Review of Angus Deaton's the Great Escape

Journal of Economic Literature 2015, 53(1), 102–114 http://dx.doi.org/10.1257/jel.53.1.102

A Review of Angus Deaton’s The Great Escape: Health, Wealth, and the Origins of Inequality†

David N. Weil*

This book explores the relationship between the material standard of living and health, both across countries and over time. Above all, Deaton is interested in the question of whether income growth contributes significantly to better health. His answer is no: saving lives in poor countries is not expensive, and there are many episodes of massive health improvements in the absence of income growth. As an alternative, he argues that the cross-sectional correlation between health and income is induced by variation in institutional quality, while over time, parallel improvements in income and health have been a result of advancing knowledge. (JEL E23, I12, I14, I15, O15, O47)

1. Introduction of GDP per capita, for which we can calculate compound growth rates and, with some- obert Lucas famously wrote of eco- what more difficulty, make comparisons Rnomic growth that once you start think- across countries. Another dimension along ing about it, it is hard to think about anything which there has been enormous change over else. But what is economic growth? One time is human health. Reminding oneself of aspect of growth is change in the goods the ubiquity of premature death, suffering, and services that an economy produces. and disability that characterized the lives of Compared to our ancestors, or to most of the previous generations, and that still charac- other residents of our planet, those of us who terizes the lives of many people in develop- live in developed countries today enjoy the ing countries today, is a good way to get some benefits of a much better consumption bas- perspective on the importance of income as ket: big houses, cars, air conditioning, restau- measured in conventional GDP.1 rant meals, and so on. These are the things Whether one includes health improve- that are captured in conventional measures ment as part of “economic growth,” or

* Brown University and NBER. 1 A brave soul might even ask how these improvements † Go to http://dx.doi.org/10.1257/jel.53.1.102 to visit the compare in terms of their effect on human welfare. See article page and view author disclosure statement(s). Weil (2014) for a discussion of this literature.

