Modelling the economy is easy, right?
Here’s how it works: An increase in output increases employees’ wages (the more you make the more you earn), which leads to more consumption (the more you earn the more you spend). More spending results in lower savings (the more you spend the less you save) which impacts on investment. The change in investment affects interest rates and output, which of course affects exports and imports, which go on to affect wages, spending, saving, output, consumption etc. etc.
The whole thing quickly becomes an intractable mess.
Add to this a workforce consisting of high- and low-skilled workers, education, immigration, government policy and it’s easy to see why many people effectively give up trying to model the economy altogether.
If you’ve ever read an economics journal article, you’ll be familiar with bombastic openings such as “the world consists of two countries with two products. Land is the only thing required to make these products and all people look the same.”
Now, I completely understand that a model has to be simple to be tractable, and that simple models can have a lot to say about the world, but starting by inventing a completely fictitious world with two countries and two products seems to me to be throwing away a lot of what we actually know about the real world, the one we really live in. (And the one we spend a lot of time gathering data about.)
What I think is really needed is not a model which explains how much of good A country i sells to country j, but a description of the world as it currently, really is. A comparison between the shape of one economy, and the shape of another economy, or the same economy ten years ago.
This description of the shape of an economy is just what is done by something called an input-output (IO) table. These things are pretty ancient: they were invented by Wassily Leontief, a German-Russian working at Harvard after the war. He later won a Nobel prize for his work. IO tables record the extent to which parts of the economy (‘sectors’) buy and sell from one another in the production of goods and services. A complete description is built up of where the goods and services are ‘going’ in an economy and, by extension, where and to whom the money flows. It is in this sense that IO tables give a representation of the ‘shape’ of an economy.
It’s little more than a neat way of writing down survey results (“how much, Mr Colgate, did you buy from Johnny’s Mint Supplies this year?”) but it has some powerful analytics associated with it, and is still used today in the calculation of GDP.
But beyond the parochial backwaters of governments’ statistics offices, these things are relatively little used. Economists continue to imagine “a contiuous spectrum of goods from 0 to 1” (if you don’t already know what this means, trust me, don’t ask!) and worlds with two countries, when these descriptions of the shape of economies have existed, and been published, for fifty years.
So why the reticence to replace ingenious fiction with data? Partly, I think, because each country has their own way of dividing up their economy into sectors. This makes comparison extremely difficult. Also, each country does its own surveys in its own time: we may have an IO table for Luxembourg in 2005, but Belgium hasn’t published one since 1993. So what can we do?
Well, around three years ago, the European Union began funding a large project called the World Input-Output Database (WIOD), whose goal was to gather all these economy-shape-descriptions, standardise them to make them comparable and put them all in one place. They also did some whizzy maths to use publicly available time series of things like total output, which are available for all years, to ‘fill in’ the IO tables for the years when none was published.
The project is now finished and the resulting IO tables are available for download for free from the WIOD website.
Here at CASA in London, we’re excited about the possiblities that these IO tables present, but also puzzled by the lack of interest shown by the economics community in this seemingly highly valuable resource.
Now, there’s a world of caveats which I fully understand–the principal one being that much IO analysis requires the assumption of linear production functions–but I for one am going to be doing a lot of work with these over the coming years.
Most excitingly, my colleague and I are going to try to use them to estimate the shape of the economies of developing countries, who haven’t traditionally published the type of data needed to put IO tables together. This could then be used to inform development aid policy at a macro level in a completely new way.
If anyone is already doing this, I’d love to hear from them. If they’re not, expect exciting results in the coming year or so.