Can happiness surveys predict the desire to migrate?

The Gallup organization has found in its surveys that about 15 per cent of the world’s adults would like to move to another country permanently if they had the chance. The rate varies substantially between different parts of the world, with about 38 per cent of adults in Sub-Saharan African countries saying that they would like to move permanently if they were able.

About 80 per cent of those who wish to leave low-income countries would like to go to high-income countries, with the United States the most popular destination in terms of absolute numbers. The desire to move tends to be higher in countries with medium to low human development, according to the UN’s Human Development Index.

Gallup has constructed a Potential Net Migration Index (PNMI) which relates the desire to move into and out of particular countries to their population. The PNMI is the estimated number of adults who would like to move permanently into a country if the opportunity arose, subtracted from the estimated number who would like to move out of it, as a percentage of the total adult population. There are a substantial number of countries with a PNMI score above 100 per cent (which implies that the population would more than double under free migration) and a substantial number with a PNMI score below 50 per cent (which implies that the population would fall below half current levels under free migration).

Can PNMI scores be viewed as indicators of the perceived wellbeing in different societies? Unless we have reason to believe otherwise, it would be reasonable to expect societies with high PNMI scores to have potential to provide high levels of wellbeing and societies with low PNMI scores to provide low levels of wellbeing.

On that basis, we might expect that happiness levels (i.e. indicators of subjective wellbeing) in different countries would predict PNMI scores. If indicators of subjective wellbeing are not good predictors of PNMI scores, we would need to consider the possibility that PNMI scores reflect factors other than wellbeing levels in different countries and/or that wellbeing indicators are biased by cultural or other factors.

The subjective wellbeing indicators that seem most relevant are the Gallup estimates of the percentage of people thriving and suffering in each country. Gallup classifies survey respondents as thriving, struggling or suffering, depending on their evaluations of their current and future lives using the Cantril ladder. The percentages thriving could reasonably be viewed as a ‘pull factor’, encouraging immigration, while the percentages suffering could be viewed as a push factor, encouraging emigration.

I have been able to match the PNMI and life evaluation data for 111 countries. There is some correspondence between countries in which a relatively high proportion of the population is thriving and high PNMI scores. The top 10 countries on both criteria include four countries in common (Sweden, Australia, New Zealand and Canada). Among the countries not included in the top 10 in terms of percentage thriving is Singapore, ranked first in terms of PNMI scores, but with only 34 per cent of the population classified as thriving. Of the countries included in the top 10 in terms of percentage thriving, Brazil had the lowest PNMI score (ranked 59^th ) even though 58 per cent of the population of that country was classified as thriving.

At the other end of the scale, there is no correspondence among the 10 countries with highest levels of suffering and lowest PNMI scores. The 10 countries with lowest PNMI scores are Haiti, Sierra Leone, Zimbabwe, Nigeria, Ethiopia, Liberia, El Salvador, Comoros, Senegal and Ghana (all with scores below -40 per cent). Of the countries included in the top 10 in terms of percentage suffering, Bulgaria (with 45 per cent classified as suffering) had the highest PNMI score (ranked 32^nd i.e. well above Brazil).

For those who are technically minded, the estimated coefficients of a regression analysis explaining PNMI in terms of percentage thriving and percentage suffering had the expected signs, but only the coefficient on the thriving variable was significantly different from zero at the 95 per cent level.

This analysis suggests that happiness levels in different countries are better at predicting the attractiveness of different countries as destinations for migration than at predicting the desire to emigrate. That is consistent with Gallup’s research findings suggesting that people who want to migrate are disproportionately young and educated and more likely to have relatives or friends who have lived in foreign countries.

However, the analysis doesn’t do much to improve my confidence in subjective wellbeing indicators. If 59 per cent of people are thriving in Brazil, why isn’t it a desired destination for migration? Again, if only 34 per cent of the population of Singapore are thriving, why would so many people want to move there?

Why hasn't more use been made of ACSA for measurement of progress?

What is ACSA? It seems to be an acronym for a lot of different things, but the particular ACSA I am referring to is Anamnestic Comparative Self-Assessment. This is an approach to measuring progress which was first suggested by Jan Bernheim about 30 years ago.

The distinctive feature of ACSA is that it asks survey respondents to rate their current wellbeing by comparison with their memory of the best and worst periods of their own lives (with the best period being given a rating of +5 and the worst period being given a rating of -5).

ACSA is an alternative to the conventional question which asks people to rate their current lives using abstract universal anchors. For example, the Cantril scale gives ‘the best possible life’ a rating of 10 and ‘the worst possible life’ a rating of zero.

In terms of measuring progress, ACSA has the merit of using anchors that could reasonably be expected to more stable over time than perceptions of the best possible life. As explained in recent posts (here and here), when people are asked to rate their own lives relative to the best possible life, they are likely to be making that assessment relative to a moving target. If they see their own lives improving in line with their perceptions of the best possible life, they can be expected to give similar ratings to their lives in successive surveys. It should be obvious to everyone that it is a mistake under those circumstances to interpret stable ratings as implying an absence of progress.