102 Weil: A Review of Angus Deaton’s The Great Escape 103 whether one restricts that term to apply only undergraduates and lay people can eas- to income, is just a matter of labeling. But ily understand, but also enlightening and there is a related substantive question: What challenging to even the most experienced do the material standard of living and health scholar. Deaton takes the reader on a richly have to do with each other? The Great detailed tour through a landscape of historical Escape tells the stories of the enormous narrative, science, data from across the world, improvements in health and income that and scholarly debate. And he is a superb have taken place in the last few centuries, guide: erudite, lucid, humane, and witty. and of the huge gaps that persist today, both between and within countries. At the book’s 2. Health and Wealth core is an examination of the causal relationship between income and health, and par- Improvements in health and economic ticularly the question of whether increasing growth (in the narrow sense of rising income in poor countries is a good way to income) have much in common in terms of bring about health improvement. Deaton’s their timing, geographical origins, spread, answer is, in brief, that income affects and underlying causes. The sustained health outcomes much less than you prob- economic growth that began in Europe with ably think. the Industrial Revolution was preceded by This review focuses on the relationship millennia in which the conditions of life between income and health, but it would hardly changed at all, and in which cross- be a disservice to the potential reader to country differences were relatively modest. give the impression that this is all that the The material standard of living has been book is about. In fact, The Great Escape utterly transformed in the countries that encompasses a far broader range of topics. started growing first, and enormous income Deaton embeds both health and wealth in gaps have opened up between countries. a framework of “well-being,” and discusses From its starting point in northwestern how to define and measure this concept. He Europe, the contagion of economic growth also addresses other measurement prob- spread to other parts of Europe and lems, including international comparisons North America in relatively short order, of income and the construction of poverty and later to Japan and South America. In thresholds. Issues outside the usual domain the post–World War II period, growth of economics, both moral (for example, the has spread further still, with late starters extent of a person’s responsibility to help such as The Republic of Korea and China others in need) and philosophical (how putting on great bursts of speed in which to evaluate the welfare consequences of a they grew at rates far greater than anything larger population) are touched upon, as well. experienced by the early starters. Throughout the book, there is a persistent In the case of health, the pattern was sim- focus on the welfare of those worst off. The ilar. There is little evidence of trend change book’s overarching metaphor, and the source in health prior to the middle of the eigh- of its title, is the idea that humanity’s escape teenth century, unless one goes back to the from material deprivation and premature transition from hunting and gathering to death has been tied up with inequality: at agriculture, at which time things got worse. first, only a few escape, and many are left And while there existed health differences behind. among countries prior to industrialization, All of these topics are woven together in with the tropics being particularly unhealthy, an elegant narrative, written at a level that the gap was small in comparison to what was 104 Journal of Economic Literature, Vol. LIII (March 2015) to follow.2 The same countries that led the advanced countries—but this is something pack in terms of income growth saw health that Deaton spends relatively little time on improve first, and as with income, many in his book, since his concern is more with countries that started their health improve- those not lucky enough to be at the frontier. ments later experienced gains at a speed The most notable relationship between far faster than anything the leaders had income and health is in cross-country data. achieved. This was particularly true during Preston (1975) first plotted and interpreted the “international epidemiological transi- the relationship between income and life tion” in the middle of the twentieth century, expectancy. Deaton begins his analysis by when a number of health technologies were showing a plot of the “Preston curve,” and transferred rapidly from the developed to much of his book is devoted to thinking about the developing world (see Acemoglu and what the curve means. The Preston curve Johnson 2007). certainly fits well. Using data from 2010, the There is also a similarly eerie constancy in correlation between the log of GDP per cap- the trend growth rates of income and health ita and life expectancy is 0.84 when countries in the most advanced countries. As discussed are weighted by population (and only slightly by Jones (2002) and Lucas (2000), among lower if they are not weighted). The relation- others, the rate of growth of output per cap- ship is not far from linear, with a doubling ita over the last 140 years in the United States of GDP per capita being associated with an (the world largest rich country for most of the increase in life expectancy of roughly five period) has been nearly constant, at roughly years. An interesting point is that the fit of 2 percent per year. Similarly, in the analysis the Preston curve has been improving over of Oeppen and Vaupel (2002), life expec- recent decades. Using data for 1980 (and tancy in the “best practice” countries (those the same sample as the previous calcula- with the highest life expectancy in the world) tion), the weighted correlation between the has increased linearly since 1840 at a pace of log of GDP per capita and life expectancy is three months per annum, with no sign of a only 0.52. Almost the entire improvement slowdown. In each case, these steady head- in fit can be attributed to a single country, line results reflect ferment beneath the sur- China, which was an enormous outlier (high face. For life expectancy, the sets of diseases life expectancy relative to income) in 1980, that were being controlled, the means used and has since moved back in line. The cor- to conquer them, and the ages at which death relation between income and life expectancy was being rolled back have changed dramati- is echoed with other measures of health, cally. In the case of income, increased invest- including absence of anemia, fraction of ment rates in human and physical capital, babies that are low birth weight, and years massive structural change, the demographic lost to disability (Weil 2014). transition, and a sea-change in the nature of There is also a significant relationship R&D—what Jones calls a series of “grand between income and health that is observ- traverses”—have netted out to constant able within countries. Gwatkin et al. (2007), overall growth. Whether these mysterious pooling data from Demographic and Health linear trends remain in place going forward surveys for fifty-six developing countries, is an issue that will have enormous impact show the wealth gradients for a number of on the welfare of residents of the most health indicators. For example, the under- five mortality rate varies by a factor of almost two, from 135.4 to 73.5 per thousand, moving 2 Depetris-Chauvin and Weil (2013). from the bottom to the top wealth quintile. Weil: A Review of Angus Deaton’s The Great Escape 105

