Suppose I want to learn the effect of land confiscation from church on education investment. I have data on all land expropriated by the government from churches, so I can see which municipalities have churches expropriated. I'm going to assume that the rest of the municipalities have no land owned by churches because I don't have the data on churches' land distribution across municipalities. Can I compare municipalities with churches expropriated and those where churches hold no land if I can show that parallel trend holds for municipalities with churches expropriated and the rest of the municipalities, or do I need to compare municipalities with churches expropriated and those with churches NOT appropriated, something that I cannot do?
$\begingroup$
$\endgroup$
4
-
$\begingroup$ So you’re assuming the municipalities where you have no data of the existing land owned by churches had none of that land expropriated by the government? $\endgroup$– Thomas BilachCommented Mar 20, 2022 at 5:56
-
$\begingroup$ yes, I'm assuming so. Would that bias the results somehow? $\endgroup$– RainroadCommented Mar 20, 2022 at 14:29
-
$\begingroup$ Yes very obviously biased. The "rest" of the municipalities are a mixture of those without any church land and those with church land that hasn't been expropriated. The effect of land confiscation is a comparison between municipalities that have church land but get to keep it and municipalities that have church land which is confiscated. $\endgroup$– dipetkovCommented Mar 20, 2022 at 15:57
-
$\begingroup$ hmm if i can show that those with church land expropriated and those without church land have parallel trends pre-treatment, would that help? $\endgroup$– RainroadCommented Mar 20, 2022 at 16:05
Add a comment
|