A major study comparing results obtained using ACSA and a conventional measure of life satisfaction for a large number of adult hospital patients suggests that ACSA is indeed less subject to biases of various kinds. For example, the results obtained using ACSA were more responsive to a major objective change in the prospects of end-stage liver disease patients following liver transplantation. The conventional measure of life satisfaction did not capture adequately the impact on wellbeing of the life-threatened situation of these patients prior to transplantation, or the fact that transplantation restored them to an almost normal life. The study is reported in Jan Bernheim et al, ‘The potential of anamnestic comparative self-assessment (‘ACSA) to reduce bias in the measurement of subjective well-being’, Journal of Happiness Studies (2006). An ungated article providing a brief discussion of ACSA is available here.

The potential strengths of ACSA relative to conventional measures of life satisfaction are most obvious where the focus of research is on changes in the wellbeing of individuals over time. A potential weakness of ACSA arises in comparing ratings of different individuals, even though research findings suggest that there are common elements in memories of different people concerning the best and worst periods of their lives (the best periods often involve such things as birth of a child and the worst periods such things as unemployment). It seems likely that many people in high-income countries would perceive that the worst periods in their lives were not as bad as those experienced by vast numbers other people in the world. They might also perceive that the best periods of their lives were better than those of people with fewer opportunities.

One possible way to combine the ACSA ratings of different people would be to place them on the same scale as conventional ratings using the Cantril scale.  When I did that for myself, I gave a rating of 8.5 to my current life, a rating of 9.5 to the best period of my life and a rating of 6.0 to the worst period of my life. That implies an ACSA rating of about 2 [10*(8.5-6.0)/(9.5-6.0) – 5]. That is also the ACSA rating I gave to my current life when I asked myself the ACSA question directly. Such introspective exercises don’t necessarily mean much, but this one suggests to me that the underlying concepts used in ACSA are compatible with the Cantril scale. I urge other people to do the exercise to see if they also get sensible ACSA estimates. 

As far as I can see there is no reason why surveys could not ask people to give a rating to the best and worse periods of their own lives on the Cantril scale, immediately after asking them to rate their current lives on that scale. The Cantril scale is far from perfect as a methodology for making interpersonal comparisons of well-being, but the results it provides in that context seem to make more sense than in making comparisons over time. The calculation of ACSA scores in conjunction in longitudinal surveys using the Cantril question provides potential for development of meaningful measures of perceptions of progress.

I don’t know the answer to the question I asked at the beginning of this post. More use should be made of ACSA. It seems to me that including ACSA type questions in longitudinal studies, such as HILDA, has potential to provide useful information.

Why have happiness researchers been so slow to recognize the problems in using surveys to measure progress?

In my last post I pointed out that it is not possible to measure perceptions of progress accurately by using surveys to measure average life satisfaction at different times and then observe to what extent it has risen or fallen. As a result of changing reference norms, people who value an expansion of economic opportunities cannot necessarily be expected to show rising satisfaction with their lives in successive happiness surveys.

I have just discovered that a similar point was made by Francis Heylighten and Jan Bernheim over a decade ago, in an article that seems to have attracted little attention. The authors made the point as follows:

‘Progress could in principle be measured through the change over time of average scores of subjective well-being. However, the existing longitudinal data show little improvement. These survey results are intrinsically insensitive to developments over time, because SWB is typically evaluated relative to proximate, and therefore salient, reference points, such as peers or expectations based on recent experience’. See: Heylighen F. & Bernheim J.(2001): "Measuring Global Progress  Through Subjective Well-Being", in: Proceedings of the III Conference of the ISQOLS.

One of the suggestions that Heylighten and Bernheim made to correct this distortion was to develop a progress indicator from variables that explain a high proportion of cross-country differences in life satisfaction.

If that approach was followed to develop an indicator to measure perceptions of  progress, recent research by John Helliwell and Christopher Barrington-Leigh suggests that the relevant variables to include might be: the log of household income; whether the respondents had relatives or friends to count on if needed; whether the respondents were satisfied with their freedom to choose what to do with their lives; whether corruption was widespread in business and government; and whether they had donated money to a charity in the past month. Their analysis suggests that people in both high-income and low-income countries place about the same value on log income (use of logs allows for declining marginal utility of income) but people in high-income countries place more value on variables other than income. See: ‘Measuring and Understanding Subjective Well-Being’ Canadian Journal of Economics, 43 (3), 2010.

However, I’m not sure that the suggested approach would entirely solve the problem. It seems likely that perceptions that people in low-income countries have of the best possible life would involve a less opulent life-style than the perceptions of people in high-income countries i.e. perceptions of the best possible life rise with increasing wealth (and the marginal utility of income may not decline as rapidly as cross-country regressions seem to imply). In my view, that means it would be preferable to measure perceptions of progress directly using the method suggested in my last post, i.e. by comparing the answers that survey respondents provide when asked to rate their past lives at the same time as their current lives. An even better approach to measurement of progress, as suggested in the book I am writing, would be to identify the characteristics of good societies and measure to what extent societies were adopting those characteristics.

