[Main Page of Report | Contents of Report]
The Los Angeles-New York City Immigrant Survey (LANYCIS) is a large survey that was designed to be representative of members of immigrant (that is, foreign-born) households in Los Angeles County and New York City. It was conducted from August 1999 to July 2000, with the bulk of interviews done between November 1999 and May 2000. This appendix provides brief documentation of the survey methodology. The survey was designed by staff of the Urban Institute and conducted by staff of the Survey Research Center, which is part of the Institute of Social Science Research at the University of California at Los Angeles. The survey was conducted in five languages: English, Spanish, Russian, Vietnamese and Mandarin Chinese.
The households that were eligible for the survey were those in Los Angeles County or New York City in which there was a foreign-born adult (someone 18 or older). We call these "immigrant households." and the groups of related individuals within them are "immigrant families." All respondents were adult members of the household, who also reported about other members of their households. Typically, the respondent was a foreign-born woman, such as the mother or wife in a family, but a substantial share of respondents were men.
Within each household, we sampled families and "focal people." All families with immigrant adults were eligible for sampling within the household. (Thus, if there were two families in a household and one was native citizen, that family would not be eligible. Similarly, unrelated people who were native-born were not eligible.) In most cases, there was just one family per household, however.
Within each immigrant family, we selected "focal people" about whom more detailed questions were asked. The focal people were:
In addition, we sampled a second respondent (R2) who was another foreign-born adult in the household who did not have children.
In general, LANYCIS is representative of all individuals in immigrant households living in Los Angeles County or New York city, with limited exceptions. For example, it does not cover native citizens who are living in immigrant households but are not part of their families (e.g., native citizen roommates or boarders). Nor does it cover families in which the parents are native citizens but the children are immigrants (e.g., families that have adopted a child from outside the United States).
LANYCIS had a complex, stratified sample. Because of the interest in food stamp recipients and low-income families, we oversampled households that met those criteria. In general, our goal was to obtain roughly equal sample sizes in both cities. In each city, it was hoped that about half the unweighted sample of households had received food stamps in 1996 or 1997, about one-quarter had incomes below 200 percent of the poverty level but did not get food stamps in 1996 or 1997, and one-quarter had incomes over 200 percent of the poverty level.
The survey used an amalgam of three sampling approaches.
1. The main sampling approach was a random digit dial (RDD) telephone survey of residences in Los Angeles County and New York City.
2. The second method used a list sample of addresses of food stamp recipient households, based on administrative data of 1997 participants provided by the local welfare agencies. Where possible, the appropriate households were telephoned. When there was no telephone number (or where the listed phone number did not work), interviewers were sent to the listed address for an in-person interview (or to get a phone number for a telephone interview). In this component of the sample, we did not require that the household interviewed be the same one listed on the administrative list, but any eligible household at that address. That is, the administrative lists can be viewed as a list of addresses with a high probability of being immigrant food stamp households.
3. There was originally a small area sample for in-person interviews of households without telephones. This yielded very few respondent households, however, so we terminated this approach early. To help represent non-telephone households, we used a Keeter adjustment that adjusts weights on the assumption that people with periodic phone service interruptions are similar to non-telephone households (Keeter 1995; Brick et al. 1999).
All interviews were conducted by computer-assisted methods, using either computer-assisted telephone interviews (CATI) fielded from UCLA's telephone facility or using computer-assisted personal interviews (CAPI) by professional interviewers based in Los Angeles or New York City using laptop computers. Most interviews were conducted by telephone, but some were done with in-person interviews, particularly those drawn from the food stamp administrative lists.
All households were informed of the purpose of the interview and of their right to refuse to participate or to not answer any question and they were promised confidentiality, following procedures approved by the Urban Institute's Institutional Review Board. A limited number of respondents, particularly those contacted in person, were offered incentive payments of $10 to participate in the study.
The unweighted sample sizes for the survey were: (52)
Category |
Los Angeles | New York City | Total |
|---|---|---|---|
Number of households |
1,893 | 1,554 | 3,347 |
Number of focal people: |
|||
Total focal |
4,750 | 3,096 | 7,846 |
Respondents |
1,893 | 1,554 | 3,447 |
Spouses |
1,066 | 654 | 1,720 |
Children 0-5 |
634 | 275 | 909 |
Children 6-17 |
968 | 469 | 1,437 |
Elderly* |
84 | 98 | 182 |
Second respondents |
105 | 46 | 151 |
Total number of people, including non-focal |
7,031 | 4,201 | 11,232 |
* There are also some people 65 or older who are classified as respondents or spouses. |
|||
Although the total sample size of people was substantially larger in Los Angeles, the number of households in each city was relatively close. Immigrant households in Los Angeles tend to have more members than those in New York, so the total unweighted sample size differed. As noted below, sample weights were used to scale the relative differences in the two cities.
