Are Happy Life Years Satisfying?: Fairly, but not Very (Long Post)

For all its merits, some combination of life satisfaction and happiness/affect is insufficient as a measure of well-being. It ignores life expectancy and gives no indication if the current level of well-being is sustainable in the future.

Ruut Veenhoven has tried to overcome the first deficiency with his “Happy Life Years” index that simply multiplies life satisfaction and life expectancy. It’s very straightforward and leads to a somewhat different ordering of countries. No longer is Denmark ranked first, whose residents eat a diet almost as atrocious as that in the U.S. and also are more willing to smoke, but that honor goes to Costa Rica, whose citizens are both very satisfied and live slightly longer.

Yet is this simple multiplying of satisfaction and life expectancy ethical? Is this the proper trade off? I want to know if the U.S. is considering a policy that could reduce (increase) happiness but increase (decrease) life expectancy, what is the appropriate tradeoff? (This would also affect the ordering of countries in future “happy life years” rankings, but I don’t really care which country wins the gold medal in the Olympics of Happiness. My goal is policy-focused, and I want to know how the U.S., using a variant of the U.S. Army slogan, can be the best it can be.)

I have only found one article critiquing Veenhoven’s index (subscription likely required) by Prof. Yew-Kwang Ng, who writes that life satisfaction should be rescaled to represent distance from the midpoint answer of five out of ten. In other words, the life satisfaction index would no longer be from 0 to 10, but instead run from -5 to +5. In contrast, there’s a significant literature critiquing and analyzing the implicit tradeoffs between income, life expectancy and education/literacy in the U.N.’s Human Development Index. Prof. Ng as well as another approach looking at the value of statistical life years (VSLYs) derived mostly from analysis of the income premium demanded to work in more dangerous jobs suggest that the happy life years approach puts about 2-4 times too much emphasis on life expectancy relative to life satisfaction. Yet given the substantial uncertainty in estimating changes in life expectancy and especially life satisfaction with a given policy, the simple happy life year approach is a decent start, even if it may not be perfect. Happy Life Years suggests an increase in life expectancy in the U.S. is good as long as it doesn’t reduce life satisfaction by more than 0.09 points – which I’ll call nine “happiness basis points” in honor of the economic use of the term – while Ng’s version is more restrictive at closer to three basis points.

The best way to start to think about the reasonableness of Happy Life Years is to look at the possible tradeoffs between life satisfaction and life expectancy. For instance, according to Veenhoven, the average person in the U.S. lives about 58 “happy life years,” a combination of a life satisfaction of 7.4/10 and a mean life expectancy of 77.9 years. With Veenhoven’s formula, the U.S. would be the same if it had satisfaction of 7.0 but a life expectancy on par with Japan of 82.4 years. Ng’s formula effectively puts much, much more weight on life satisfaction, so an equal reduction in life satisfaction would require nearly 250% more of a gain in life expectancy to an astronomical 93.5 years. The idea is that a change of a couple of tenths in terms of life satisfaction makes much more of a difference around a small number like two rather than a larger one like seven.

Another method is to try to think through more possible combinations of satisfaction and life expectancy, but it is very difficult to understand just what it means for a country to have different average life satisfaction. For instance, “happy life years” also implies the U.S. would have the same well-being with a life satisfaction of eight and life expectancy of 72.1 years, effectively a combination of the happiness of Switzerland with the life expectancy of Romania – or just to show how quickly health has improved, also the U.S. about 35 years ago. This tradeoff seems reasonable, or at least not awful, but it is difficult to say it is correct. I have great difficulty thinking through what it would mean for the citizens of the U.S. to rate themselves with a life satisfaction of seven compared with eight and so on.

Probably the only definitive problem with happy life years is its implication that low life satisfactions with extremely long lives is good; for instance, a life satisfaction of 5.0 with a life expectancy of over 115 years would equal the U.S.’s happy life years today. However, from a policy standpoint, such an issue at the extremes is meaningless today because there is no possibility of keeping people alive that long. Ng’s -5 to +5 scale appears to only be relevant within a narrow band of life satisfaction , but this is generally fine when trying to analyze policies in the context of the U.S. where dramatic long-term happiness or life satisfaction tradeoffs should be relatively rare.

While this method is illustrative, it does not resolve any ethical question about which is more appropriate. For this, I used existing work on the value of life years and the impact of income on life satisfaction and found a midpoint estimate in between the estimates with the HLY and Ng approaches.

Deaton (2008) and Stevenson and Wolfers (2008) show that using the Gallup World Poll’s Best Possible Life question, each doubling of per-capita GDP is equal to a gain of roughly 0.6 points on an 11-point scale. According to the IMF, the U.S. had a per-capita GDP of about $46,500 in 2009. Aldy and Viscusi (2007) note that there’s quite a bit of disagreement over the value of a statistical life year, but they provide an estimate of $300,000 for the U.S.

In order to make use of all these numbers, assume there is a policy that will increase life satisfaction by increasing income, but it will also decrease average life expectancy by one year. It is as if each person has lost $300,000 in lifetime per-capita GDP. According to standard economic theory, they could be compensated for this loss if they get this additional income while they are still alive. Disregarding discount rates for now, if they live 76.9 years, they would need to make $3,900 extra per year of their life to be compensated. This extra income can be translated into additional life satisfaction of almost exactly 0.05 points per year using the .

But this tradeoff leads to fewer happy life years. Actually, in order to keep the same happy life years, that extra income gain would have needed to add almost twice as much life satisfaction as it likely would. Like Ng’s analysis, the VSLY approach emphasizes changes in life satisfaction more.

This VSLY approach is quite uncertain, but with arguments that it both emphasizes life satisfaction and life expectancy too much, hopefully the issues cancel each other out. It may weight life satisfaction too much because the VSLY could easily be higher. Aldy and Viscusi (2007) find a range between $269,000 and $2.26 million, and their estimate of $300,000 is on the low end of this range, tending to weight an extra year as less valuable. Additionally, my example only looked at an income gain and a life expectancy loss, but the results should change if the situation is reversed. A loss of income should be twice as important as a gain, both due to loss aversion and the structure of the income-life satisfaction relationship. That same amount of income lost should decrease life satisfaction by twice the amount.

On the other hand, the VSLY approach may not give enough weight to life satisfaction. Chances are that absolute income changes do not account for all of the 0.6 points of life satisfaction gain per doubling of per-capita GDP. It ignores issues such as relative income, governance, employment, and air and water quality that if included would almost certainly weaken the income-life satisfaction link. If weaker, more income would be needed to increase life satisfaction by the same amount. Discounting also has an effect. Presumably a policy will change life satisfaction in the future, so the present value of the SLY is less than $300,000. The compensating increase in income begins sooner, but since less is needed, life satisfaction rises by a smaller amount throughout one’s lifetime.

All in all, the relatively mindless multiplying of life satisfaction and life expectancy seems reasonable, although there are hints that it does not emphasize happiness enough. This would be quite remarkable considering, after all, that an academic who has devoted his career to happiness developed it.

Advertisement