makes a very clear claim. very nice.
also comment, if you have evidence, on the actual effect you're
studying or what evidence would be necessary to study it if you don't have
it in the data.
slight comment below..
Gary
On Tue, 1 May 2007, mastrangelo, paolo wrote:
Thank you all in advance for comments and suggestions!
~~
Do constitutions /really/ cause large governments?: ``quasi-experimental"
evidence using reasonable counterfactuals.
We show that the estimated effects of proportional elections and
presidential regimes on size of government presented in Persson and
Tabellini (2002) result from model dependence. In addition, we highlight
i think you mean 'are model dependent' or are dependent on assumptions
that are not justifiable or equally valid assumptions lead to different
results
the problem of post-treatment bias in their analysis.
We document the
fragility of their matching estimates by estimating a range of propensity
scores, and improve their empirical strategy in three ways, using only
the data in their original analysis. First, we reduce post-treatment bias
by analyzing the data only using covariates plausibly unaffected by the
treatment. Second, we employ alternative matching algorithms to achieve
balance. Third, we estimate treatment effects only for those
counterfactuals that can be addressed given the data. Our estimates of
the effects of individual constitutional rules on the size of government
are substantially different than those in the original paper. Moreover,
we find no evidence of any joint additive effect of electoral and
political systems on size of government.