Abstract
Gay, C. 2012. “Moving to Opportunity: The Political Effects of a Housing Mobility Experiment.” Urban Affairs Review 48 (2): 147–179. 10.1177/1078087411426399.
The author requested a corrigendum to clarify attribution. The changes do not affect the substance of the analysis or the conclusions of the article, and mainly concern the use of standard statistical terminology. Although the author did not find that the level of attribution as reflected in these clarifications would have been necessary at the time or is warranted now in all instances, the author suggested these amendments to clarify any instances of ambiguity.
The corrections to the article are listed below:
Amend the following text on p. 154: ‘While the MTO demonstration was not designed to address issues of racial and ethnic concentration directly, the racial characteristics of the new neighborhoods differed modestly from the original locations and between voucher groups.’
The amended text will read: ‘While the MTO demonstration was “not designed to address issues of racial [and] ethnic concentration directly,” the racial characteristics of the new neighborhoods differed modestly from the original locations and between voucher groups (Larry et al. 2003, p. 36).’
Amend the following text on p. 159: ‘Under this assumption, one knows that the treatment impact for the noncompliers (i.e., individuals who did not use the voucher) is zero, making the ITT estimate πITT a weighted average of the impact on compliers and the zero effect on noncompliers (Bloom 1984); the weights are the portions of the sample that are compliers and noncompliers. As a result, the TOT impact can be estimated by dividing πITT by the program compliance rate for the treatment group (for similar approaches to estimating MTO impacts, see Katz, Kling, and Liebman 2001; Jens et al. 2008; Larry et al. 2003).’
The amended text will read: ‘Under this assumption, one knows that the treatment impact for the noncompliers (i.e., individuals who did not use the voucher) is zero, making the ITT estimate πITT “a weighted average of the impact on compliers and the zero effect on noncompliers—the weights are the portion[s] of the sample that are compliers and … noncompliers” (Jens et al. 2008, p. 153, citing Bloom 1984). As a result, the TOT impact can be estimated by dividing πITT by the program compliance rate for the treatment group (for similar approaches to estimating MTO impacts, see Katz, Kling, and Liebman 2001; Jens et al. 2008; Larry et al. 2003).’
Amend the following text on p. 159: ‘There are two estimates of interest that follow from MTO’s experimental design and are reported in the tables: the intent-to-treat (ITT) and the treatment-on-treated (TOT) effects. The ITT effect, estimated from the difference in mean outcomes for the treatment and control groups as a whole, is the effect of being offered the voucher, regardless of subsequent compliance (i.e., whether or not the family offered the voucher actually used it to lease up). The TOT effect is the effect of the voucher on the compliers—the MTO sample adults who actually leased up using the program voucher. Whereas the offer of a voucher was extended to every member of the experimental and section 8 treatment groups, not every group member used their assigned voucher.’
The amended text will read: ‘As Jens et al. (2008, p. 146) observe, there are two estimates of interest “that follow from MTO’s experimental design” and are reported in the tables: the intent-to-treat (ITT) and the treatment-on-treated (TOT) effects.’
Amend the following text on p. 159: ‘Thus the TOT effects, unlike the ITT effects, are nonexperimental, in the sense that they are not directly observed for whole randomly assigned groups.’
The amended text will read: ‘Thus the TOT effects, unlike the ITT effects, are “non-experimental, in the sense that they are not directly observed for whole randomly assigned groups” … (Larry et al. 2003, p. 9).’
Amend the following text on p. 159: ‘The TOT impact is inferred from the ITT impact πITT based on the weak assumption that the effect of the treatment occurs entirely through moving using a program voucher.’
The amended text will read: ‘The TOT impact is inferred from the ITT impact πITT based on “the weak assumption that the effect of the treatment occurs entirely through moving using a program voucher” … (Larry et al. 2003, appx. B, p. 7).’
Amend the following text on p. 170: ‘Past research potentially has confused the effects of neighborhoods with the effects of the (unobserved) characteristics of individuals who live in different types of neighborhoods.’
The amended text will read: ‘Past research “potentially confused the effects of neighborhood[s] with the effects of the [(unobserved)] characteristics of” individuals who live in different types of neighborhoods (Larry et al. 2003, p. 1).’
Amend the following text on p. 173: ‘The standard errors for the TOT estimates are similarly adjusted. Thus, while TOT impact estimates are substantially larger than ITT estimates (because they are not attenuated by zero effects of the intervention on the noncompliers), they are statistically significant only if the ITT estimate is significant.’
The amended text will read: ‘The standard errors for the TOT estimates are similarly adjusted. Thus, while TOT impact estimates are “substantially larger” than ITT estimates (because they are not attenuated by zero effects of the intervention on the noncompliers), they are “statistically significant only if the… ITT estimate[] [is] significant” (Larry et al. 2003, p. 9).’
The Journal Editor confirmed that the changes to the article do not change the conclusions of the article.
This corrigendum supersedes the information provided in the Expression of Concern published on 30 January 2024.