In this study the impacts of the Immigration Act of 1990 are estimated. The "impacts" that are the focus of the study are numbers and various shares of legal U.S. immigrants compared to an estimate of what corresponding numbers and shares would have been in 1995 and 1996 if the "old" immigration law had remained in force. The study is concerned only with part of U.S. immigration, but a large part. Refugees are eliminated from the analysis along with certain groups of immigrants whose admittance was due to a "one-time" program that is inherently impossible to model. Thus, immediate relatives of U.S. citizens, family-based restricted immigrants, employment-based restricted immigrants, and certain special immigrants are studied. In every case, principals are distinguished from derivatives, and where necessary to address a specific issue, other distinctions are made in the data, such as age, occupation, country or region of birth, adjustments by nonimmigrant entry class, new immigrants versus adjustments of status, and specific immigrant entry class

Effective October 1, 1991, U.S. immigration law changed considerably. These changes were the result of the Immigration Act of 1990 which, while retaining the basic principles of the earlier legislation, provided the most comprehensive change in legal immigration since 1965. Data on which we base our forecasts relate to the old law. The new law has many similarities to the earlier law. For example, the new law, like the old, provides for the unrestricted immigration of certain immediate relatives of U.S. citizens. The new law places more emphasis on employment considerations, but the old occupation preferences (third and sixth) also place emphasis on the same types of considerations. Under the old law legal resident aliens were able to reunite with immediate members of their family under the second preference, and a similar preference (though expanded in terms of available visas) also is available under the new law. Moreover, the new law provides for "diversity" immigrants and explicitly excludes from this class any natives of countries that are oversubscribed.

Several changes made by the 1990 Act are of particular relevance. Under the previous law, the annual allocation of numerically restricted visas was 270,000. The 1990 Act established a "flexible" worldwide cap on family-based, employment-based, and diversity immigrant visas. Beginning in fiscal year 1995, after a "transition" period during which the annual quota was set at 700,000 the worldwide limit is 675,000. Separate ceilings are set for each of the immigrant categories: for family-sponsored visas, 480,000; for employment-based visas, 140,000; and for diversity visas, 55,000. While immediate relatives of U.S. citizens remain exempt from numerical limitation, the number of spouses, minor children and parents of U.S. citizens are subtracted from the overall numbers available for family sponsorship. However, under no circumstances can the number of numerically restricted family-sponsored visas be less than 226,000. Therefore, if the number of immediate relatives of U.S. citizens exceeds 254,000 (i.e., 480,000 - 226,000), the flexible worldwide cap of 675,000 may be pierced.

In addition to setting a higher overall limit on admissions, the 1990 Act altered the per-country limitations used to determine how many immigrants may enter the United States each year. Previously, the per-country quota was set at 20,000 visas per year. The 1990 Act provides that family-based and employment-based visas made available to citizens of a single independent foreign state may not exceed 7 percent of the total available. Given the minimum of 226,000 family-sponsored and 140 employment-based allocations, the per-country ceiling for an independent country is raised to 25,620. Additional flexibility is provided for potential migrants by the fact that the 7 percent per-country limit is not subdivided between family-sponsored and employment-based allocations.

In many respects the new law concerning family-based immigrants is similar to the previous law. Immediate relatives of U.S. citizens remain exempt from numerical limitation. Moreover, the annual floor of 226,000 numerically restricted family-sponsored visas represents only a relatively small increase over the 216,000 available within the previous law�s family-related preference categories. However, the new law makes certain provisions that should alter the mix of immigrants within the family-sponsored categories. The major change in the family-sponsored categories involves spouses, minor children and unmarried adult children of permanent resident aliens (i.e., second preference). For these immigrants, the new law increased the allotment from 70,200 to at least 114,200. Moreover, at least 77 percent of these visas are designated for spouses and minor children of permanent resident aliens and three-quarters of these are not subject to per-country limits. The other family-based categories (i.e., first, fourth, and fifth preference), either had their allocation remain essentially unchanged (i.e., fifth) or reduced by the new law.

