Procedure to impute property taxes paid by renters

[jo-fleck]

1. Match renter households in ASEC by state and county (where possible) to similar households in the ACS (same matching variables as we used for owners) and impute an estimate of rent paid to each ASEC renter household, $\hat{r}_{j}$.

2. Use the Zillow price rent ratios by state to estimate home values $\hat{P}_{j}$.

[See here for an issue regarding the Zillow data]

3. At the state and county level (where possible) estimate the property tax rate paid given the ACS household income level.

• We estimate property tax rates using reported home values and property taxes for homeowners

• For each homeowner we have an individual property tax rate defined as property taxes divided by home value

• We fit a function to these property tax rates at the state (and county) level as $\tau{i}(y)=\alpha {0,i}+\alpha {1,i}\log (y)+\alpha {2,i}\log (y)^{2}$ where $i$ denotes state and county. We run this regression at the individual household level.

• Then, when we have estimated the coefficients, we impute property taxes for renter household $j$ in ASEC as $proptax{j,i} = ( \alpha{0,i}+\alpha{1,i}\log (y{j})+\alpha{2,i}\log (y{j})^{2} ) \hat{P}_{j}$

4. Then the last step was to make a scatterplot where we plot property taxes against log household income for both homeowners and renters in the same state, with the homeowner dots in blue and the renter dots in red

[jo-fleck]

After talking with Jiaxi on May 30 2023 (she will write a comprehensive description using the tex file I shared with her):

1. Use the Zillow price rent ratios by state to impute home values $\hat{P}_j$ for every renter in ACS using their self-reported monthly rent payments $r_j$.
2. Use the self-reported property taxes and home values of ACS owners $i$ to fit a function for property tax rates $t^p$ at the household level $t^p{i}=\alpha {0,c}+\alpha _{1,s}\log (yi)+\alpha {2,s}\log (yi)^{2}$ where $c$ denotes county and $s$ state so $\alpha {0,c}$ is a county fixed effect and $\alpha {1,s}$ and $\alpha {2,s}$ differ by state.
3. Using the estimated coefficients $\alpha$, we impute property taxes for ACS renters $j$ as $proptax{j,i} = ( \alpha{0,c}+\alpha{1,s}\log (y{j})+\alpha{2,s}\log (y{j})^{2} ) \hat{P}_{j}$
4. Using a nearest neighbor matching procedure (same matching variables as for owners) match renters in ASEC by state and county (where possible) to ACS renters and assign their property taxes.

[jo-fleck]

@jiaxitan Please write up a comprehensive description using the tex file I shared. Also include some of the explorations and robustness exercises we did while developing the procedure (graphs etc).