The overall response rate for LANYCIS was 69 percent, although the rate was
lower in New York City than Los Angeles. (53)
Although there are few comparable surveys, we felt that the response rate
was reasonable, given the expected difficulties of contacting and interviewing
immigrants (particularly low-income immigrants) in urban areas. The response
rate is based on responses of the primary respondent; the rate for the second
respondent (R2) was substantially lower (34 percent). After a lengthy interview
with one person in a household, it was hard to secure an interview with another
household member. The table below summarizes various components of the survey
and their respective response rates for the primary respondent.
Response rates for primary respondent |
Los Angeles Co. | New York City | Both Cities |
|---|---|---|---|
RDD |
70% | 65% | |
In-person, from food stamp list |
83% | 73% | |
Telephone, from food stamp list |
77% | 63% | |
Area |
97% | 45% | |
Overall |
72% | 65% | 69% |
To compensate for stratification in the sample design and for non-response, weights were developed for LANYCIS. There are alternative sets of weights, depending on whether the unit of analysis is the household, family, or focal person. The weights were designed to account for the sampling approach used for each responding household and the probability of selection of that household. To ensure that the data correspond--at least in basic parameters--with Census data, we added post-stratification adjustments to the weights based on Current Population Survey data for Los Angeles County and New York City, averaging three years of the March 1997, 1998, and 1999 samples. The LANYCIS weights were designed to bring population totals for each city into correspondence with Census data when considering the following factors: poverty level and composition of families (i.e., those with and without and elders), and gender, age, schooling and country of origin for family members.
[Go To Contents]
In any survey, a certain amount of data are missing, underreported, or misreported, which can be at least partially corrected through data editing and imputation. To help adjust for these problems, survey researchers typically edit (or clean) their data and also impute some data to compensate for missing data. In LANYCIS, we focused on editing and imputation in three areas: income, health insurance, and immigration status.
One common problem in surveys is that respondents know that someone in the household has a certain type of income or health insurance, but might not know how much income or what type of insurance. Sometimes they are just unsure whether there is any type of income or insurance at all. For LANYCIS, John Coder of Sentier Research imputed income and insurance status of individuals using "hot deck" imputation methods, which are also used in surveys like the Current Population Survey (CPS) and National Survey of America's Families (NSAF). These essentially impute values for missing responses using responses sampled from similar types of people. Imputation may lead to erroneous (or questionable) values assigned for individual cases, but it should generally improve the distribution of cases for the overall sample.
We inquired about 1998 annual income by source (e.g., earnings, self-employment income, pensions, child support, interest or dividends, rent, social security income, welfare income, etc.) and by household member. After imputations, we computed each person's total income and created measures of family and household income. These were then compared with federal poverty guidelines (the Department of Health and Human Services guidelines used for program eligibility) to create income poverty ratios.
For insurance status of each focal person, LANYCIS asked whether insurance was employer-sponsored, other private (nongroup), Medicare, Medicaid, other state, or other insurance. As in NSAF, for people with no source of insurance we asked a follow-up question to confirm that the person was uninsured (Rajan, Zuckerman, and Brennan 2000). Anyone who reported receiving TANF or SSI was assigned Medicaid coverage. People could have multiple types of insurance coverage. We also created hierarchical measures of insurance coverage for each individual, in which each person was assigned one type of coverage: Medicaid is ranked at the top of the hierarchy, followed by job-based insurance, other private coverage and then other public coverage (usually either Medicare of State Children's Health Insurance Program).
Immigration status was sometimes imputed using logical editing processes. For instance, we generally used the status of the respondent and spouse to impute the status of their children. In the event that parents had discordant immigration statuses, we assigned the status of the parent who entered the United States at the same time or from the same country as the child. As is standard in surveys of this type, we did not directly ask whether people were undocumented but instead assigned undocumented status to people who do not otherwise report a legal basis for being in the United States. We edited some responses to modify impossible (or very unlikely) status codes, and a number of people were assigned undocumented alien status. In many cases, we used information from the in-depth qualitative interviews to further verify status.