While maintaining the strong orientation toward family reunification that characterized the 1965 Amendments, the new law accommodates more skill-based immigrants, and it provides for more source-country diversity. Prior to the 1990 Act, 54,000 visas were available for occupation-based immigrants. The new law allows up to 140,000 employment-based visas and also places more emphasis on skilled migrants within this category. The so called "diversity" immigration allocation was made available for the first time by the 1990 Act (although a relatively small number of diversity visas were allocated by means of a lottery before the new law took effect). The diversity immigrant allocations are designed to facilitate the entry of potential migrants from countries adversely affected by the 1965 law. Effective in 1995, the diversity quota is 55,000. These 55,000 visas are to be allocated to natives of a country that has sent fewer than 50,000 immigrants to the United States over the previous five years. No single country may receive more than 7 percent (3,850) of the number available worldwide. To be eligible for a diversity visa, a prospective immigrant must have at least a high school education or its equivalent and at least two years of work experience in an occupation that requires at least two years of training or experience. Diversity immigrants are therefore a kind of occupational immigrant.

Table 1 bridges immigration under the old law and immigration under the Immigration Act of 1990. Data in this table are averages for 1990 and 1991 (the last years of the old law), 1992-1994 (the three transition years of the new law), and 1995 and 1996 (the first two years during which the more or less permanent cap and quota numbers were in effect). Between 1990-1991 and 1995-1996, immigration subject to the numerical cap increased by 27.0 percent from 535,993 to 681,209. Family-based immigration increased by 16.9 percent to 528,551, whereas employment-based immigration increased by 72.3 percent to 101,418. Immigration not subject to the numerical cap fell by 7.7 percent, and in 1995-1996 accounted for 16.3 percent of total immigration compared to 21.1 percent during 1990-1991.

Whereas total immigration subject to the numerical cap increased somewhat due to the new law, many of the percentages noted above did not change markedly with the exception of the percentage increase in employment-based immigrants. Thus the Immigration Act of 1990 appears clearly to have had the effect of boosting employment-related immigration, which was one of the major objectives of the new law. This conclusion is further substantiated in the analysis provided below. However, professionals with an advanced degree (employment 2^nd preference principals) have declined steadily since the first year that the new law was in effect: 58,401 in 1992, 29,468 in 1993, 14,432 in 1994, 10,475 in 1995 and 8,870 in 1996. This pattern suggests a pent-up demand for entry by such individuals that was relieved the Immigration Act of 1990.

As noted above, the goal of this study is to determine the extent to which the Immigration Act of 1990 changed the number and composition of various types of legal U.S. immigrants relative to what they would have been if the former law remained in effect. The "old law" was operational from FY 1968 through FY 1991 and thus allows the development of a considerable history of legal entrants. Because data are available on each legal resident alien accepted by the United States for FY 1972 through FY 1991, we are able to generate 20 annual observations for any cross classification that is possible given the information reported in the Immigration and Naturalization Service�s Public Use Tapes. However, due to the different treatment of the Eastern and Western Hemispheres until 1976, the study�s historical background consists of data beginning with 1977.

Simple linear extrapolation is the methodology used in this study to predict what immigration of various types under the old immigration law would have been had it continued through 1995 and 1996. The form of forecasting equation is

IMM_it = b ₀ + b ₁T + e_{it, ,}

where IMM_it is immigration of type i in year t, T is a simple time trend (1, 2, . . .,15), and e_it is random errors. Each forecast refers to 1995 and 1996, which values are forecasted by carrying the time trend to 19 and 20 respectively. With the estimated constant term (or intercept) and the estimated coefficient on T, which indicates the direction and strength of the trend, each estimated relationship yields a "predicted value" that represents immigration under the old law if the 15-year trend continued for four (1995) and five (1996) more years. These values then are compared to a corresponding comparable value that occurred under the Immigration Act of 1990 (for 1995 and 1996). Comparable classes of immigrants, or the mapping from the old law to the new law, are reported in Appendix A. The ratio of "actual" to "predicted" yields an estimate of the percentage difference that exists for the new law compared to the old law.

For those immigrant classes subject to a ceiling under the old law, various shares were forecasted (e.g., principals versus derivatives). The ceilings were assumed to have been met, so that the issue became not how many but rather who was admitted under the ceiling. Even when predictions are developed for aggregate numbers, the ceilings are recognized in the components that make up the aggregate prediction.

