Friday, June 1, 2018

Rejoinder To Rauch's Response To Me On The Happiness Curve Overhyped

On May 15 I posted here on "Overhyping the Happiness Curve," a critique of the recent book by Jonathan Rauch, The Happiness Curve: Why Life Gets Better After 50. After it was linked to on Marginal Revolution, author Jonathan Rauch wrote a Response to my post on May 25, which was also linked to on MR. I did not immediately reply as I was in Santa Cruz and did not have my copy of the book. I shall now comment on his reply.  He makes three main points.

The first is that he says I made a false dichotomy between unadjusted and adjusted studies of the age-happiness relation, and that I failed to recognize his discussion on pp. 69-75 of how factors besides age affect happiness.  Certainly he recognizes that other factors impact happiness, even as his focus is on the particular effect of age, which requires focusing on adjusted relations taking account of the impact of those other factors. He cites Blanchflower and Oswald (the apparent originators of the U-curve idea and among its strongest advocates) to the effect that going from age 20 to 45 reduces happiness by as much as a third of what becoming unemployed does, which is a lot.  Some other studies along such lines are cited.  Then a formulation from psychology is brought in that says that happiness is a function of one's "set range" (basically one's general happiness propensity), of circumstances not under one's control, and of things under one's control.  While he cites Martin Seligman, this argument has been widespread in psychology, with a common finding being that 50% is the set range, 10% is circumstances, and 40% is under one's control, although Rauch does not mention this finding.  As it is, he proposes time as a separate variable, although offhand it would seem to fit in the category of those things we cannot control, "circumstances."  (I have some serious questions about what is under our control and what is not, but let us leave that aside.)

This is all well and good, more or less, but it does not deal with the point I made that a quite a few of these other things that can impact happiness tend to be pushing one towards being happier in middle age than not, which can easily lead to a finding that age makes one unhappy in middle age when one pulls out the estimated effects of these other variables.  Again, we are talking about employment, income, marital status, health, and broader social relations, among others.  Rauch simply never notes this, although it is implicit in his contrasting a global finding that shows unadjusted happiness tending to gradually rise from 20 to 64 with a global finding adjusting for non-age factors that shows the U-curve bottoming at 50.  So, yes, he talks about how other factors are important, but he never addresses their own relationship with age, which is what makes this so difficult to parse out.

The second is that he is unhappy that I said he "cherry picked" studies.  Let me withdraw this term as I recognize that it suggests something worse than what he has done.  Bad cherry picking involves taking a container mostly containing raspberries with just a few cherries, and then  picking out only or mostly cherries and declaring, "Ah hah!  I have container full of cherries!"  This is not what he has done.  What he has done is more like taking a container that is 85% cherries and 15% raspberries, and only picking out maybe 1% raspberries so as to be able to say, "While there may be a very few raspberries, it is nearly all cherries!"  Indeed, he accuses me of cherry picking because I mentioned only one study that questioned his findings, Steptoe, Deaton, and Stone in The Lancet.  But as he admits there are indeed more, although he dismisses them as being "few" and "mostly old," and so on.

I agree that the number of studies that find a U-curve with age in high income nations after adjusting for other variables way outnumbers those that do not.  But, in fact Rauch in his response partly mischaracterizes Steptoe et al, which is quite recent.  He says they do no adjusting for other variables, but in fact they do, although not for as full a set as other studies do.  They mostly adjust for health, which is not surprising given that The Lancet is one of the world's leading medical journals.

As it is, some of those other doubting studies are also by some prominent figures in the field, aside from Nobel Prize winner Deaton.  Probably the most prominent of all is a figure Rauch quotes and discusses, while never mentioning his views on this matter of the age relation: the father of happiness economics himself, Richard Easterlin, whose paper on this was published in 2006 in the Proceedings of the National Academy of Sciences. Here is the irony. I largely accept the argument that there is a U-curve with age in high income nations after adjusting or other variables (and in many even with no adjusting).  I have even argued for it with Dick Easterlin in person about  a year and a half ago.  I was struck by how strongly he disputed the proposition, although indeed he was doing so on the basis of the unadjusted relation.  It is probably the case that what really lay behind my complaint on this matter was a feeling that I needed to channel somewhat the views of Easterlin unreported in the book.

The third response involves the matter of country comparisons, with him citing more recent work by Blanchflower and Oswald on groups of countries supposedly showing the U-curve holding after adjustment.  I do not wish to get into a detailed discussion of their findings other than to note that these relations varied quite a bit across the groups and some did not seem to be as strong as they claim in their summaries, although mostly it seems to be there.  This was indeed the issue where I dragged in Steptoe et al, who failed to find U-curves in some places that Rauch claimed to find them (or he did not mention).  On the particular case of Latin America he uses Carol Graham, who I think is more expert on Latin American data than is Angus Deaton, although again these studies vary on which measure of happiness one is using, time periods, precise nations, and which other variables are being adjusted for and how.  The U-curve shows up, so the argument for a "tendency" has genuine basis (which I granted in my original post), but Rauch has tended to downplay the amount of questioning or disagreeing studies, not wishing to distract readers with how unsettled much of this still is.

Let me close with two comments.  The first is that as I said I think he is basically right on a carefully stated tendency to a U-curve in many nations, and the fact that I argued with Dick Easterlin on signals my basic agreement.  So really my main complaint has been that there was insufficient caveating of all this in the book, although there was some ("Don't be Russian!").  I get it that this messes up the story, but I really think it would not have been that hard to recognize more these contrasting studies, even if this was largely confined to the endnotes.  I realize this may make me seem like Will Ferrell in the old Saturday Night Live skits of recording music where he would always say, "More cowbell!"  Here I am saying "More caveating!"

The final point is that I really do think it is an excellent and well written book.  It has a broad perspective, and the discussions in the later parts are wise and useful.  I did not and do not criticize them.  Indeed, while visiting family I discussed this with a daughter who is a psychiatrist in a VA hospital and who is very interested in this topic.  I strongly recommended the book to her.  So, maybe it could have used more cowbell, I mean caveating, but it is largely a worthy and admirable work.

Addendum:  In private communication with the parties mentioned in this post, David Blanchflower informs me that he and Andrew Oswald have convinced Richard Easterlin that there may be a U-curve related to age in some countries, even for unadjusted data, and is for adjusted data in many others.  It is my observation that Easterlin's concern relates to policy, in particular strong evidence that nations that seem to have declining age-happiness curves have poor social safety nets for old people, and that when those are put in place, happiness for older people improves and one starts to see more of a U-curve, with China perhaps being a recent example.

Barkley Rosser
June 1, 2018

No comments: