PR for #247: Submit to AER (#249)

* Move to "Online Appendix" terminology for AER #247

* Restart counter in online appendix #247

* Compiles #247

* Compile paper #247

* 98-words abstract + compiles paper #247

* Update `/reply/` #247

* Improved disclosure statements #247

paper_slides/figures/dynamic_stacked.tex

@@ -16,7 +16,8 @@
         and the QCEW.
         The figure mimics estimates in Figure \ref{fig:dynamic_baseline} 
         using a ``stacked'' sample.
-        We construct the sample as explained in Appendix Table \ref{tab:stacked_w6}.
+        We construct the sample as explained in Online Appendix Table 
+        \ref{tab:stacked_w6}.
         95\% pointwise confidence intervals are obtained from standard errors 
         clustered at the state level.
     \end{minipage}

paper_slides/figures/map_housing_exp_chicago_jul2019.tex

@@ -11,7 +11,7 @@
         Notes:
         Data are from the \textcite{IRS} and the \textcite{hudSAFMR}.
         The figure shows housing expenditure shares as computed in
-        Appendix \ref{sec:measure_housing_expenditure}, namely,
+        Online Appendix \ref{sec:measure_housing_expenditure}, namely,
         by dividing the SAFMR 40th percentile rental value for a 2-bedroom 
         apartment by average monthly wage per household divided, both
         for 2018.

paper_slides/paper/appendix.tex

@@ -155,7 +155,7 @@ \subsection{Assigning Minimum Wage Levels to USPS ZIP Codes}
 based on the names of the county and the place.
 We define the statutory MW at each census block as the maximum of the federal,
 state, county, and place levels.
-Then, based on the original correspondence table described in Appendix 
+Then, based on the original correspondence table described in Online Appendix 
 \ref{sec:blocks_to_uspszip}, we assign a ZIP code to each block.
 Finally, we define \textit{the statutory MW} for ZIP code $i$ and month $t$, 
 $\MW_{it}$, as the weighted average of the statutory MW levels in its
@@ -189,7 +189,7 @@ \subsubsection*{Locating minimum wage earners}
 \footnote{More precisely, we compute a tract-to-ZIP-code correspondence from
 the LODES correspondence between blocks and tracts, available in 
 \textcite{CensusLODES}, and the geographical match between blocks and ZIP codes
-from Appendix \ref{sec:blocks_to_uspszip}.
+from Online Appendix \ref{sec:blocks_to_uspszip}.
 For each tract, we compute the share of houses that fall in each ZIP code, and 
 we assume that the share in the tract-ZIP code combination equals the share of
 houses times the estimated number of MW workers in the tract.}
@@ -442,8 +442,8 @@ \section{The Effect of the Minimum Wage on Income Across Space}
 %%    wages and everything works out
 %%
 
-Appendix Table \ref{tab:static_wages} shows estimates of $\varepsilon$ for 
-different specifications of the model given in \eqref{eq:wage_level_model}.
+Online Appendix Table \ref{tab:static_wages} shows estimates of $\varepsilon$ 
+for different specifications of the model given in \eqref{eq:wage_level_model}.
 Columns (1) through (3) estimate the effect of the workplace MW on log total
 wages under different specifications.
 The point estimates of $\varepsilon$ fluctuate between 0.0909 and 0.1275.
@@ -460,13 +460,14 @@ \section{The Effect of the Minimum Wage on Income Across Space}
 that a 10 percent increase in the MW will increase total wages by 
 $(6.8/10.1)\times 10\times 0.15 \approx 1.01$ percent.
 
-Column (4) of Appendix Table \ref{tab:static_wages} replicates column (3)
+Column (4) of Online Appendix Table \ref{tab:static_wages} replicates column (3)
 but interacts the workplace MW measure with the standarized share of MW workers
-estimated as explained in Appendix \ref{sec:assigning_mw_levels}.
+estimated as explained in Online Appendix \ref{sec:assigning_mw_levels}.
 As expected, we find that a higher share of MW workers makes the effect of
 workplace MW increases larger.
-Column (5) of Appendix Table \ref{tab:static_wages} shows, as a falsification
-test, estimates of the same model as in column (3) but using the log of total
+Column (5) of Online Appendix Table \ref{tab:static_wages} shows, 
+as a falsification test, 
+estimates of the same model as in column (3) but using the log of total
 dividends as dependent variable.
 We obtain a positive but much lower effect that is statistically 
 indistinguishable from zero, suggesting that dividends do not respond to 

paper_slides/paper/context_data.tex

@@ -35,15 +35,15 @@ \subsection{Minimum Wage Policies in the 2010s}
 levels.
 In these cases we assign a weighted average of the statutory MW levels in its
 constituent census blocks, exploiting an original correspondence table between 
-blocks and ZIP codes detailed in Appendix \ref{sec:blocks_to_uspszip}, 
+blocks and ZIP codes detailed in Online Appendix \ref{sec:blocks_to_uspszip}, 
 where weights correspond to the number of housing units.
 The result is a ZIP code-month panel of statutory MW levels in the US between
 January 2010 and June 2020.
-More details on the construction of the panel can be found in Appendix 
+More details on the construction of the panel can be found in Online Appendix 
 \ref{sec:assigning_mw_levels}.
 