All forecasting regressions and all corresponding forecasts distinguish "principals" and "derivatives." Principals enter the United States by dint of their relationship with a U.S. citizen or resident alien, or by dint of their skills, whereas derivatives enter as the spouses and/or minor children of the principals.

Estimating linear trends is not without shortcomings. First, in certain instances nonlinear relationships may provide better fits than linear relationships. Second, in other cases step functions may fit the historical data better, and by imposing a linear trend on the data such a pattern could impose a strong trend to a relationship that would better be carried forward at a constant level. Third, certain categories of immigration occurred under a ceiling. Nevertheless, trends may be apparent under the ceiling as levels approached the maximum. In such cases ceiling levels are preferable to extrapolating the trend. Finally, some behavioral model may underlie the historical data. The problem with using more complex forecasting models is that forecast values of the independent variables of the model must be used to generate out-of-sample forecasts and these may be unavailable.

The advantage of using a similar linear trend is its simplicity. The trend line is easily extrapolated out of sample and the technique can be applied to numerous 15-year series required for this report. However, the assumptions that underlie the approach should be kept in mind.

Appendix A contains considerable detail on the methodology used here, as well as the mapping of immigrant classes from the old law to the new law. In what follows, we begin each section with a specific question regarding immigration under the new law relative to the old law.

What is the difference between actual and predicted total immigration for 1995 and 1996? The Immigration Act of 1990 appears to have slightly increased total immigration for 1995 (by 4.4%) and to have greatly increased that for 1996 (by 35.2%). The respective ratios of actual to predicted are 1.044 and 1.352. In Table 2 and in subsequent tables, a ratio of 1.0 indicates that actual equals predicted, whereas values less than 1.0 indicate that the new law yielded immigration of type i that was less than predicted under the old law. Values greater than 1.0 indicate that the new law increased immigration compared to the numbers expected under the old law. Table 2 also indicates that whereas the actual numbers are greater than those predicted for both principals (1.4% and 33.1% for 1995 and 1996, respectively) and derivatives (14.2% and 42.3%, respectively), the derivatives rose relatively more than the principals with the consequence that the share of total immigration accounted for by derivatives increased compared to that predicted under the old law (by 9.4% in 1995 and 5.2% in 1996). (Appendix A provides further detail on the methodology used to generate predicted values.)

In Table 3, the above question is addressed by occupation. This table shows that professional, technical, and kindred (or PTK) immigration was increased considerably by the Immigration Act of 1990. For both principals and derivatives each component of the PTK group (i.e., other professional, health professional, and technical specialty) increased relative to predicted values. The same conclusion holds for the managerial and executive group. For principals, other occupational groups are generally down (except operators, fabricators, and labors � up 41.8% in 1996; farming, forestry, and fishing � up 5.2% in 1996; and services � up 2.9% in 1996). All occupational categories of derivatives increased relative to predicated values (except farming, forestry, and fishing, which is down each year).

Table 3 has two components. Table 3A provides shares of all immigrants who report an occupation. Thus, for each year values in the columns for predicted and actual shares sum to 100% for principals and derivatives taken as a whole. Shares in Table 3B sum to 100% for principals and to 100% for derivatives. In each table the shares refer only to those immigrants who report an occupation. For principals, each table indicates a strong increase in the shares for the PTK groups, as well as for managers and executives. Thus, the Immigration Act of 1990 not only increased the numbers of highly educated and skilled individuals relative to what they would have been under the old law, but also it tilted the overall composition of U.S. immigration toward the more highly skilled groups. Moreover, for derivatives, relatively large increases occurred in both the numbers of highly skilled immigrants and the percentages of such individuals compared to what was expected under the old immigration law.

Table 4 is similar to Table 3, except that Table 4 breaks out two age groups that correspond to working ages (20-34, 35-64). Principals among the PTK immigrants tend to be concentrated in the younger ages, whereas principals among managers and executives tend to in the older ages. Among derivatives, the most highly educated and skilled individuals tend to be concentrated in the older age categories.

