Friday, August 05, 2005

Respect for reality

You've heard that all economists do is assume, assume, assume. Well we do a lot of that, sure. It is essential to our work. As long as the set of assumptions "make sense", and generate hypothesis that are consistent with the data, then we stick to our assumptions. For example, rational choice: as long as people respond to incentives the way rational choice models predict, we feel free to maintain the assumption in most of our work.

Which brings us to data. An economist, like any self-respecting scientist, is primarily interested in being confirmed or refuted by data, rather than simply framing hypotheses. Moreover, analysis of data is the only way to quantify the strength or weakness of hypothesized relationships. For example, for a specific product, sold in a specific market, over a specific period, by how much will quantity purchased fall when price increases, all other factors constant (all changes expressed in percentage terms)? Notice how wordy that is - it's part of having to make everything precise, systematic, and credible.

It's not enough claim that national consumption is a function of national income. One must get data on national consumption, national income, and other related variables, and analyze the data to extract the relationship - if any.

For the macroeconomic forecasting model of Ateneo, (see post below) this is what a team of economists have done. The forecasts we generate are based on a model that has a sound basis in data, i.e. observed economic reality. In fact we have done this for thirteen hypothesized relationships. Then we impose other conditions, which consist of either accounting identities, or economic relationships which have a basis in the literature (i.e. on earlier, observation-based studies). Then we put it all together into a model of the Philippine macroeconomy. So it's not like the numbers are plucked from air.

Right now, I'm thinking of how to confirm whether, statistically speaking, the AMFM generates accurate forecasts. Since we regularly come up with forecasts, ex post (after the fact) comparisons are possible. However you need to wait a while, because the sample is still small (we started in 2002). I'd better dig in the literature for other models that have done this type of study (i.e. ex post forecast accuracy). The "systematic" and "precise" part of economics - of any science, for that matter - takes time and painstaking care. Reality deserves no less.

No comments: