A sample text widget

Etiam pulvinar consectetur dolor sed malesuada. Ut convallis euismod dolor nec pretium. Nunc ut tristique massa.

Nam sodales mi vitae dolor ullamcorper et vulputate enim accumsan. Morbi orci magna, tincidunt vitae molestie nec, molestie at mi. Nulla nulla lorem, suscipit in posuere in, interdum non magna.

Indirect economic effects & the need for econometrics

According to the Pope and Archbishop Pell promoting condom use in Africa will reduce the incidence of AIDS among those who use them but will encourage promiscuity thereby increasing the overall incidence of AIDS.  Contrary views dominate the literature.

According to some transport economists encouraging more fuel efficient cars will reduce fuel usage over given travel distances but will create incentives for people to drive further since it now costs less thus possibly increasing aggregate fuel usage via a rebound effect.

According to some opposed to harm minimisation policies in relation to drug use such institutions as Safe Injecting Rooms will reduce the probability of fatal overdose deaths among illicit drug users who use these rooms but provide improved incentives to initiate drug use thus potentially increasing drug deaths overall.

Only very recently colleagues of mine have suggested that improving water use efficiencies – reduces water losses – will reduce water availability downstream from the sites of efficiency gain.

Finally, as has long been argued installing seat belts in cars reduces mortality in accidents but induces less safe driving thus increasing accident numbers.

I don’t want to argue the issue of whether these claims are or are not sound.  I only want to point out the pervasiveness of such arguments. All these arguments point to a definite theoretical claim whose validity depends on the empirical size of direct versus indirect effects. There are many other analogous contentions such as the very significant second-best effects of environmental taxes on labour markets that wipe out the case for environmental taxes if the taxed good is a substitute for leisure – this need not be the case if the goods are complements.

I have never pursued empirical econometrics preferring theory. But we need theory to identify plausible hypotheses and need econometrics to choose between seemingly equally plausible a priori theoretical predictions in many important areas of applied economics.  You cannot ‘think’ you way around such important policy issues – evidence is needed.

10 comments to Indirect economic effects & the need for econometrics

  • conrad

    There’s a great Darwin quote on this:

    “About 30 years ago there was much talk that geologists ought only to observe and not theorise; and I well remember someone saying that at this rate a man might as well go into a gravel pit and count the pebbles and describe the colours. How odd it is that anyone should not see that all observation must be for or against some view if it is to be of any service!” [September 18. 1861]

    It think the real problem is that theorizing for many is hard (perhaps it’s not bad in economics, but where I work, it’s data-without-theory central. The number of PhDs I see like this is astounding). I also think it’s hard for the general public to understand even the simplest of theories.

  • Uncle Milton

    The problem with econometrics is that it isn’t a powerful enough tool to conclusively determine which theories are wrong. Even best practice econometrics leaves so much discretion – choice of specification, choice of econometric method, choice of data set – that any half competent economist can find econometric evidence to support their pet theory. The result is that no one takes econometric evidence seriously. Sure, people who devise really smart new econometric methods get taken seriously, and some of them win Nobel Prizes, but econometrics just isn’t a useful way of decising this or that theory is wrong and should be discarded,

    Harry, in your career, how many times have you, as a theorist, decided against pursuing a theory because of prior contrary econometric evidence?

  • hc

    I agree that the evidence is often ambiguous but so too is the theory. I think progress is being made although slowly. I have recently been working on ‘double dividend’ aspects of environmental taxes on transport. The standard wisdom is that these taxes place huge costs on labour markets – so much so that the double dividend of using these taxes to cut distortionary labour may in fact be negative. This assumes that the demand for travel and the demand for leisure are substitutes. Recent evidence suggests they might be complements. This is useful information though I agree the debate is still unsettled.

    The evidence on traffic accidents and seatbelts has converged a bit. The original Peltzman type arguments that seatbelts cause accidents has been softened and the effect he identified have been shown to be low to moderate. There might be more accidents but there are less severe accidents.

    Generally your comment makes much sense. Thinking back through the years on issues that were supposed to hinge on an empirical regularity (e.g. the Keynesian/Monetarist debate) the issue always remained inconclusive.

    I refuse to attend seminars on empirical finance – the issues there are even worse than mainstream economics.

  • Mark U

    Uncle Milton says “The problem with econometrics is that it isn’t a powerful enough tool to conclusively determine which theories are wrong. Even best practice econometrics leaves so much discretion – choice of specification, choice of econometric method, choice of data set….”

    This statement may be true, but to then argue that econometrics should be discarded is like throwing the baby out with the bathwater. How exactly can you go about resolving competing theories without applying some empirical approach? You may never be able to disprove a theory, but over time the weight of empirical evidence may begin to fall one way or another. For example, despite Card and Krueger, the preponderance of empirical evidence on the impact of minimum wages on employment shows a negative relationship.

  • Uncle Milton

    Mark, of course some empirical approach is necessary, but the question is what value is brought by econometrics. Perhaps it is helpful in resolving very narrow, very tightly defined questions, like how big are certain labour supply elasticities. But these questions take the theory as given. The only issue then is the size of parameters.

    But on resolving big picture theoretical debates, like the Keynesian/Moneratist debates, econometrics has proven to be worse than useless.

  • Mark U

    Uncle Milton

    So what has proved useful in resolving these debates? I am also not clear why you claim it to have been “worse than useless”.

  • Uncle Milton

    Mark, experience and observation resolved these debates. The experience of the 70 and 80s showed very clearly that monetary policy has powerful effects on the economy; that there is no long run trade off between inflation and unemployment, and that the growth in money supply is not an exogenous cause of inflation. Thirty years of econometrics failed to show any of these things. The econometrics was worse than useless because it was a distraction that led to entrenched and stupid policy positioning. It also hindered the development of our understanding of the economy.

  • derrida derider

    I think the real problem is that theorizing for many is hard (perhaps it’s not bad in economics, but where I work, it’s data-without-theory central. The number of PhDs I see like this is astounding.) – conrad

    In fact the problem is the exact reverse in economics. Its really, really easy to think up all sort of theories – why demand curves might slope upwards (conspicuous consumption, crowding out of intrinsic motivations), why monopoly might foster innovation (Schumpeter plus economies of scale in R&D), why fiscal deficits might reduce consumption (Ricardian equivalence plus uncertainty), why infant industry policy might work (increasing returns leading to path dependence), etc. The hard bit is rigorously testing these with data, because the world is a giant uncontrolled experiment where every event is overdetermined.

    Had I been present at the creation, I would have advised God to set up a series of blind (God being God, double blind is a bit hard to arrange) randomised controlled trials.

  • hc

    This is a valuable discussion. I’ll stick to my guns and assert the importance of econometrics and of testing theory with data but I have sympathies with the views of Uncle Milton. So often econometrics doesn’t deliver much. That’s not necessarily a criticism of econometrics but almost certainly reflects the way things are.

    No-one has got specific with the issues I raised and how econometrics has or hasn’t helped narrow down the range of what is reasonable.

  • Mark U

    Econometrics is a relatively young discipline. There have been a advances in econometrics in the 80s and 90s that have led to better guidance on how to undertake good econometric practice (eg. cointegration) and limited the ability of researchers to “massage” results. However, the main obstacle remains the reliability of data and the inability to perform controlled experiments.

    I realise this does not address your question, Harry. My intuition is that direct effects will generally be larger than indirect effects. For example, if you are in an accident and wearing a seatbelt you are far more likely to survive than if you are not wearing seatbelt, and this would seem to be sufficient to outweigh the higher probability of an accident. But I am not familiar enough with the literature in these areas to be able to assess the empirical evidence.