search

diegogentilepassaro / min_wage_rent

GNU General Public License v3.0
0 stars 0 forks source link

Estimate the effect of the MW on rents using the new ZORI variable #260

Closed santiagohermo closed 12 months ago

santiagohermo commented 1 year ago

In the new submission we promised to show a new analysis using the ZORI variable, which attempts to control for housing charaterics. In this issue we will

To start after #254 is merged.

Notes

Extract from new cover letter

We will add estimates using Zillow's new monthly rental housing index (ZORI). This index, based on methodology by Ambrose et al. (2015), directly attempts to control for changes in the composition of posted rents on the platform by estimating change in rents for the same rental unit over time. We hope that analyses based on this variable provide more confidence on our conclusions, in particular with respect to the concern that selection of listings may be driving our results (as suggested by R1).

santiagohermo commented 1 year ago

I opened a new issue to show some quick estimates that I attempted. All use the data as of April 23 from PR #252.

These estimates use CBSA-time FE. I've found consistent results with this model. Using only time FE I've found less robust results.

Estimates using xtevent without controlling for residence MW

xtevent ln_zori_23 if fullbal_sample_zori23, policyvar(mw_wkp_tot_17) ///
    pre(3) overidpre(3) overidpost(2) post(0) ///
    reghdfe addabsorb(cbsa_num##year_month) cluster(statefips) nocons

image

Controlling for the residence MW

xtevent ln_zori_23 mw_res if fullbal_sample_zori23, policyvar(mw_wkp_tot_17) ///
    pre(3) overidpre(3) overidpost(2) post(0) ///
    reghdfe addabsorb(cbsa_num##year_month) cluster(statefips) nocons

image

Using our usual model without the residence MW

reghdfe D.ln_zori_23 L(-6/6).D.mw_wkp_tot_17 if fullbal_sample_zori23, ///
    absorb(cbsa_num##year_month) cluster(statefips) noconsa

image

and with the residence MW

reghdfe D.ln_zori_23 L(-6/6).D.mw_wkp_tot_17 D.mw_res if fullbal_sample_zori23, ///
    absorb(cbsa_num##year_month) cluster(statefips) noconsa

image

Using our model without CBSA-time FE things and the unbalanced panel

reghdfe D.ln_zori_23 L(-6/6).D.mw_wkp_tot_17 D.mw_res, ///
    absorb(cbsa_num##year_month##yr_entry_to_zillow_zori2023) cluster(statefips) nocons

image

santiagohermo commented 1 year ago

Opening comment updated

santiagohermo commented 12 months ago

This was completed in #268 @diegogentilepassaro @gabrieleborg. In particular, see https://github.com/diegogentilepassaro/min_wage_rent/commit/b3e4280607109dcf2c05357d3505de7072449af7