-Appendix Figure \ref{fig:mw_policies} shows the different levels of binding MW 
-policies over time in our data.
+Online Appendix Figure \ref{fig:mw_policies} shows the different levels of 
+binding MW policies over time in our data.
 Panel A focuses on state-level MW policies.
 There are $\stateBindingMW$ states with MW policies in 2010--2020, all of 
 which started prior to January 2010.
@@ -75,10 +75,11 @@ \subsection{Relationship Between Income and Housing}
 While approximately $\TopDecPrRent$ percent of households in the bottom quintile 
 are renters, only around $\BottomDecPrRent$ percent of households in the top 
 quintile are.
-Appendix Figure \ref{fig:ahs_rent_sqft} shows that, among households that rent, 
-rents per square foot are surprisingly constant across household income levels.
-Appendix Figure \ref{fig:ahs_unit_types} shows the type of building households
-live in by household income decile.
+Online Appendix Figure \ref{fig:ahs_rent_sqft} shows that, among households 
+that rent, rents per square foot are surprisingly constant across household 
+income levels.
+Online Appendix Figure \ref{fig:ahs_unit_types} shows the type of building 
+households live in by household income decile.
 
 %%
 %% Things that we may also say:
@@ -110,15 +111,15 @@ \subsection{Relationship Between Income and Housing}
 rental value from the HUD.%
 \footnote{We impute a small share of missing values using a regression model 
 where the ZIP code-level covariates include data from LODES and the US Census.
-See Appendix \ref{sec:measure_housing_expenditure} for details.}%
+See Online Appendix \ref{sec:measure_housing_expenditure} for details.}%
 \textsuperscript{,}%
 \footnote{This computation will be a good approximation for the housing 
 expenditure share insofar total housing expenditure and total wage income are 
 roughly proportional to their averages under the same constant of 
 proportionality.
 This computation also assumes away differences in the number of bedrooms across 
 ZIP codes.}
-Appendix Figure \ref{fig:map_hous_exp_chicago} maps our estimates for the 
+Online Appendix Figure \ref{fig:map_hous_exp_chicago} maps our estimates for the 
 metropolitan area of Chicago.
 There is considerable variation in housing expenditure over space, with poorer
 areas generally spending a higher share of their income in housing.
@@ -132,7 +133,7 @@ \subsection{Relationship Between Income and Housing}
 Thus, to get a sense of the spatial distribution of minimum wage earners we 
 construct a proxy variable using the number of workers across income bins
 in the 5-year 2010-2014 American Community Survey \parencite[ACS;][]{CensusACS}.
-See details in Appendix \ref{sec:assigning_mw_levels}.
+See details in Online Appendix \ref{sec:assigning_mw_levels}.
 Our variable for the share of MW workers is negatively correlated with median 
 household income from the ACS (corr.\ $=\corrShWkrMedInc$) and 
 positively correlated with our estimate of the housing expenditure share 
@@ -175,21 +176,21 @@ \subsubsection{Rents Data}
 covers the most common US rental house types \parencite{Fernald2020}.
 We focus on rents \textit{per square foot} to account for systematic differences
 in housing size.
-In fact, as shown in Appendix Figure \ref{fig:ahs_rent_sqft}, this variable does 
-not seem to vary much with household income.
+In fact, as shown in Online Appendix Figure \ref{fig:ahs_rent_sqft}, this 
+variable does not seem to vary much with household income.
 Our main outcome variable represents the median rental price per square foot in 
 the SFCC category among units listed in the platform for a given ZIP code and 
 month.
-Appendix Figure \ref{fig:trend_zillow_safmr} shows that this series follows a 
-similar trend over time when compared to SAFMR.
+Online Appendix Figure \ref{fig:trend_zillow_safmr} shows that this series 
+follows a similar trend over time when compared to SAFMR.
 We show results using median rents per square foot in other rental categories 
 available in the data as well.
 
