Xinhua Coverage and Chinese Foreign Aid

Contributors are:

Think about future direction:

look at the sentiments for news articles covering existing Chinese aid programs
use the salience of countries in Xinhua's news articles to predict aid allocation in the upcoming years (win!)

Hypothesis: Chinese government suppresses news coverage over where they give aids.
Potential explanation: aids are not so popular among Chinese publics; maybe safer to keep it secret rather than promote it.
Conventional studies: more coverage -- > higher salience -- > justification of more aids to the recipient countries from the perspectives of aid-giving developed democratic countries;
China may be an odd case. So we want to loo at whether Chinese aid allocaion can predict a drop in the media coverage in the subsequent years. More aids -- > less coverage

Look at summary statistics of average countries mentioned so we have a baseline of media coverage across aid-recipient countries
Try a bunch of prediction models to increase the accuracy
Think about other predictors we can put in X (more leeways in doing predictions)
A research design that speaks more directly to our theory

look at the distribution of data; outliers and stuff
measure of salience: counts; frequency; ranking
look at the preliminary results in the imputed data and the non-imputed raw data to decide whether we should try another imputation method
try different model specifications

the aid data is also scrapped from the news, but we defend ourselves in saying those data not only use Xinhua source, so the data sources are not completely overlapping.

Lagged instruments -- 2LSL: regress coverage on a bunch of stuff and previous coverage; then regress predicted coverage on aid
Some alternative instrument
GMM
Just using lag of coverage

Between-recipient coverage at a given point in time (subset, African continents for example)
Within-recipient coverage over time (key recipients)