The next series of questions relates to employment-based immigration. Under the old law the employment preferences (often called "occupational preferences") were the third (P3) and sixth (P6) preferences. These are the preference categories used for forecasting purposes.

To what extent have EB1 versus EB2 versus EB3 visa issuances increased the skill composition of U.S. immigration? These categories refer to aliens with extraordinary ability, outstanding professors or researchers, and multinational executives and managers (EB1 or E1); professionals holding advanced degrees (EB2 or E2); and skilled workers and professionals as well as needed unskilled workers (EB3 or E3). Table 5 reports four groupings (or mappings) of the data relating to the old and new laws: (1) E1/P3; (2) E3/P6; (3) (E1 + E2)/P3; and (4) (E1 + E2 + E3)/(P3 + P6). These groupings allow comparisons between the highest skills (E1 and P3) and skills in short supply (E3 and P6), as well as other comparisons that are more general, but remain within the employment preferences. The preference categories that require the very highest skills and/or education (E1and P3) suggest that for both 1995 and 1996 both principals and derivatives (who came with principals admitted under the employment preferences) were down considerably relative to the predictions. The shortfall of principals in 1995 was 0.546 of the predicted number and in 1996 was 0.892 in employment-based preference EB1 compared to what would have been expected under old preference category P3. (The respective shortfall in derivatives was 0.724 in 1995, but in 1996 an increase occurred of 12.7%).

For skills in short supply (E3 and P6), the numbers are up considerably. For principals EB3 admittances in 1995 were 130.5% higher than anticipated and in 1996 were 189.4% higher. When the mappings are expanded to include all employment preferences (E1 + E2 + E3 relative to P3 + P6), principals also were up sharply for both 1995 and 1996 (55.0% and 117.2%, respectively). Derivatives too were up, but by correspondingly smaller percentages (37.3% for 1995 and 90.5% for 1996). In Table 5, the figures for derivatives indicate nothing about the skills of derivatives because this group of immigrants does not need to satisfy any skill requirement (and indeed most of them report this occupation as our category 9, which is "homemakers, unknown, retired, etc.").

Table 5 also reports EB4 admissions, which under the old law were numerically exempt special immigrants. Under the new law they are also special immigrants, including ministers and others. These immigrants were forecasted by means of a simple time trend (unconstrained), which yields predicted 1995 and 1996 values for them. Under the Immigration Act of 1990 these immigrants are up in numbers for both principals (63.8% in 1995 and 90.7% in 1996) and derivatives (3.2% in 1995 and 13.5% in 1996).

Are the derivatives of the new employment-based categories more skilled than those of the old occupational preferences? This question is addressed by examining occupation-specific forecasts based on the sum of P3 and P6 with occupation-specific actuals based on the sum of E1, E2, and E3. Table 6 corresponds to Table 5, except that in Table 6 occupations are distinguished and EB1, EB2, and EB3 taken as a whole relative to the forecast for P3 and P6 is shown. For principals, as expected, the professional, technical, and kindred (PTK) as well as the managerial and executive groups show major increases. Health professionals stand out as the group with the largest increases for both 1995 and 1996. The answer to this question appears to be a clear "yes." Among derivatives, in 1996 health professionals (numbering 916) were 535.6% over the predicted number. Other professionals (1,375 in 1996) were 305.9% greater than predicted, and technical specialty workers (1,711 in 1996) were 145.1% greater than predicted. Most derivatives fall in the "homemaker, unknown, retired, etc." class, but among those who report an occupation the PTK groups, especially health professionals, are up considerably. However, the numbers in these groups are understandably small compared to their counterparts among the principals.

Has "dual intent" on adjustments from nonimmigrant entry classes (in particular from H1, H1B, and L) increased skill composition? Table 7 addresses this questions. Table 7A reports principals and Table 7B shows derivatives of these principals. Principals are further divided into new entrants and those who adjust their status, with those who adjust distinguished as H1B, L1, and other adjuster. Among nonimmigrant entry classes, H1B refers to specialty occupations (and under the old law referred to admissions of "distinguished merit or ability"), and L1 refers to intracompany transferees. In addition to H1B, the H1 category includes registered nurses (H1A). Regarding H1B adjusters, the answer to the question is that professional groups, as well as managers and executives, definitely are up compared to what is predicted under the old law. For L1s, managers and executives are up, but PTK workers are not heavily represented in this group. Other adjusters are in the various high-skill occupations, but they are up in most other occupations as well. With the exception of the "other professional" category, new immigrants are lower than expected or are only slightly higher (in 1996 only). Although their numbers are not great, derivative adjusters increased skilled immigration, but less-skilled immigration also. Derivatives of new immigrants clearly increased skills.

