Private provision of local public goods and housing market responses: Evidence from London

• Both referees inquired why the effects in the main event studies fade after period 5. We believe that the main reason for the drop in effects is a reduction in the number of control groups over time. This estimator compares properties in earlier BIDs with properties in later treated ones. Hence, t=0 it compares the BIDs that first open with later treated ones, in t=1 compares those that open in t=0 and t=1 with those that open later and so on. As we move further, the number of control areas reduce. To attempt and answer this concerned we have added a robustness exercise where we consider BIDs that opened by 2010, and use instead never treated units as controls. The event-studies of this exercise reveal now more persistent effects, with a positive and significant coefficient being found in the 10-year post-treatment period. We have added the corresponding figures to the appendix. • To address the comments from Referee 2, we have included an additional specification using local authority trends, with results remaining largely consistent. We have also added a figure that further explores BID heterogeneity by splitting around the median share of deprived households and vacant space, using data from the 2001 Census. The corresponding Table and Figure were added to the appendix. • Both referees inquired about the rationale behind using a two-stage process, here firstly prices are residualized using observable characteristics, then we apply the Callaway and Sant’Anna (CS) DID estimator to these residuals. We have clarified, both in the letters to the referees and in the manuscript, that the reason why we do not incorporate these observable characteristics directly into the estimator as controls is because the csdid Stata package sets the value of the covariate to be equal to the value of the covariate in the “base period”. In post-treatment periods the base period is the period immediately before observations in a particular group become treated (when there is anticipation, it is before anticipation effects start too), and in pre-treatment periods the base period is the period right before the current period. While this is of clear interpretation with analyses of “fixed units” over time (such as individuals, blocks, municipalities), we believe that applying the same reasoning to repeated cross-sectional settings like ours would not be correct.