Corruption and investment

I came across an empirical article that reports that corrupt countries attract more private investment in infrastructure, and another one (also empirical) that reports that companies that appoint former politicians as directors do not benefit from it. Are these two findings compatible? I believe they would hardly be compatible in the same data base. That is, if the measure of corruption is the presence of former politicians in firms (more lobbying than corruption), then if firms enter systematically countries where they need to appoint politicians to obtain profits, they should make profits from this. Alternatively they would be making two mistakes: appointing useless former politicians, and entering countries where they make no profits. I do not exclude it, but it seems difficult. More likely is simply that these two results are compatible because, as they do, they use two totally different data bases. One is a cross-country data base about the institutional determinants of private investment in infrastructures, and the other is a study of political connections in Australian firms. The first finds that firms invest in corrupt countries perhaps because they expect that corruption is a protection against the risk of expropriation in industries that require huge sunk investments. The second is not the first study on political connections that reports a negative effect of political appointments on shareholder value, perhaps because managers or shareholders make behavioural mistakes in the market for directors (like soccer clubs make mistakes in the transfer market). It could be that the corruption that makes infrastructure investment attractive is not achieved through political connections, but through other means. And/or it could be that in countries other than Australia the firms that make political appointments do benefit from it. In other words, one should be very cautious making extrapolations and inferring external validity from the results obtained with very specific measures and samples.