 The Zillow data have several limitations.
 First, Zillow's market penetration dictates the sample of ZIP codes available.
-Appendix Figure \ref{fig:map_zillow_sample} shows that the sample of ZIP codes
-with valid SFCC rents data typically coincides with areas of high population 
-density.
+Online Appendix Figure \ref{fig:map_zillow_sample} shows that the sample of ZIP 
+codes with valid SFCC rents data typically coincides with areas of high 
+population density.
 Second, we only observe the median per-square-foot rental value among listings.
 We do not observe actual rents paid by tenants in a given period, 
 the distribution of rents among listings in the given ZIP code and month, nor
@@ -212,7 +213,7 @@ \subsubsection{The residence and workplace minimum wage measures}
 We collected the datasets for ``All Jobs.''
 The raw data are aggregated at the census block level. 
 We further aggregate it to ZIP codes using the original correspondence between 
-census blocks and USPS ZIP codes described in Appendix 
+census blocks and USPS ZIP codes described in Online Appendix 
 \ref{sec:blocks_to_uspszip}.
 This results in ZIP code residence-workplace matrices that, for each location 
 and year, indicate the number of jobs of residents in every other location.
@@ -256,8 +257,8 @@ \subsubsection{The residence and workplace minimum wage measures}
 illustrates the difference in these measures by plotting the change in the 
 residence and workplace MW measures in the Chicago-Naperville-Elgin CBSA in
 July 2019.
-For completeness, Appendix Figure \ref{fig:map_rents_chicago_jul2019} shows
-the changes in our main median rents variable around the same date.
+For completeness, Online Appendix Figure \ref{fig:map_rents_chicago_jul2019} 
+shows the changes in our main median rents variable around the same date.
 
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
@@ -305,7 +306,8 @@ \subsubsection{Other data sources}\label{sec:data_other}
 In order to describe our sample of ZIP codes we collect data from the ACS 
 \parencite{CensusACS} and the 2010 US Census \parencite{CensusDecennial}.
 We collect these data at the block or tract levels, and assign it to ZIP codes
-using the correspondence table described in Appendix \ref{sec:blocks_to_uspszip}.
+using the correspondence table described in Online Appendix 
+\ref{sec:blocks_to_uspszip}.
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 \subsubsection{Estimation samples}\label{sec:data_final_panel}
@@ -315,8 +317,8 @@ \subsubsection{Estimation samples}\label{sec:data_final_panel}
 This panel contains $\ZIPMWeventsUnbal$ MW changes at the ZIP code level, 
 which arise from $\StateMWeventsUnbal$ state-level changes and 
 $\CityCountyMWeventsUnbal$ county- and city-level changes.
-Appendix Figure \ref{fig:mw_changes_dist_zillow} shows the distribution of 
-positive increases in our statutory MW variable among all ZIP codes available 
+Online Appendix Figure \ref{fig:mw_changes_dist_zillow} shows the distribution 
+of positive increases in our statutory MW variable among all ZIP codes available 
 in the Zillow data.
 Given that ZIP codes enter the Zillow data progressively over time affecting 
 the composition of the sample,
@@ -350,7 +352,7 @@ \subsubsection{Estimation samples}\label{sec:data_final_panel}
 we conduct an estimation exercise where we re-weight our sample to match the 
 average of a handful of characteristics of those.
 
-Finally, Appendix Table \ref{tab:stats_est_panel} shows sample statistics 
+Finally, Online Appendix Table \ref{tab:stats_est_panel} shows sample statistics 
 of our baseline panel.
 The distribution of the residence and workplace MW measures is, as expected,
 quite similar.

paper_slides/paper/counterfactual.tex

@@ -69,7 +69,7 @@ \subsection{Empirical Approach}\label{sec:emp_cf}
 While gauging the spillover effects of the MW on wages across ZIP codes is 
 not the main goal of the paper, estimates of $\varepsilon$ are not readily 
 available in the literature.
-Appendix \ref{sec:mw_on_income} discusses the details of our estimation 
+Online Appendix \ref{sec:mw_on_income} discusses the details of our estimation 
 strategy, and compares the results with estimates of the effect of the MW 
 on income of workers in the same jurisdiction.
 We also show that our results are heterogeneous depending on the share of 
@@ -143,8 +143,8 @@ \subsubsection{Counterfactual increases in residence and workplace MW levels}
 and (ii) the assumed value for the federal or city MW in January 2020.%
 \footnote{To be more precise, we take the maximum between the MW levels of 
 different jurisdictions at the level of the block.
-Then, we aggregate up to ZIP codes using the correspondence table in Appendix 
-\ref{sec:blocks_to_uspszip}.
+Then, we aggregate up to ZIP codes using the correspondence table in 
+Online Appendix \ref{sec:blocks_to_uspszip}.
 We do so to account for the fact that the new MW policy may be partially 
 binding in some ZIP codes.}
 Then, we compute the counterfactual values of the residence MW and the workplace
@@ -154,7 +154,7 @@ \subsubsection{Counterfactual increases in residence and workplace MW levels}
 \paragraph{Federal increase}
 
