DISCREPANCIES FROM THE PUBLISHED ARTICLE

Replication date: 20 October 2026
Software used: Mathematica 15.0; Stata 19.5; MATLAB R2026a

This package reproduces the authors' calculations. The programs write the
calculated values; nothing is adjusted toward the printed values. The
following differences from the published article remain.

Figures 1 and 2: the equilibrium is computed from the model primitives
by mathematica/HousingEquilibrium.wl. Its price and allocation curves differ 
from the original grid by less than 0.01 away from the end points, the figures 
are visually identical. Fonts, line widths and the EPS page size differ 
between Mathematica 9 and 15; see mathematica/README.md.

Figure 4: the published figure was computed from the 1998 survey, while the
text describes the 2007 experiment. Figure4_1998 reproduces the published
figure; Figure4_2007 is the experiment described in the text.

Table 1: 21 of 24 entries agree at the published precision. The exceptions are:

Entry                                             Replicated       Published
Houston, percentage change, elasticity 0.4          +1.0281%          +0.1%
Baltimore, dollar change, elasticity 0.5            -11.4842 thousand -11.0 thousand
Boston, percentage change, elasticity 0.5           -5.8476%          -5.9%

Their causes have not been established; transcription errors or different 
historical calculation versions are possible.

Table B1: 61 of 66 entries agree at the published precision. Minneapolis's 2007
annual-income Gini is 0.351957, rounding to 0.35 rather than the published 0.36.

The metropolitan area income-house-value rank correlations are computed on 
a different sample from the other columns. The article states that all results 
exclude the bottom 5% of house values and the top-coded houses, and the counts 
and Ginis are reproduced on that sample. The published rank correlations are 
reproduced only on the sample which does not remove households with top-coded 
house values; instead of dropping the bottom 5% it removes households 
reporting a house value or income of $2 or less (21 households in Tampa 1998 and 
5 in Tampa 2007; between 0 and 28 elsewhere), house values below $1,000 are read as 
thousands of dollars (3 and 11 households in Tampa), the bottom 2% of house values 
and the bottom 1% of incomes are dropped. Eight correlations agree with 
the published values. The remaining four are:

Entry                                             Replicated       Published
Baltimore, rank correlation, 1998                   0.4771            0.38
Baltimore, rank correlation, 2007                   0.4601            0.48
Boston, rank correlation, 1998                      0.3804            0.48
Boston, rank correlation, 2007                      0.3804            0.46

The published Baltimore and Boston values appear to be exchanged: three of
the four replicated values equal the other metropolitan area's published
value. The published Baltimore 2007 value is not reproduced either way.

Smoothing bandwidth: Tampa's 2007 income smoother uses 0.12, which reproduces
the authors' 2010 inputs; the article states 0.11.
