The Two-Income Trap - Elizabeth Warren [95]
Districts from five different states, each from a different region in the country, provided some geographic diversity. In addition, by drawing from some of the same districts as the 1981 and 1991 samples, the 2001 study maintains some continuity for researchers interested in longitudinal research. The choice of the specific districts within each state increased the size of the pools from which the sample was drawn, increasing the representativeness of the sample within the geographic location.
The five metropolitan areas captured in the core sample represent another form of diversity: the incidence of bankruptcy filings in the local population. To the extent that the decision to file for bankruptcy might be influenced by local legal cultures, variance in the filing rates per thousand may indicate differing bankruptcy practices as well.5 According to data compiled by SMR Research Corporation, one area, Nashville, has higher rates of bankruptcy filings than most other U.S. cities, at 9.70 bankruptcy filings per 1,000 adults. Closer to the other end of the spectrum were Philadelphia (4.91) and Dallas (5.02), which have lower rates of bankruptcy filings than the U.S. population generally. In between were Chicago (5.83) and Los Angeles (8.02).6
This core sample varies from a representative national sample of bankrupt debtors in that it represents relatively few districts. The distribution and completion of the questionnaires also introduce an opportunity for nonresponse bias. This bias could be introduced if the people who complete the questionnaires differ systematically from the people who do not complete questionnaires in ways that might affect the variables of interest. It is possible, for example, that people who are very uncomfortable in the courthouse would reject the questionnaire. This discomfort (an unmeasured characteristic) might be linked to some other characteristic that was measured (such as size of total debt). By their very nature, such biases are difficult to detect.
Two events in 2001 might cause some speculation that our sample was prone to seasonality bias: the events of September 11, 2001, and their subsequent economic impact, and the publicity given the debates in Congress concerning revision of the bankruptcy laws. In fact, however, the sample was drawn earlier in 2001, before either of these events had occurred.
Supplemental Samples
To expand the analysis and to probe additional issues, two supplemental samples were developed. A central focus of the research is the relationships between housing, mortgages, and bankruptcy. To study these relationships with greater reliability, the research design included supplementary collections of homeowners in bankruptcy, drawn from Los Angeles, Chicago, and Philadelphia. In order to examine financial problems outside urban centers, an extended sample was drawn from rural areas in Tennessee and Iowa.
TABLE A.1 Summary of core and supplemental samples
The quota sampling technique was also employed for both supplemental samples (homeowners and rural debtors), with the addition of more qualifying criteria. The supplemental rural sample consists only of those debtors living in certain predetermined rural zip codes; all respondents not in those zip codes were eliminated from the supplemental sample. The Los Angeles supplemental housing sample was developed by continuing to collect questionnaires beyond the original target of 250 until a total of 100 homeowners had been collected; nonhomeowner cases were discarded from this supplemental sample. The supplemental housing sample for Philadelphia and Chicago used two filters: The debtors were homeowners and they agreed to be interviewed by telephone; nonhomeowners and those who did not wish to be interviewed were omitted from those supplemental samples.
To recap: The core sample consists of 1,250 debtors, 250 from each of the first five states. This sample was supplemented