In the United States, mortality probabilities (2) y h . i = β i + ϵi for most demographic groups fall with lifetime income (Cristia 2007). For example, Discussion of the relationship between men aged 50–64 in the top quintile have health and income can then be conducted in mortality probabilities 60 percent lower, and terms of the slope coefficients, as well as the women 40 percent lower, than those in the variances and covariances of the two error bottom quintile. Other measures of health terms. The next two sections of this review are also correlated with income. Height is discuss these structural equations in turn. often used as a summary measure of the cumulative effects of nutrition and health 3. The Effect of Income on Health insults prior to adulthood. As Deaton shows, in developed countries, height has risen on The idea that the Preston curve rep- average roughly ten centimeters since the resents, at least in some significant part, middle of the nineteenth century. Within a causal effect of income on health hangs countries, height is correlated with income, over much of Deaton’s book. It is an issue the relationship being stronger in poorer to which he repeatedly returns, in part countries. In the United States and United because the idea is so seductive: “If the dis- Kingdom, Case and Paxson (2010) estimate eases of poor countries are indeed ‘diseases semielasticities of wages with respect to of poverty’ in the sense that they will van- adult height (controlling only for ethnicity) ish if poverty is reduced, then direct health of between 0.48 percent and 1.1 percent per interventions may be less important than centimeter. Vogl (2012) and Thomas and economic growth. Economic growth would Frankenberg (2002) find semielasticities of be ‘twice blessed’; it would increase material wages with respect to height in centimeters living standards directly and improve health of 2.5 percent in Mexico and 3.1 percent in as a bonus.” (p. 106). This is exactly the error Indonesia. away from which Deaton works hard to steer In addition to this between- and the reader. within-country variation, there is also the Causality running from income to health temporal covariation noted above: after could have manifestations in both the time millennia of stagnation, health and income series and cross-sectional domains. In the began improving at roughly the same time. time domain, the idea was associated origi- All of these data establish that in a statisti- nally with Thomas McKeown, but was pro- cal sense, income and health are certainly pounded most vigorously in recent years by related. The question is, what is the nature Robert Fogel. Looking at the historical expe- of that relationship? riences of developed countries, McKeown Any well-trained undergraduate can famously claimed that the bulk of observed answer that question, at least in a superficial declines in mortality from a number of infec- fashion: The correlation results from income tious diseases took place prior to the deploy- causing health, from health causing income, ment of both medical treatments such as from some other factor(s) causing both, or drugs and vaccines, on the one hand, and from some combination of these three chan- public health measures such as clean water, nels. Putting that idea in a simple frame- on the other. Fogel focused more specifically work, we can think of income (y) and health on nutrition, citing evidence on both caloric (h) being simultaneously determined: intake and changes in body size over time. A central pillar of his analysis is the so-called (1) h y Waaler surface, showing the relationship i = α i + μi 106 Journal of Economic Literature, Vol. LIII (March 2015) between height, weight, and relative mortal- in the relationship between life expectancy ity risk for middle-aged Norwegian men in and income. The increase in life expectancy the 1960s and 70s. Superimposing changes experienced in a country over time can be in average height and weight over the last decomposed into the part due to higher two centuries on the Waaler surface (for income (movement along the Preston curve), example, the French went from roughly 161 and the shift in the curve itself. Preston’s cal- cm and 45 kg in 1705 to 172 cm and 72 kg culation was that less than one quarter of the in 1975) predicts a large drop in mortality. average mortality improvement observed Fogel (1997) calculates that this change between 1930 and 1960 was due to move- alone explains 90 percent of the reduction in ment along the curve, with the remainder French crude death rates between 1785 and due to shifts in the curve. In other words, 1870, and a further 50 percent of the reduc- income gains were not the primary source tion between 1870 and 1975. Floud et al. of health improvements (although it is worth (2011) calculate that over a roughly similar noting that the Fogel/McKeown argument is time period, age-standardized calorie con- primarily meant to apply to a period prior to sumption in France rose by 65 percent. the one examined by Preston). Deaton argues that McKeown was right Deaton’s own narrative of the health about the large mortality declines from improvements over time puts less weight on infectious diseases in early developing coun- nutrition and more on knowledge. In partic- tries preceding the arrival of medical treat- ular, he focuses on the improved understand- ments, but not about their preceding the ing of disease and its control as the driving deployment of public health interventions. force that produced the greatest improve- As an example of the importance of pub- ment in life expectancy. In this story, no sci- lic health interventions, Cutler and Miller entific advance is more important than the (2005) estimate that water filtration and germ theory of disease, which allowed for chlorination alone accounted for 43 percent the introduction of effective public health of the decline in mortality in a sample of infrastructure, particularly clean water, as U.S. cities over the period 1900–1936. While well as numerous other changes in behavior Deaton does not reject the idea that better that reduced the burden of infection. Other living standards promoted better health, he vital pieces of knowledge include: downplays it. Indeed, Deaton’s Great Escape began life as a review of Fogel’s book, The • The discovery that smoking was harm- Escape from Hunger and Premature Death, ful for health, made salient by the U.S. 1700–2100, in the pages of this journal Surgeon General’s report of 1964, (Deaton 2006). Among the pieces of direct although much of the information was evidence he brings to bear, he points out available before then that until 1750, the well-fed British nobil- • The development of a series of cheap, ity had no higher life expectancy than the effective treatment regimens for man- general population. He also argues that the aging high blood pressure, which pro- bigger bodies on which Fogel focuses were duced a massive decline in death rates the result not only of more nutrition, but also from cardiovascular disease in devel- less infectious disease. oped countries after 1970 The most important piece of evidence • Oral rehydration therapy for diarrhea, against the Fogel/McKeown view is the invented in 1973 during a cholera out- one noted by Preston himself in his origi- break in Bangladeshi and Indian refugee nal article: the shifting upward over time camps, and described by The Lancet as Weil: A Review of Angus Deaton’s The Great Escape 107

