[Previous entry: "dream night 9.26.2007"] [Next entry: "bhurma"]
09/26/2007: "ideas"
1. schooling in capitalist amerikkka
is it possible to see the channels through which schooling affects outcomes later in life? christiano wants to see the effect of social clubs on wages later on by implementing a stuctural analogue to andy's height paper. is it possible to identify whether or not school helps people by making them better worker bees? bowles and gintis didn't do a good job of this in their 2002 paper (which may be why it was relegated to the journal of the sociology of education). even if someone has a better attendance record than someone else, this is plausibly (or even "more plausibly", some argue) due to home conditions or motivation that is entirely derived from reasons not related to having been conditioned to arrive at work on time. selection effects are tough to get around. it seems that schooling in capitalist america used marxism as an identification assumption.
also interesting: how about conspicuous waste aspect of social clubs? conditional on the educational outcomes, background, whatever, more clubs mean you're doing the same as others with less time. and how the fuck does one test this as well?
2. efficacy of political campaigns
if i have information on prior knowledge of voters, how much they were subjected to advertising, posterior knowledge of voters, and votes, i might be able to fit a model of political campaigns as advertisements. identification assumption: more informed voters will be less affected by the informational affect from advertising.
obvious problems:
a. how to voters obtain different priors? i'd need a panel data set to account for individual heterogeneity. outside of this, i'd need to do a keane-wolpin workback to some initial heterogeneity, which isn't appealing b/c i'll have the answer that all the differences are due to differences in initial endowments, and i'm not sure that this will allow me to think of political campaigns much at all, since the initial distribution would eat up most variance.
b. political parties clearly strategize w.r.t. advertising campaigns between states. i don't think i have any way of estimating this right now. would be a much better story if i estimated the equilibrium instead of the one-sided decision theoretical problem. what data would i need for this?
more coming later. i'm just afraid this computer might shit off randomly.