 The distributions of counterfactual increases in the MW measures are displayed 
-in Appendix Figure \ref{fig:cf_hist_res_and_wkp_mw}.
+in Online Appendix Figure \ref{fig:cf_hist_res_and_wkp_mw}.
 Out of the $\zipcodesFedNine$ ZIP codes that satisfy our criteria, 
 $\zipNoIncFedNine$ (or $\zipNoIncPctFedNine$\%) experience no increase in 
 the residence MW at all.
@@ -167,8 +167,8 @@ \subsubsection{Counterfactual increases in residence and workplace MW levels}
 are still visible in the histogram of the workplace MW.
 However, we observe more places experiencing moderate increases in this measure.
 
-Panel A of Appendix Figure \ref{fig:map_chicago_cf_wkp_res} maps the changes 
-in the residence and workplace MW in the Chicago-Naperville-Elgin CBSA.
+Panel A of Online Appendix Figure \ref{fig:map_chicago_cf_wkp_res} maps the 
+changes in the residence and workplace MW in the Chicago-Naperville-Elgin CBSA.
 Unlike in Figure \ref{fig:map_mw_chicago_jul2019}, we observe the MW increasing 
 from the outside of Cook County and spilling over inside it.
 
@@ -182,8 +182,8 @@ \subsubsection{Counterfactual increases in residence and workplace MW levels}
 In this case, there are $\zipIncChiFourteen$ ZIP codes whose 
 residence MW are affected by this change and $\zipNoIncChiFourteen$ 
 that remain directly unaffected.
-Panel B of Appendix Figure \ref{fig:map_chicago_cf_wkp_res} shows the changes 
-in both MW measures after this policy.
+Panel B of Online Appendix Figure \ref{fig:map_chicago_cf_wkp_res} shows the 
+changes in both MW measures after this policy.
 As expected, we observe large increases in the workplace MW in the city, 
 which become smaller as one moves away from it.
 
@@ -195,8 +195,8 @@ \subsubsection{The share of extra wage income pocketed by landlords}
 Following the results in Table \ref{tab:static}, we take 
 $\beta = \betaCf$ and 
 $\gamma = \gammaCf$.
-Based on the results discussed in Appendix \ref{sec:mw_on_income}, we take
-$\varepsilon = \epsilonCf$.
+Based on the results discussed in Online Appendix \ref{sec:mw_on_income}, we 
+take $\varepsilon = \epsilonCf$.
 We follow the procedure outlined in the previous subsection to estimate the 
 incidence of the counterfactual policy.
 
@@ -213,15 +213,15 @@ \subsubsection{The share of extra wage income pocketed by landlords}
 
 Panel A of Figure \ref{fig:map_chicago_cf_shares} maps the estimated shares 
 in the Chicago-Naperville-Elgin CBSA.
-Panel A of Appendix Figure \ref{fig:map_chicago_cf_rents_wages} shows
+Panel A of Online Appendix Figure \ref{fig:map_chicago_cf_rents_wages} shows
 estimated increases in rents and wage income.
 We estimate a larger share pocketed in Cook County.
 The reason is that these ZIP codes experience the new policy only through
 their workplace MW and, as a result, rents increase relatively more than
 wage income.
 We also observe a larger incidence on landlords in the south of Cook County,
 where the housing expenditure share is larger 
-(as reflected in Appendix Figure \ref{fig:map_hous_exp_chicago}).
+(as reflected in Online Appendix Figure \ref{fig:map_hous_exp_chicago}).
 
 The top rows of Panel A in Table \ref{tab:counterfactuals} show the medians of 
 the key estimated objects for two groups:
@@ -262,7 +262,7 @@ \subsubsection{The share of extra wage income pocketed by landlords}
 $\rhoMedCentsIndirChiFourteen$ cents for the median not directly treated one.
 
 Panel B of Figure \ref{fig:map_chicago_cf_shares} maps the shares.
-Panel B of Appendix Figure \ref{fig:map_chicago_cf_rents_wages} shows the 
+Panel B of Online Appendix Figure \ref{fig:map_chicago_cf_rents_wages} shows the 
 estimated changes in rents and total wages.
 Unlike the previous exercise, the share pocketed by landlords is now higher 
 right outside of Chicago City.