How did the composition of source countries change due to the fact that the Immigration Act of 1990 did not subject three quarters of FB2As to per country limits? Under the Immigration Act of 1990, the former second preference became the second family based preference class (FB2). This preference class is for spouses and children of legal permanent residents. The Act further distinguishes between FB2A (spouses and children under 21 years of age) and FB2B (unmarried children 21 years of age and older). Table 8 reports information relating to this questions. Mexico received the major benefit from the three-quarter exclusion from country limits. In 1995, Mexico alone had 51,502 FB2A admissions (relative to a predicted number of 3,740). In 1996, actual FB2A admissions from Mexico were 86,390 compared to a predicted number of 3,424. The Philippines also benefitted from FB2As, but benefitted more from FB2Bs. In short, source-country composition was strongly tilted toward Mexico and to a lesser extent toward the Philippines due to the exclusions from country limits.

Did the Immigration Act of 1990 increase the overall numbers of family-based immigration? Table 9 distinguishes principals and derivatives both for family-based numerically restricted immigration and for numerically exempt immediate relatives of U.S. citizens. The family-based restricted under the old law (Preference Classes P1, P2, P4, P5) entered under a numerical limitation that was in the late 1980s and early 1990s about 216,000 per year. Thus, for this group we predicted shares of principals (derivatives) under an annual limit of 216,000. However, immediate relatives of U.S. citizens were (and are) exempt from quota limits. Thus, immediate relatives were predicted by means of an unconstrained linear regression.

Family-based restricted principals increased considerably relative to predicted numbers (29.1% in 1995 and 68.6% in 1996). Immediate relatives of U.S. citizens were below the predicted number in 1995 (0.775), but slightly above in 1996 (1.024). In terms of shares, during 1995 and 1996 numerically restricted principals increased by 40.2% and 44.0%, respectively, relative to predicted shares. Family-based restricted derivatives fell relative to predicted numbers (0.875 in 1995 and 0.965 in 1996). Do the above numbers differ by restricted/exempt? From the above discussion, we can see that the answer is "yes" for numerically restricted immigrants (and strongly so), but is "no" for numerically exempt immigrants. Overall, the Immigration Act of 1990 increased immigrant numbers.

The final block of questions relates to diversity issues.

Did the Immigration Act of 1990 result in more diversity by country of birth? Tables 10, 11 and 12 relate to this question. Table 10 shows principals (9A) and derivatives (9B) by class of immigrant and country or region of birth. Country/region predictions are based on forecasts of shares (constrained to sum to 1.0) for those groups that enter under a restriction. The number of restricted employment-based principals in 1995 and 1996 exceeds predictions for most countries/regions, but especially for China, which for each year had over 10 times the predicted number. The Philippines, Canada, India, Africa, and Europe all had 1995 and 1996 numbers well in excess of their predicted numbers.

Family-based restricted principals exceeds predictions especially for Mexico, but for the Philippines as well, which were due primarily to the FB2A visas discussed above. For principals, exempt immediate relatives in 1995 were below the predicted numbers, but in 1996 were above for many countries/regions.

The most notable increase in employment-based derivatives was for China, but Canada and India also had relatively large increases. The largest relative increase (compared to predicted numbers) in restricted family-based numbers was for Africa , but the total numbers for Africa are small in comparison to other regions/countries.

The answer to the question immediately above depends upon perspective. The major source of U.S. immigrants is Mexico, and Mexico had by far the largest increases in 1995 and 1996 relative to predicted numbers. These increases were due in part to FB2A category of admissions discussed above. For example, restricted family-based principals from Mexico accounted in 1995 for 10.0% of admissions listed in Table 10 and in 1996 for 11.9%. However, more immigrants are originating in most other parts of the world and in this sense immigration is becoming more diverse.

