Tuesday, April 04, 2006

The Social Security Trustees violate the law by not delivering their actuarial report by April fool's day (an appropriate deadline given their scam of measuring productivity over "complete business cycles" last year which would imply that US citizens would be delighted to learn their pension system is OK if and when there is a recession).

However REPEC delivers their ratings of economists on time. Can't be both quick and reliable and they rate me 803rd (but look what happens when citations are divided by number of authors on the paper -- ouch). Also I am number 21 in Italy.


Update: The trustees are innocent or more exactly non existent. That is there isn't a full complement of trustees, because Bush and the Senate are arguing over whether two current trustees should be re-appointed.

Barry asks in comments "What's wrong with using complete business cycles? It's either that, or adjusting for incompleteness, or producing garbage."

Thanks Barry for a genuine authentic non spam comment.

I claim neither originality nor ability to link to the original arguments made by people like Atrios and Yglesias, but I think they were convincing. The claim is that the trustees switched to measuring productivity growth over complete cycles, because the extremely high productivity growth since the last recession would imply, according to calculations using the older method, that the social security trust fund is solvent without reform.

Thus one of the reasons to object to using complete cycles is that it is a new method, that the change in method was not stressed in the report, that the old method was not presented for comparison and that there was a clear political motive for the change in method. Conceding for the moment that the new method is reasonable (it is not) there is still a very serious problem with allowing policy makers to change methods of data analysis to get the results they want. Thus there is a long tradition of sticking to old methods of presenting data even if researchers think there is a better way. This is a case of rules vs discretion. To support a rule such as statistical methods are to be changed rarely and only after a public debate, one does not have to think that current methods are the best possible. It is enough to assume that flexibility will be abused.

In any case, if a change in assumptions is critical to the conclusions it really must be clearly stated in the executive summary -- a footnote is not enough. Indeed if an arguable assumption is critical (as is always true) sensitivity to assumptions should be discussed. The trustees report 3 estimates for this reason. Their record of forecasting during the Bush presidency is as aweful as the rest of forecasting by Bush appointees. I suspect for the same reason, that the aim of data analysis is to support a politically determined conclusion.

The 2005 trustees report strikes me as an effort at intellectual fraud. A similar effort from an academic would not just cause me to recommend rejection of a paper if I were a referee but also has caused me to conclude that one such academic should not be hired even though I had not read all of his publications.

OK now how should productivity growth be forecast ? Is the complete cycle method a good method ? It has one clear problem that recent data is ignored completely. All data since 2001 aren't considered at all because the current cycle is not complete. This can lead to absurd results. estimates of productivity growth made in 2000 would have ignored all data since 1991. Clearly a forecast which ignores data completely is not optimal.

The complete cycle approach implies that a recession can cause a large increase in estimated productivity growth. Using last years method this will happen when the next recession comes, since suddently 21st century data will be considered. This is clearly not reasonable.

The problem with comparing current productivity to productivity at the last business cycle peak is that productivity at a peak is abnormally high and probably unsustainable (that is really miss measured as increased depreciation due to delayed maintenance is not calculated). Thus estimates of productivity from last peak to now can easily be biased down. This is not a reason to use still lower estimates as the trustees did.

It is possible to deal with the cycle by averaging productivity growth over a longish period. This is the old method. It is crude but clear. It is also possible to deal with the cycle by correcting with measures of intensity of work. In particular the ratio of overtime hours worked by production workers to straight time hours is strongly correlated with cyclical fluctuations in productivity. Thus estimating that relationship (on HP filtered data and no HP does not stand for Hewlett Packard) and using the coefficient to project what productivity would be for a fixed ratio of overtime hours to straight time hours is a reasonable estimate of true long lasting productivity growth.

There are in the literature a gazillion other methods for trend cycle decomposition of everything including productivity growth. Each is imperfect and the debate provides gainful employment to many time series economists. Each is probably better than peak to peak. I haven't surveyed the literature. Does each and every such decomposition shows a productivity speed up since 2000 which is ignored by the trustees ? Does each and every academics trend cycle decomposition give better predictions so far than the trustees method ?

I don't know, but the fact that I think it is possible that the trustees have managed to do worse than people some of whom are completely out of touch with reality and some of whom are principally motivated by the need to do something new to get published leads me to call the trustees 2005 report a scam.

Ahhhh 2005 what a wonderful year. Blogging was all about social security. Like Atrios and Yglesias I miss it. Like Atrios and unlike Yglesias, I have an excuse -- we are economists delighted that someone is interested. Yglesias is twisted.

I am very grateful to Bruce for a geniune authentic non spam comment.

1 comment:

  1. Anonymous6:34 PM

    What's wrong with using complete business cycles? It's either that, or adjusting for incompleteness, or producing garbage.

    If you want an indepth examination of BS, go to Bruce Webb's page:
    http://bruceweb.blogspot.com/

    ReplyDelete