“potentially the most important medical doctors their first effective treatment for advance of this [the twentieth] century” many bacterial infections. Jayachandran, Lleras-Muney, and Smith (2009) estimate These gains in knowledge often led to syn- that these drugs raised life expectancy in the chronized declines in mortality across sets United States by between 0.4 and 0.8 years of countries that were at different stages of over the period 1937–43. Scientifically, sulfa economic development. drugs were spun off from the dye industry, The knowledge-driven model of health based on synthetic chemicals derived from improvement provides a natural explana- coal tar. The drug development looked much tion for the temporal covariation between like R&D as described by modern growth income and health: growth in health knowl- theorists: an industrialized invention pro- edge had its source in the same scientific cess undertaken by a profit-maximizing revolution, and the attitude of experimenta- firm looking to use both secrecy and pat- tion originating in the Enlightenment, that ents to reap monopoly profits (Lesch 2007). brought forth new technology for producing (Unfortunately for I.G. Farben, the company output. Like two rockslides triggered by the that did the drug development, it turned out footsteps of the same careless mountaineer, that the active component in sulfa drugs was these two intellectual juggernauts proceeded a molecule whose patent, issued in 1909, had in parallel to reshape the human condition. already expired.) Examples like this suggest Of course, there were important intellec- that without ongoing economic growth, the tual links between the development of growth of health knowledge would have health-producing technology and the devel- stalled out at some point in time. opment of output-producing technology, so The knowledge-driven model of health the description of them as parallel landslides improvement provides a natural explanation is something of a caricature of Deaton’s nar- for the upward shifts of the Preston curve, rative. Still, my own feeling is that Deaton especially in the second half of the twen- understates the role of economic growth and tieth century, when the infrastructure for increased income in producing the health spreading medical advances worldwide was knowledge that was so instrumental in saving firmly in place. The converse of this obser- lives. Indeed, to the extent that he stresses vation is that knowledge, at least in its pure income’s role in producing that knowledge, form, is not a good candidate for explaining it is on the demand side: industrialization cross-country differences in health outcomes brought about infection-prone agglomera- today. Premature death in poor countries is tions of people, which made discovery of the largely attributable to the same set of dis- means to fight disease all the more import- eases that were responsible for most of the ant. But economists who study technologi- premature death several centuries ago in cal progress more broadly would emphasize today’s rich countries. There are exceptions two other channels by which income growth to this rule, the most notable being malaria produced health knowledge: by producing (a mostly tropical disease) and HIV (a new better scientific tools, and by raising the disease). But the old-time killers—diarrhea, willingness to pay for health discoveries. lower respiratory infections, tuberculosis, Both of these channels are exemplified in and conditions arising shortly after birth— the discovery of sulfonamide antimicrobial still do a large share of the grisly work. The agents in the 1930s. Though soon overshad- knowledge of how to defeat these conditions owed by antibiotics such as penicillin, “sulfa is widely available, at least in the sense that drugs” were an enormous advance, giving even in the poorest country, there are some 108 Journal of Economic Literature, Vol. LIII (March 2015) professionals who know what needs to be undone by behavioral change and the advent done. The germ theory of disease does not of antiretroviral therapy. But it is hard to see have to be rediscovered. differences in income per capita as having If knowledge does not explain the played a large role in determining which cross-country relationship between income African countries were most afflicted by the and health, what does? It would certainly be scourge. Botswana, long the economic suc- possible for the upward shift in the Preston cess story of the region but with a quarter of curve to be the result of worldwide advance adults infected with HIV, is a case in point. of knowledge, while at the same time the Finally, the United States, with its bloated cross-sectional income–health relationship health spending and mediocre average out- was still due to causality from income to comes, provides yet another piece of evi- health. Once again, however, Deaton puts dence that more money does not always buy little stock in this story. The most import- better health. ant piece of evidence against this view is the If causality from income to health does not observation that many of the health improve- explain cross-country differences in health, ments that could save lives in poor countries and if the pure knowledge story does not are not very expensive. This is made partic- explain it either (in the sense that sufficient ularly clear from close examination of epi- knowledge to drastically reduce mortality sodes of rapid health improvement. Caldwell exists even in very poor countries), we are (1986) studied “mortality breakthroughs,” left with the question of what does explain for example Sri Lanka over the period 1946 mortality gaps. Deaton’s answer is that the to 1953, where life expectancy rose by twelve source of variation is in the application of years. His conclusion was that such episodes knowledge—in particular through govern- are more a matter of political and social will ment actions. Many of the important com- to address health issues than of the avail- ponents of health improvement are public ability of economic resources. Along these goods: “Turning the germ theory into safe lines, Cutler, Deaton, and Lleras-Muney water and sanitation takes time and requires (2006) point out that almost all of China’s both money and state capacity; these were remarkable improvement in infant mortal- not always available a century ago, and in ity took place before economic growth took many parts of the world they are not avail- off in 1980, and similarly that the accelera- able today” (p. 97). tion in growth in India following economic A final question to be addressed regard- reforms in the early 1990s was accompanied ing causal links from income to health is the by a slowdown in the rate of decline in infant importance of this channel in explaining the mortality. Similarly, in Bolivia, Honduras, correlation of these two variables within and Nicaragua, gains in life expectancy on countries. Institutions are not a great can- the order of twenty years took place during didate to explain this relationship, since to periods of modest or even negative income some extent these are the same for all cit- growth (Soares 2007). izens in a country (only to some extent, of The most important negative health shock course—the quality of institutions to which of our time, HIV, also suggests a relatively an individual has access can vary according small role for income’s effect on health. The to income, ethnicity, or location.) Similarly, spread of HIV starting in the 1980s knocked knowledge in its pure form should be equally as many as fifteen years off life expectancy in available to everyone in a country. Deaton’s several sub-Saharan African countries, with answer again focuses on the application some of that damage being subsequently of knowledge, in this case by individuals. Weil: A Review of Angus Deaton’s The Great Escape 109

