«Moving to opportunity voluMe 14, nuMber 2 • 2012 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
GAD required that the adult experience a period of at least 6 months in which they had excessive anxiety about multiple things and at least three of the following symptoms: restlessness, easy fatigue, difficulty concentrating, irritability, muscle tension, and sleep disturbance. In addition, the anxiety Our TOT estimates assume that families who are offered a voucher but do not take up the offer (that is, do not use a program voucher to move) are not affected by the offer itself. This assumption may not be strictly true, because families may have changed some of their thinking or behaviors (such as looking at apartments) as a result of the offer. We think, however, that the effects of actually using the voucher are likely to be much larger than any effects of being offered the voucher and not using it, and that it is reasonable to assume that the ITT effects are driven strictly by effects on compliers.
120 Moving to Opportunity The Long-Term Effects of Moving to Opportunity on Adult Health and Economic Self-Sufficiency had to cause significant distress or impair respondents’ functioning. Our measure of PTSD used a subset of the CIDI questions and then imputed the probability of PTSD from those responses using data from a national study.5 To be categorized as having PTSD, the respondent had to have experienced, witnessed, or been confronted by a traumatic event that involved actual or threatened death or serious injury to themselves or others,6 and the trauma had to invoke at least three of the following symptoms: avoiding activities, places, or people that arouse recollections of the trauma;
reduced interest in activities; feelings of detachment; restricted range of feelings; and a foreshortened sense of the future. PTSD also involves difficulty falling or staying asleep, an exaggerated startle response, and impairment of functioning.
We measured alcohol or drug dependence using the Severity of Dependence Scale (SDS). The SDS consists of five questions about out-of-control use, anxiety or worry about missing a fix or a drink, worry about use, frequency of desire to end use, and difficulty of going without use (Gossop et al., 1995). The SDS ranges from 0 to 15, and we consider a score of 3 or greater to indicate substance dependence.
Physical Health Measures We studied physical health outcomes using a combination of survey questions, physical measurements, and assays from dried blood spot samples. The survey asked adults if their health was excellent, very good, good, fair, or poor; if they had suffered an asthma or a wheezing attack in the past year; and if their health limited them in climbing several flights of stairs or lifting or carrying groceries (Wiener et al., 1990). To assess obesity, interviewers measured each respondent’s height and weight.7 We calculated BMI by dividing respondents’ weight in kilograms by the square of their height in meters. We looked at effects stratified by three levels of obesity (BMI of 30 or greater, 35 or greater, and 40 or greater)8 because of evidence that very high BMI values may be strongly associated with subsequent adverse health outcomes (NHLBI Obesity Education Initiative, 1998).
A woman who is of average height in the United States (5 feet, 4 inches) would need to weigh 204 pounds to have a BMI of 35 and would need to weigh 233 pounds to have a BMI of 40. In addition to height and weight, interviewers took respondents’ seated blood pressure using an automated cuff.9 We used the average of two readings to assess hypertension10 and considered individuals to be hypertensive if their average systolic pressure was 140 millimeters of mercury or higher or their average diastolic pressure was 90 millimeters of mercury or higher (Chobanian et al., 2003).
We used data from the National Comorbidity Survey Replication (Kessler and Merikangas, 2004) to calculate the likelihood of PTSD based on the questions asked in the MTO survey and taking into account age, gender, race, and ethnicity.
The MTO survey asked about the following events: beaten up as a child by a primary caregiver, a spouse or romantic partner, or someone else; raped or sexually assaulted; mugged or threatened with a weapon; unexpected death of a loved one; traumatic event experienced by a loved one; witnessed physical fights at home; witnessed death or saw a dead body or someone seriously hurt; or some other traumatic event.
We measured height and weight in accordance with the protocols developed for the Health and Retirement Survey (ISR, 2008).
Ludwig et al. (2011) reported findings on MTO’s effects on obesity and diabetes.
The automated sphygmomanometer used was Omron model HEM-711DLX.
We considered a reading valid if diastolic blood pressure (pressure when the heart is at rest between contractions) was more than 40 millimeters of mercury, the systolic blood pressure (pressure when the heart is contracting) was more than 60 millimeters of mercury, and the systolic was at least 10 points higher than the diastolic.
We measured diabetes using both survey questions and blood samples. Interviewers asked respondents whether they had diabetes (or high blood sugar) or had received treatment for it during the past year. Because many people with diabetes are unaware of their condition, we collected blood spot samples from respondents and measured glycosylated hemoglobin level as an indicator of diabetes.11 Economic Measures Our last set of measures focused on the economic self-sufficiency of adults. We measured employment, earnings, household income, and use of government assistance programs through both survey questions and administrative records. We drew on employment questions from the Current Population Survey and considered MTO respondents to be employed currently if they reported working for pay during the week prior to the interview (or reported being temporarily absent from their job because of illness or vacation). Interviewers asked respondents how much money they earned in the previous year, whether they were currently receiving food stamps or Temporary Assistance for Needy Families (TANF), and how much income their household (all members combined) received in the previous year. Information on household income enabled us to determine whether their household was above or below the U.S. Census Bureau’s poverty threshold.
For example, a family consisting of one mother and two children would be below the poverty threshold if they had an income of less than $17,285 in 2009. We also matched the MTO sample to administrative data on quarterly earnings from state unemployment insurance (UI) agencies and to TANF and food stamps records from state (or county) agencies. We used the matched data to look at employment and earnings in 2007 and at receipt of TANF or food-stamp benefits over the 2-year period from July 2007 through June 2009.12 MTO Effects on Adult Outcomes In this section we present our estimates of MTO’s effects on the mental health, physical health, and economic self-sufficiency of adults in the program. Our impact estimates are based on the regression model from equation 1.
Mental Health Effects Exhibit 3 shows evidence of beneficial MTO effects on the mental health of adults in terms of lower depression and levels of psychological distress, but the experimental group also shows an increase in substance dependence. All remaining exhibits are structured the same way as Exhibit 3. Each row presents the findings for the outcome listed in the left-hand column; the first outcome is depression. The second column, Control Mean, shows that approximately 20.3 percent of adults in the control group suffered from major depression during their lifetime. The third column displays the ITT effect of being offered an experimental voucher, estimated by comparing the entire experimental group with the entire control group. Adults in the experimental group met the criteria for major depression at a rate that was 3.2 percentage points less than the rate for adults in the Glycosylated hemoglobin (HbA1c) captures the average glucose level in the blood during the past several months. The American Diabetes Association (2010) recommends using HbA1c levels of 6.5 percent or higher to diagnose diabetes.
Data availability limits our analyses of TANF and food stamps to participants from Boston, Chicago, and Los Angeles.
122 Moving to Opportunity The Long-Term Effects of Moving to Opportunity on Adult Health and Economic Self-Sufficiency
control group (p.10, as indicated by the tilde). The standard error (shown in parentheses) is
1.7 percentage points.13 In the fourth column, our estimates suggest that moving using an experimental voucher, or the TOT effect, reduces the prevalence of lifetime depression by 6.6 percentage points (about twice the size of the ITT effect). In the fifth and sixth columns, we turn to the effects (ITT and TOT) for the Section 8 group. Being offered a traditional Section 8 voucher reduces the prevalence of lifetime depression by 4.8 percentage points, and actually using the voucher reduces depression by an estimated 7.7 percentage points (p.05). The final column of the table shows the number of observations used in the analysis.
The standard error indicates the estimate’s precision. The effect plus or minus the standard error, multiplied by 1.96, captures the 95-percent confidence interval around the effect, in this case implying a confidence interval ranging from an increase of 0.1 percentage points to a decrease of 6.5 percentage points in depression.
Similar percentages (about 6.5 percent in each group) of adults in both the control group and the experimental group met the criteria for generalized anxiety disorder (lifetime). Adults in the Section 8 group were marginally less likely than adults in the control group to have had GAD (p =.057). Of adults in the control group, 22 percent met the criteria for PTSD, and the prevalence for the voucher groups was not significantly different from this.
Turning to a more global measure of mental health—psychological distress—we find beneficial program effects. The average psychological distress levels of adults in the experimental and Section 8 groups are about one-tenth of a standard deviation less than adults in the control group. The impact estimate achieves statistical significance for the experimental group (p =.011) but not the Section 8 group (p =.084). (To make the K6 results easier to interpret, we standardized the units by subtracting off the control group mean and dividing by the control group standard deviation to create what is known as a z-score.) We observe an adverse effect on substance dependence. About 5.5 percent of adults in the control group met the criteria for substance dependence during the past month and assignment to the experimental group was associated with a prevalence that was 2.9 percentage points higher than that of the control group. For the Section 8 group, we do not detect a statistically significant effect on dependence.
Physical Health Effects As shown in Exhibit 4, MTO appears to reduce the share of adults with diabetes and the likelihood of severe obesity, but we do not detect any treatment effects on several other health measures. About 56 percent of adults in the control group indicated that their current health was good or better;
reports by adults in the experimental and Section 8 groups were similar. About 29 percent of adults in the control group reported having had an asthma attack in the past year. The rates were slightly lower for the two treatment groups, but the differences were not statistically significant. The average weight of adults in the control group was about 190 pounds, and about 58 percent of controls met the criteria for obesity (BMI of 30 or greater). Although we detect no statistically significant effects on the likelihood of having a BMI of 30 or more, we do detect beneficial program effects at more extreme obesity levels. The experimental and Section 8 groups are 4.6 and 5.3 percentage points, respectively, less likely to have had a BMI of 35 or greater compared with adults in the control group who had a prevalence of 35 percent (p.05). These estimates imply that actually moving using a voucher reduces the prevalence of a BMI of 35 or greater by about 9.5 percentage points for the experimental group and 8.6 percentage points for the Section 8 group. For a BMI of 40 or greater, the ITT effect was a 3.4-percentage-point reduction for adults in the experimental group (p.05) and a 2.9-percentage-point reduction for adults in the Section 8 group (not significant).
Both types of diabetes measures (self-reports and blood samples) point in the direction of lower diabetes rates among adults in the treatment groups compared with adults in the control group, although the exact magnitudes and levels of statistical significance vary. When we rely on respondent self-reports to measure diabetes, we find that the estimated decline in diabetes prevalence for adults in the experimental group relative to adults in the control group is not quite statistically significant, whereas adults in the Section 8 group have a significantly lower rate (ITT of 6.1 percentage points) compared with adults in the control group. Compared with self-reports, the blood test results 124 Moving to Opportunity The Long-Term Effects of Moving to Opportunity on Adult Health and Economic Self-Sufficiency
show that a greater share of adults in the control group had diabetes. Using the blood test results, we find a statistically significant 5.2-percentage-point experimental ITT effect (and corresponding 10.8-percentage-point TOT effect), but we do not detect an effect for the Section 8 group. (We previously reported diabetes results for MTO adult women in Ludwig et al., 2011.) The last two health measures we present are physical limitations and hypertension, on which we only detect an effect on health limitations among adults in the experimental group. Of adults in the control group, 51 percent reported that their health limited them in everyday activities, and adults in the experimental group were 4.8 percentage points less likely to report this type of limitation.
The difference between the control group and the Section 8 group was not statistically significant.
Nearly 32 percent of adults in the control group had hypertension, and the incidence of hypertension for adults in the experimental and Section 8 groups was similar to that of adults in the control group, with no significant differences detected.