Thursday, May 18, 2006

Maybe we can learn from the world

Recently, as part of on-going work into an inquiry into public support for science and innovation in Australia, the Productivity Commission (PC) released a very large (595 page) econometric analysis to uncover the social rate of return for business R&D expenditure in Australia. You can access that report here.

Now most econometric studies of this kind have found that the social rate of return to R&D is very very high (greater than 100 percent). One might have expected that the PC would find similar things. However, they do not. They find it hard to find a significant impact of domestic R&D expenditure on productivity growth although foreign knowledge does have a large impact.

This sort of result can lead to the following conclusion: perhaps we shouldn't support R&D in Australia at all and instead get knowledge from the rest of the world. After all, the economy has travelled well without it for the last 14 or so years. I call this conclusion: the leech policy (yes, clearly it is not a value free term but it is accurate).

However, in my reading of the report, there is an irony to the PC's findings. In contrast to most other studies of the social rate of return of R&D in small countries, the PC focus exclusively on Australia. Hence, their time series sample size is just 35 years. Now when you have a small sample size, the problem is that it is hard to get precise estimates. The PC realise this but instead of getting more data (you know as others have done) they appear to throw their hands up and claim that we just don't know what the impact is.

But here is the irony. The leech policy is that we should learn exclusively from the rest of the world. However, the PC's own econometric analysis explicitly assumes that they do not believe we can learn anything from the rest of the world to inform us as to the likely social rate of return on R&D! That is, they do not use the data and hence, experience of other countries -- controlling for obvious differences but respecting similarities (e.g., the global economic cycle) -- to assist them in understanding what the R&D does in Australia. That knowledge is useful. In other areas of public policy it goes under the term 'benchmarking.' In macroeconometrics, it is about the law of large numbers. When you have a small sample you add to that other data to improve the numbers.

So my message to the PC is, let's learn from the rest of the world when thinking about the likely impact of our own policies. After all, the knowledge is there for the leeching.


Andrew Leigh said...

Nice post. BTW, is there any good reason the report needed to be more than, say, 50 pages long? I spent 10 minutes looking for their preferred specification so I could estimate the 95% confidence interval, then gave up.

hc said...

Like Andrew I also gave up on this very disappointing PC report. I gave up because it was not very informative. OK it set out to be some applied econometrics and they got non-specific sorts of results but there really doesn't seem to be a lot of thinking in it.

Your post provides csome useful ideas - obviously we need to do our own R&D in specific Australian areas - agriculture and certain types of agricultural protection but we can borrow or leech a lot.

I suspect you leech more effectively if you are engaged in R&D.