There may be a case to be made that the well-being of people in high-income countries would be higher if the move toward post-materialistic societies was more rapid. But the people who want to make that case should argue it openly, rather than pretending that responses to happiness surveys indicate that most people do not place much value on material progress.

Can happiness surveys help us to measure progress?

I have written about similar questions here before, but I’m not sure that I managed to get the message across to many people. The issues are not all that complex. I probably just need more practice in trying to explain them in simple terms.

The most obvious way to use happiness surveys to measure progress would be to use such surveys to measure average life satisfaction at different times and then observe to what extent it has risen or fallen.

Where is the problem in that? The main problem is that as a result of changing reference norms people who value an expansion of economic opportunities cannot necessarily be expected to show rising satisfaction with their lives in successive surveys.

What do I mean by changing reference norms? When we are asked to rate our satisfaction with life we do so relative to reference norms, such as by comparing our standard of living with that of people we know. Some surveys ask people to rate their lives relative to ‘the best possible life’, but our perceptions of ‘the best possible life’ may also change. For example, education may cause people to expand their horizons so they become less satisfied with a modest standard of living. The same kind of thing can happen when people move from rural to urban areas or obtain access to TV and the internet.  

So, if education tends to make people less satisfied with a modest standard of living, does that mean that they do not value the opportunities that education provides? It obviously doesn’t. Some people make large sacrifices to obtain educational opportunities, so it would be difficult to argue that they don’t value them.

The same reasoning applies to the benefits of technological progress. No-one could expect that people living in 1950 could have felt unhappy or dissatisfied - or sad, or angry even - because they did own personal computers or any of the numerous other amenities of modern life that had not then been invented.

The fact that we do not feel dissatisfied that we do not yet possess the products of future technological progress does not mean that such products will not enhance our future wellbeing and that of our descendants. It just means that we are fortunate to have emotional systems that enable us to give a high rating to our current lives if we can attain a standard of living that is somewhere near the upper bound of what it is currently possible for humans to attain.

Changing reference norms help our emotional system to adapt to changes in external circumstances, but that doesn’t mean that we should allow them to bias our judgements about changes in the quality of our lives.

Derek Bok, former president of Harvard, unwittingly provided a good example of the distorted perception that can arise when we ignore changing reference norms when he wrote:

‘As Americans adapt and yesterday’s luxuries turn into today’s necessities, people are naturally unwilling to give them up, but that does not mean that they are any happier than they were before the process began. Neither does it suggest that the products they yearn for in future will bring them any greater pleasure. What then is the justification for future economic growth?’ See: ‘The Politics of Happiness’, 2010, p 67.

The fallacy in that argument becomes obvious if it is applied to advances in medical science. Does the fact that people in high-income countries have adapted to advances such as the development of antibiotics, and now tend to view them as a normal part of life, mean that such advances have no value? In deciding whether or not we would be happier without advances in medical science, or any other product of technological change, the pertinent question to ask is whether we are obtaining a net benefit from it now. Adaptation may cause us to take for granted the benefits of technological progress, but it is our judgement of where our interests lie that makes us unwilling to give up the those benefits.

One way to eliminate the possible impact of changing reference norms is to ask survey participants to rate their lives at some point in the past (for example, five years ago) at the same time as they are asked to rate their current lives. Responses to such questions enable levels of individual flourishing to be gauged against historical benchmarks to show to what extent people feel that their lives have improved over time. My analysis of such data collected by the World Gallup Poll suggests that people tend to perceive the greatest improvement in their lives over the previous five years in countries where a high percentage of people consider that the national economy is ‘getting better’ and where rates of economic growth have been relatively high.*

Happiness surveys can help us to measure progress if they are used in the right way.

 ----------------------------------------------------------------------------------------------

*The estimated regression equation is as follows:

LIFETODAY = -0.330 + 1.003*PASTLIFE + 0.015*ECONOMY + 0.037*GROWTH + 0.299*IMPGOV

                       0.290)  (0.044)                     (0.003)                      (0.017)                    (0.243)

Adjusted R² = 0.84. The figures in brackets are standard errors of the estimated coefficients.

102 countries were included in the analysis.

LIFETODAY is the average rating ‘life today’ from the Gallup World Poll (around 2008) which asks respondents to rate their current lives on a ladder scale with the ‘best possible life’ as the top rung.

PASTLIFE is the average rating of ‘life five years ago’ from the Gallup World Poll.

ECONOMY is percentage of participants in the Gallup World Poll who perceive that economic conditions in their country are getting better.

GROWTH is the estimated rate of growth in per capita GDP (rgdpl) from Penn World Tables over the preceding five years (2002-07).

IMPGOV is the improvement over the period 2002-07 in the average of the six World Bank Governance Indicators.