What is the skill composition of the new permanent diversity program? Table 11 sheds more light on the diversity questions by breaking out countries that were not qualified for "diversity visas" from those that were. (The set of diversity-qualified countries may change over time, but we took Mexico, Philippines, and India as not diversity qualified.) Table 11 reports simple cross tabs for diversity immigrants versus non-diversity immigrants from diversity-qualified countries and for immigrants from non-diversity qualified countries. In each case principals are distinguished from derivatives and occupational detail is provided. For diversity-qualified countries, diversity immigrants accounted for higher shares of the PTK groups and for smaller shares of the less skilled occupational groups. This statement holds for principals and derivatives. Thus, the qualifications for the diversity program appear to have increased skills. The non-diversity qualified countries tend to have greater shares of immigrants in less-skilled occupations than diversity-qualified countries, but health professionals is an exception.

Table 12 provides overall predictions for the various occupational groups for diversity-qualified and non diversity-qualified countries and for principals versus derivatives. For principals from diversity-qualified countries, the diversity programs appear to have increased skills. The high-skilled occupational groups all tend to have actual values in excess of predicted values for diversity-qualified countries. So do the derivatives from these countries, but for them all occupational groups tend to have risen relative to predicted numbers (although the increases are relatively largest for those with the highest skills). The permanent diversity programs thus appear to have increased skilled immigration compared to what it would have been under the old immigration law.

The INS Public Use data tapes for the years 1977-1996 served as the source of the underlying data for this study. (Data for the years 1972-1976 were not employed because the Western Hemisphere was treated differently than the Eastern Hemisphere in these years.) These raw data were aggregated into a processed data set whose data dictionary is presented below. The aggregated data and, in some cases, further aggregations thereof were used to produce the number used in the regressions and cross-tabs presented in this study. This appendix describes this process including details such as how the structured missing data present in the data for FY80-FY83 were processed. (The INS Public Use tapes contain flaws for the years 1980-1983 that require attention.)

The data were aggregated as described in the following outline. The outermost level of the outline describes the variables appearing in the processed data set (with difference between the "old" and "new" data indicated.) The deepest level of the outline indicates the actual values the specific variable may take in the processed data set. Where necessary, additional aggregation was performed to recover the inner levels.

1. Class of Admission [10,12]

   1. Employment Based [2,4]

      1. Pre-92 categories [2]

         1. third preference (P3 )

         2. sixth preference (P6 )

      2. Post-91 categories [4]

         1. EB1 (E1 )

         2. EB2 (E2 )

         3. EB3 [2]

            1. skilled (E3A)

            2. unskilled (E3B)

         4. EB4 [2]

            1. A portion of EB4 which corresponds to certain formerly exempt special classes. Therefore this category is defined both Pre-92 and Post-91 (E4A).

            2. The remainder of the new EB4 category (E4B).

         5. EB5 (E5 )

   2. Family Based [6]

      1. Unmarried Sons and Daughters of Citizens (F1 ) P1/FB1

      2. Spouses and unmarried sons and daughters of Permanent Residents. Note: Children were not distinguished from Unmarried Sons and Daughters 21 years and older prior to 1992. In these years the F2A/F2B split will be accomplished for principals based on their age and derivative immigrants are assumed to be in F2B.

         1. Spouses and Children (F2A) P2A/FB2A.

         2. Unmarried Sons and Daughters 21 years or older (F2B) P2B/FB2B

      3. Married Sons/Daughters of citizens (F3 ) P4/FB3

      4. Brothers/Sisters of Citizens 21 years of age or older (F4 ) P5/FB4

   3. Exempt Immediate Relatives (IR )

   4. Diversity (DIV)

2. Diversity - Is country of birth a 1996 qualified for diversity country? [2]

   1. Yes (Y)

   2. No (N)

3. Country of Birth [10]

   1. North America (2)

      1. Mexico (MEX)

      2. Canada (CAN)

   2. Asia and Oceania (4)

      1. Philippines (PHL)

      2. India (IND)

      3. China (CHN)

      4. Remainder (ASO)

   3. Europe (EUR)

   4. Africa (AFR)

   5. South America (SAM)

   6. Caribbean and Central America (CCA)

4. Occupation [12]

   1. Professional Technical and Kindred [3]

      1. Health Professionals (H)- contains DOC, HLD, HLT, and NUR.

      2. Other Professionals (P)- contains ARC, ENG, MCS, NSC, and SSC.

      3. Technical Specialty (T) - contains ART, COU, LAW, LIB, SWK, TCO, TCU, TNH, and TNO.

   2. Managerial, Executive (2) - contains EXC.

   3. Sales (3) -contains SLS

   4. Administrative support (4) - contains ASP.

   5. Precision Production (5) - contains PCR.

   6. Operators, fabricators, and laborers (6) - contains LAB.

   7. Farming, forestry, and fishing (7) - contains FFF.

   8. Service (8) - contains SER.

   9. Homemakers, unknown, retired, etc. (9) - contains HOU, NOT, UNR, and STC.

   10. Structured Missing (U)

5. Age [4]

   1. (A1) - contains 0-19

   2. (A2) - contains 20-34

   3. (A3) - contains 35-64

   4. (A4) - contains 65+

6. Adjuster Detail [5]

   1. New (NEW)

   2. Adjuster [4]

      1. H1 [2]

         1. H1A- registered nurse as indicated by admission class RN6 or RN7 (H1A).

         2. H1B- other H1 category (H1B).

      2. L1 (L1 )

      3. Others (OTH)

   3. Structured Missing (UUU)

7. Principal/Derivative [2]

   1. Principal (P)

   2. Derivative (D)

8. Gender [2]

   1. Male (M)

   2. Female (F)

   3. Missing (X)

   4. Structured Missing (U)

9. Fiscal Year

To produce the processed data set, we processed the files into an intermediate format that adds two pieces of information to the above: a registered nurse (RN6/RN7) admission class and an age indicator. These fields are used to produce the H1A/H1B and, in pre-92 data, the F2A/F2B distinctions as described above. The additional fields are then removed. Some important notes regarding the H1A/H1B and F2A/F2B split appear later in this appendix.

This section describes the technique used to distribute the structured missing data in the occupation, sex, and adjuster detail variables caused by structured missing data for fiscal years 1980 through 1983. Because of concerns about data strength at the fine level of detail provided by this module, a straightforward technique of distributing the structured missing data has been chosen. To ensure that the data set remains consistent when examined in less aggregated views, we have applied the proportions determined at an aggregate level to the highest level of detail provided.

Consider a module which contains two variables affected by the structured missing data: sex and intended state of residence. Sex and intended state of residence are not relevant in the current study, but they serve to illustrate how the data were treated. The distribution procedure is described graphically in Figure 1. The structured missing data are indicated by the number in the intersection of the structured missing values for sex (�U�) and state of intended residence (�UU�). These data are to be distributed into the desired categories for the sex and state. The structured missing data are distributed into the shaded region of Figure 1, which includes all sex/state pairs. On a fiscal-year basis the proportion of missing data that should be allocated to each sex/state pair is determined by dividing the total number of records for that tuple prior to distribution by the total number for all destination pairs to distribution. The data are then distributed at the finest level of detail provided for the module as follows. For each combination of the variables unaffected by the missing information, a table analogous to the one in Figure 1 is formed. Within that table, the structured missing is distributed to the destination pairs in the shaded area according to the proportions determined at the fiscal year level in Figure 1.

Performing the procedure in this manner ensures that the data for each fiscal year are consistent with the overall proportions in the data unaffected by the structured missing data while providing data at the level of detail required by this study. As a side effect, the data after distribution for the years FY80-FY83 include numbers that have a fractional part. This is necessary to ensure that totals after distribution agree with those prior to distribution.

As a check on this procedure, one should observe no change in the category totals for variables other than those affected by this problem. The relative proportions of each category (disregarding the structured missing category) should be the same before and after distribution.

For the processed data sets, the sex, occupation, and adjuster detail variables are those affected by the structured missing data problem.

