
studies to consider capital costs, asking, for example, if illegal immigration is substantial enough to lead to capital costs for prison construction. Furthermore, future studies could make an attempt to define when it is appropriate to include capital costs. Second, in thinking through the appropriateness of using marginal or average costs, analysts might consider whether particular fiscal effects are short or long run. The effect of illegal immigration might be thought generally to have a marginal effect on costs. If illegal immigration is large and has long-term effects, however, the marginal costs may be close to the average costs. Third, it would be useful to see work on possible short-cut methods for assessing costs, rather than the approach of assessing costs of each service (for many localities for state studies) and then adding the components.
Regarding research frameworks, participants suggested that analysts might consider strategies that relate key factors together within an overall consistent framework. Workshop participants noted that a consistent framework would be useful to take into account potentially countervailing effects of illegal immigration on local employment (e.g., immigrants might displace some native workers but would also have multiplier effects on employment through their expenditures). It was pointed out that consistent accounting frameworks and a common conceptual basis on costs would be useful for making comparisons among fiscal case studies. New approaches would be helpful for offering alternative assessments. Would it be possible to design some comparative observational studies of similar communities with different immigration patterns to assess costs of services that can be attributable to general immigration levels and characteristics? If possible, such studies would obviate the need for direct information on actual use and service costs. In addition, analysts could consider comparing cross-sectional estimates of costs and revenues with life-cycle estimates. Cross-sectional estimates are heavily influenced by the age composition of the immigrant population. Since the immigrant population is young, data from any cross-sectional analysis would reveal disproportionately high school enrollments and disproportionately low levels of taxes paid, tending to make immigrants' fiscal costs exceed the revenues they generate. However, examined from a life-cycle perspective, as the younger immigrants age, it is possible that a different picture would emerge. Furthermore, expenditures on health and education in childhood influence the earnings and tax payments of workers in later years, which is not taken into account in cross-sectional calculations. But if, in contrast, there is a constant flow of illegal immigrants, the cross-sectional estimates may be accurate in the short run, but they would not be accurate for the lifetime costs of a particular cohort of immigrants. Thus, workshop participants suggested that using life-cycle estimates as a comparison to cross-sectional estimates might help to provide a clearer picture of costs. Moreover, life-cycle analysis would need to take into account substantial emigration among the foreign-born population. Because many immigrants subsequently return to their countries of origin, a sizable proportion of immigrants (illegal and legal) do not reach their elderly years as U.S. residents.
