Sent three days ago, I have received no reply. If anyone has any
ideas for an explanation of the huge difference in the standard
errors, let me know.
---------------------------------
Dear Dr Lott,
I have compared the corrected tables you posted with the ones from
A&D's response.
Here are the numbers for the crime trend after the laws using the Spline Model
in table 3a for your paper, their correction and your correction.
Plassman Ayres & Lott's
Whitley Donohue's correction
correction
Murder -2.0% -1.0% -1.0%
(0.9%) (1.0%) (0.3%)
Rape -2.4% -1.4% -1.4%
(0.8%) (1.0%) (0.3%)
Robbery -2.0% -0.9% -0.9%
(0.8%) (1.0%) (0.3%)
The second column shows the numbers that Ayres and Donohue got when
they reran your regressions after correcting the coding errors. None
of the numbers are statistically significant any more. This is why
they said that correcting the errors eliminates your results.
The third column shows the numbers from the table you gave after
correcting the coding errors. The reductions in murder, rape and
robbery are all now strongly statistically significant. I guess that
is what you were referring to when you wrote:
"The figures for the paper continue to show clear drops in violent
crime for murder, rape, and robbery."
However, when I compare your correction with Ayres and Donohue's I
find that you get the same values for the size of the effect. The
only difference is that in your correction the standard errors are
dramatically smaller. That is why the values in your correction are
statistically significant and theirs are not.
Could you please explain why the standard errors in your
corrected table are so much smaller than those in your original table
and in A&D's corrected table?
--
Tim
