Well, actually, after yet another podcast today from the LSE events series (this is how I stay connected with the metropolis while driving to Maasrticht!), the point was made by Tony Atkinson that very often economists are happy to use and work with data on GDP etc, but very often have no idea about where it comes from. And I agree. And not only from working in Mozambique....
In my first real job in the Scottish Civil Service, I worked on producing the Scottish Input-Output tables, basically the tables behind a social accounting matrix, which are used to reconcile national accounts data. Since the border with England is pretty porous, as you can imagine data on exports and imports from the rest of the UK are pretty ropey. And what about the rest of the data? Well, there were periodic surveys carried out of the expenditure and sales patterns of specific sectors in the economy, but obviously there were too many sectors to do a survey every year, so some sectoral data was just scaled up versions of data collected a while ago. A long while ago in some cases. Scaled up based on what you ask? Based on the sectoral growth rate data, which came from the production surveys, also a sample-based dataset from which GDP growth rates were estimated, themselves based on some pretty subjective weights.
Just from that brief description, you can probably see some room for errors. Low response rates on surveys might cause a few problems. Major changes in inputs and/or sales may make a few changes too which can only be picked up through a survey. But even the overall growth rate of the sectors is based on sample estimates, again subject to selection bias through non-responses. And yet, this is about as good as it gets. While the various data and measures of expenditure, production and incomes were then made to reconcile through "RASing", I remember someone telling me that in Germany (perhaps pre-RAS) they had balanced their tables through a kind of auction in the cafeteria, with those with excess numbers offering to offset with those with deficits in their column. Arbitrary? A little bit. Yes.
So imagine what goes on in a developing country, where the survey design, response rate, data entry, analysis, and balancing are all even more open to error. I recall one IMF mission discovering a mistake in an excel file which meant that in fact GDP growth forecasts (themselves ropey) were being over-estimated due to the transport sector, which in fact was expected to grow at 7 percent not the 17 percent the excel error had led us to believe!
This is something dealt with in a recent blogpost by Shanta Devarajan from the World Bank. Although his focus is on the implications for poverty estimates, his point is not limited to that.
This is something dealt with in a recent blogpost by Shanta Devarajan from the World Bank. Although his focus is on the implications for poverty estimates, his point is not limited to that.
This also has clear implications for tax and tax policy. If your revenue target is estimated as an (increasing) share of GDP, if you increase your growth forecast, you also increase the actual amount of revenue which the tax authorities must collect. What will they do under that pressure? My guess is they might become a bit arbitrary and over-enthusiastic with their revenue collection, and become a little more coercive perhaps...
Or maybe not. One to investigate....
No comments:
Post a Comment