The survey asked respondents and their spouses:
We define immigrants as foreign-born persons permanently residing in the United States. Foreign-born persons not permanently residing in the U.S. (i.e., non-resident aliens) include students, tourists, and temporary workers. Non-resident aliens are excluded from most analyses in this report, unless their spouses or unmarried partners are immigrants permanently residing in the country. Thus, the study includes primarily adults with four types of immigration status: undocumented, legal immigrant (LPR), refugee and naturalized.
Undocumented immigrants are persons who entered the United States without inspection, overstayed temporary visas, or otherwise violated U.S. immigration laws but remain in the country. In some cases, respondents answered that they do not have documents allowing them to remain in the country legally. In other cases, they answered that they have some type of temporary non-resident document (tourist visa, student visa, temporary work permit, or other document). We used a series of steps to impute undocumented status to some temporary document holders (rather than treat them as non-resident aliens), given that their documents were likely invalid or expired, and they were continuing to reside in the country:
1. All tourist visa holders who last entered the United States more than two years before the survey, assuming that their tourist visas had expired by the survey date.
2. All students not enrolled in school or working 20 hours or more per week, since student visa holders are not allowed to work more than 20 hours.
3. All temporary work permit and other visa holders who last entered the United States more than five years before the survey.
4. All temporary work permit holders not working in occupations for which work permits are valid.
Assignment of legal immigrant status was based on respondents' statements that they possess resident alien or "Green Cards." At the time of the survey, some legal immigrants had already applied for citizenship. But if they had not yet been interviewed and sworn in, we considered them legal immigrants.
Legal immigrants admitted to the United States as refugees are grouped with those who were still refugees at the time of the survey. Immigrants who have been granted asylum are included in the refugee group. Asylum applicants who have not yet been granted that status, however, are included in the undocumented group.
Naturalized citizens are foreign-born persons who have been sworn in as U.S. citizens, regardless of their entry status. Refugees who naturalize are included in this group, since their eligibility for benefits is more generous as U.S. citizens than as refugees.
Family immigration status is based on the combined statuses of respondent and spouse or unmarried partner. To begin with, all immigrant families included in the analyses have at least one immigrant family member. We categorize family immigration status in a hierarchy of increasing public benefit eligibility, with undocumented status at the top, as follows:
1. Any family where either the respondent or spouse/unmarried partner is undocumented is considered undocumented.
2. Any family without an undocumented respondent or spouse, but where either the respondent or spouse is a legal immigrant, is a legal immigrant family.
3. Any family with neither an undocumented nor a legal immigrant respondent or spouse, but where either the respondent or spouse is a refugee is a refugee family.
4. Any family where the respondent and spouse are both naturalized, or where a single respondent is naturalized, is considered naturalized.
Families where both the respondent and spouse (or a single respondent) is a temporary non-resident alien or a U.S.-born citizen, are excluded from the analyses in Part II. This exclusion led us to drop 85 of 3,448 families from the data. For analyses involving all families, the sample size is 3,363. Sample sizes are smaller for subsets of the data (e.g., non-elderly, low-income and/or food-insecure families), or when some variables have missing values.
[Go To Contents]
In this paper we use the U.S. Department of Agriculture's "Standard 6-Item Indicator Set for Classifying Households by Food-Security-Status Level." USDA and its contractors analyzed answers to 59 different food hardship questions included in the April food security supplement to the monthly CPS. USDA used the Rasch measurement model to validate a scale, ranging from "food secure" to "severe hunger," based on answers to 18 of these questions. USDA has released food hardship figures for 1995 through 1999, using this 18-item scale and CPS data. USDA also validated a 6-question scale for use in other surveys; this scale is shorter in order to reduce respondent burden. The 6-item scale does not included questions concerning more severe levels of hunger, and so it allows three possible determinations: "food security", "food insecurity without hunger", and "food insecurity with moderate hunger" (Bickel et al. 2001). The six questions are:
1. "The food that we bought just didn't last, and we didn't have money to get more." (Positive if this was often or sometimes true, negative is this was never true.)
2. "We couldn't afford to eat balanced meals." (Positive if this was often or sometimes true, negative if this was never true.)
3. In the last 12 months did you or other adults in your family ever cut the size of your meals or skip meals because there wasn't enough money or food? (Yes or no.)
4. (Ask only if yes on question 3.) How often did this happen--almost every month, some months, but not every month, or in only 1 or 2 months? (Positive if almost every month or some months, negative if in only 1 or 2 months.)
5. In the last 12 months, did you ever eat less than you felt you should because there wasn't enough money to buy food? (Yes or no.)