Education is a key determinant of health sanitation—for which governments need because it allows individuals to apply knowl- money” (p. 32). edge in their own lives (and also allows them As The Great Escape is not a quantita- to know what government could be doing for tive monograph, Deaton does not need to them). This effect is easiest to see when new take a stand on the precise magnitude of the pieces of knowledge become available, and effect of income on health, in other words, then are differentially taken up. For example, the value of the parameter in equation (1). α in the late nineteenth century, prior to the Further, he would probably (and correctly) widespread acceptance of the germ theory of say that the effect of income in health var- disease, the children of doctors and teachers ies with both the institutional setting and the had only slightly lower mortality rates than state of technology. For example, new med- average. By 1925, when knowledge about ical technologies are often expensive, so in a how to control infection was available to be period when such technologies are coming applied, such children had mortality rates on line, the structural effect of income on that were one-third below average (Preston health might be temporarily large. Indeed, and Haines 1991). Similarly, at the time of the process of new cures starting as luxury the Surgeon General’s report in 1964, there goods before moving down market (exam- was little variation in rates of smoking by ples are variolation for smallpox in the eigh- education; by 1987, smoking among male teenth century and antiretroviral drugs for college graduates had fallen to 17 percent, as HIV recently) is part of the story of escape compared to 41 percent among high school and catch-up that characterizes the dynam- dropouts (Preston 1996). The same type of ics of both income and health. And finally, human capital that unlocks access to health there is good reason to think that the struc- knowledge is also rewarded in the labor mar- tural effect of income on health varies with ket, and this omitted variable induces a good the level of income itself: among the very deal of the correlation between health and poor, increases in income facilitate the type income. (Although, obviously there is also of consumption that is health-improving; in a structural effect of income on health out- rich countries, this is probably not the case. comes, via both nutrition and access to med- Summarizing all these effects in a single ical care, which is particularly important in structural parameter would be contrary to poor countries.) the approach of the book. Deaton’s conclusion is that the “diseases of poverty” that are the main killers in today’s 4. The Effect of Health on Income poor countries are overwhelmingly not caused by poverty. As a corollary, the best While Deaton takes seriously the idea that way to reduce the burden of these diseases income affects health, he pays little attention is not to try to eliminate poverty, but to focus to possibility that causality also runs from on health directly. And yet, for all the evi- health to income. It is not clear whether this dence that he provides about the income is because he thinks that the effect is small, not being too important for health, Deaton or because his primary interest is in what does not take the view that income doesn’t determines health rather than in what deter- matter at all. “Income must be important in mines income. some ways and at some times. . . . Income is An extensive literature examines the important in places where improving health effect of health on individual economic out- requires better nutrition—for which people comes, with good reason. Not only is this need money—or cleaner water and better an important question in its own right, with 110 Journal of Economic Literature, Vol. LIII (March 2015)