The nonimmigrant class of entry field in the INS Public Use data records the most recent entry class for immigrants adjusting their status. The entry class H1 indicates a "Temporary worker of distinguished merit and ability." Beginning in 1992, the documentation indicates that the class H1 is divided into H1A for registered nurses and H1B for "temporary worker with �specialty occupation�."

The manner in which the post-92 split is recorded in the Public Use data sets is not clear. For 1992 and 1993, the indicated codes are H1A and H1B. Since the field is has only two characters, there is no distinction. In 1994 and later, the documentation indicates that H1A is coded as S8 while H1B is coded as H1. In these years where the distinction is clearly made in the documentation, the number of immigrants coded as H1A is negligible.

The admission class codes corresponding to H1A are RN6 and RN7 which appear in years 1990 and later. To use this information, the admission codes RN6/RN7 are mapped to a separate value of the admission class variable (namely RNE). The adjuster detail is recorded as H1 rather than making the H1A/H1B distinction. In a post processing step, the immigrants whose adjuster detail is H1 are assigned to H1A if their admission class is RNE and H1B otherwise.

Under IMMACT90 the former second preference class became the second family based preference class: FB2. This preference class is for spouses and children of legal permanent residents. IMMACT90 further distinguishes between FB2A containing spouses and children under the age of 21 and FB2B containing unmarried children 21 years of age and older.

The admission class code appearing in the Public Use Data for the second preference class does not distinguish between children under 21 and those 21 and over. To facilitate comparison between the pre-92 and post-92 data, the following procedure was used. For immigrants whose admission class was that of a child of a legal permanent resident, the age field was examined. Those under 21 were assigned to FB2A while those 21 or older were assigned the FB2B. For immigrants whose admission class was that of a derivative of a child of a legal permanent resident, there was no way of selecting FB2A or FB2B. All such derivatives were assigned to FB2B. In effect, the FB2A/FB2B split was not made for derivatives and this fact is reflected in the analysis.

The estimates are based on regressions from fiscal years 1977 through 1991. Simple numerical regressions are appropriate for the exempt categories, but would result in estimates for 1995 and 1996 which exceed certain caps in the old law such as the 27,000 limit for the third (P3) and sixth (P6) preference classes and the 80% limit on restricted family based (FB) immigration. In an effort to reflect the anticipated effects of these caps, the estimates were produced from four separate regressions: P3, P6, FB, and exempt.

The division of a group�s immigrants into the P3, P6, FB and exempt categories is based on admission class. P3 and P6 each correspond directly to a value of the admission class variable. FB is the aggregation of the F1, F2A, F2B, F3, and F4 values of the admission class variable. The exempt category is the aggregation of IR, E4A, and RNE (recall that E4A was numerically exempt under the old law.) The remaining categories either do not appear in the historical portion of the data (the DIV and remaining "E" classes), or are excluded from the analysis (the REF and "X" classes.)

For the P3, P6, and FB categories regressions are performed on the share of P3, P6, and FB immigration associated with the group being estimated. For the exempt category a numerical regression is calculated. The raw estimated shares of P3, P6, and FB and the number of exempt immigrants are then estimated for 1995 and 1996. Any negative raw shares or numbers are set to zero. Then the shares are scaled to sum to 1. The estimate for a given group is then the sum of the its exempt estimate, 27,000 times its estimated P3 share, 27,000 times its estimate P6 share, and 216,000 times its estimated FB share.

This section describes the mapping used to aggregate the INS Public Use data into the process data sets. The utility used to perform the aggregation uses configuration files which describe how the data appearing in the Public Use Data are to be mapped into the desired variables. The appropriate portions of these configuration files are reproduced below.

Note: The admission class variables assigns all possible admission class codes to some category. In addition to the categories described in the data dictionary and the RNE class used for H1A/H1B, there is a refugee class REF and several special classes indicated by the first letter X. The REF and "X" classes are not included in any of the tables. The F2X class lists admission categories which are F2B after FY91 but are ambiguous prior to FY92. The previous section describes how this category is used.

Note: The special case of 20 is used as part of the FB2A/B split described in the previous section. The 20 age group is joined with the A2 age group in the post-processing step.

Note: In both cases, new immigrants are indicated by spaces in the field. The vertical bar character informs the processing software that the spaces are significant rather than just white space.