6. In the last 12 months were you ever hungry but didn't eat because you couldn't afford enough food? (Yes or no.)
Families providing zero or one positive response are considered "food secure." Those providing two, three, or four positive responses are "food insecure without hunger." And those with five or six positive responses are "food insecure with moderate hunger."
Following USDA guidelines, we imputed responses for missing values for the six food security questions, although item non-response was rare. If any question was not answered, and there was an affirmative answer on a question higher on the scale, but no negative answer lower on the scale, then an affirmative answer was imputed for that question. Otherwise, a negative answer was imputed. Fifty cases had no responses to all six food security questions, and they were coded as "food secure."
[Go To Contents]
The LANYCIS survey allows in-depth examination of food stamps use, loss and reduction among immigrant families. The survey over-samples food stamps recipients, using lists of recipients provided by public social service agencies in Los Angeles and New York. Sample weights have been adjusted to account for this sampling strategy. Two screener questions determine whether or not a family received food stamps at different points in time:
The second screener question is only asked in cases with a negative response to the first. Then a series of more detailed food stamps questions are asked of all families answering affirmative to either of these questions.
First, the respondent is asked if the family is currently receiving food stamps. If the answer is affirmative, he or she is asked about the amount of the monthly benefit, and whether it has been reduced.
If the family does not currently receive food stamps, the respondent is asked when and why benefits were terminated, the average monthly amount of the benefit before termination, and whether or not benefits were reduced prior to termination. If the benefit amount had been reduced, the respondent is asked when, why, and by how much the benefit was reduced.
If the family received food stamps since 1998 but not in 1996 or 1997, the respondent is asked a similar set of questions, excluding questions about benefits reductions.
If the family received food stamps in 1996 or 1997 but not since 1998, the respondent is asked when and why they stopped receiving food stamps, as well as how much they were receiving before benefits were terminated. Then he or she is asked if the benefit had been reduced prior to termination. If the benefit amount had been reduced, the respondent is asked when, why, and by how much the benefit was reduced.
Using data from the screener and follow-up questions described above we categorize food stamps use as follows:
1. Families Receiving Food Stamps Last Year (kept, not reduced): families who kept food stamps until the time of the survey, and did not have their benefit amount reduced
2. Families Receiving Food Stamps Last Year (left or reduced): families who had their food stamps benefits terminated or reduced during the 12-month span before the survey.
3. Families Receiving Food Stamps Last Year: All families receiving benefits during the 12-month span before the survey. Sum of categories 1 and 2.
4. Families Using Food Stamps Since 1996: All families receiving benefits since 1996/97 (answered "yes" to second food stamps screener question listed above).
5. No food stamps use: Families who did not receive food stamps at any time since 1996/97.
6. Food stamps use, termination date unknown (excluded as missing cases): Families that received food stamps, but later had benefits terminated. Information about when benefits were terminated is missing.
Construction of the variables for food stamps receipt within the last year involved some imputations to determine the date when benefits were reduced or terminated. Possible scenarios for determining reduction and termination dates include: (1) responses for both calendar year and month; (2) responses for calendar year and season; (3) response for calendar year but nonresponse for month and season; and (4) nonresponse for year. In the first scenario, no imputation is required. In the second, we impute month from season: winter to January, spring to April, summer to July, and fall to October. In the third scenario, if the year of termination or reduction is the same as the interview year, then we impute that the family lost food stamps within the last year. If the year of termination or reduction is the year before the survey, then we cannot determine whether or not benefits were lost or reduced less than 12 months before the interview. If the year the family lost benefits is more than one year before the interview, then that family left food stamps before last year.
52. There are small discrepancies between these data and the unweighted sample sizes shown in Table 1.1. For example, Table 1.1 classifies all people 65 or older as elderly, while the appendix table shows people on the basis of their original sample designation, and many of the respondents or spouses were elderly. In addition, the total sample sizes in the appendix are slightly higher because a couple of people had missing values on immigration status variables.
53. There are alternative methods to compute response rates, particularly for telephone surveys. The estimates cited here correspond to the RR6 method, as defined by the American Association for Public Opinion Research (1998). An alternative, stricter definition is RR3, which estimates a 51-percent response rate. The difference in rates involves varying assumptions about the eligibility of people who could not be contacted (e.g., where the telephone was never answered).
Main Page of Report | Contents of Report
Home Pages:
Human Services Policy (HSP)
Assistant Secretary for Planning and Evaluation
(ASPE)
U.S. Department of Health and Human Services
(HHS)