implications for policy and welfare, but it is frontally and centrally in any comprehensive also a question for which it is possible to find development strategy.” Fogel (1997), exam- good identifying variation. Exogenous dif- ining the historical evolution of body size and ferences among individuals or changes over calorie consumption in the United Kingdom, time in specific health inputs can be matched concludes that over the period 1780–1980, to outcomes such as wages or education to better nutrition raised labor input per work- produce well identified estimates. Examples aged adult by a factor of 1.96. of this approach include Behrman and Empirical attempts to measure the aggre- Rosenzweig (2004), who use variations in gate effect of health on income are rare, birthweight among identical twins to identify in part because of the difficulty of achiev- the effect of fetal nutrition on education and ing identification. The most important wages among adults; Almond (2006), who macro-level paper addressing this issue is shows that individuals exposed to Spanish Acemoglu and Johnson (2007), who use Influenza in utero had lower education cross-country variation in the exogenous attainment and higher rates of disability than component of increased life expectancy surrounding cohorts; and Bleakley (2007), during the international epidemiological who uses geographic variation in hookworm transition to instrument for health improve- prevalence, combined with rapid eradication ments. In their analysis, the effect of health in the American South, to show a long-run on income is negative—that is, countries effect of exposure to the parasite during that experienced larger exogenous health childhood on education and wages. These improvements saw lower gains in income and other studies create the presumption per capita. Acemoglu and Johnson attribute that a country that is in aggregate health- their finding to two factors: first, the underly- ier should, ceteris paribus, be richer. Thus, ing effect of health on individual productivity the parameter in equation (2) should be is small; and second, improvements in health β greater than zero. But microeconomic stud- resulted in large increases in population, the ies, because they only look at one aspect of effects of which undid any positive effect of health at a time, give little insight into just productivity on income per capita.3 Similarly, how large should be when the indicator of Ahuja, Wendell, and Werker (2007) find no β health in equation (2) is a summary measure, evidence of a negative effect of the HIV such as life expectancy. health shock on average income in Africa, A leading proponent of the view that there using circumcision as an instrument. is a large structural effect of health on income My own work (Weil 2007) assesses the is Jeffrey Sachs. For example, in Sachs importance of aggregate health improve- (2001) he writes: “Improving the health and ments by building up from well-identified longevity of the poor is an end in itself, a microeconomic estimates of the effects of fundamental goal of economic development. health improvements on productivity. Unlike But it is also a means to achieving the other Acemoglu and Johnson, I find that the effect development goals relating to poverty reduc- of health is positive, but small. For exam- tion. The linkages of health to poverty reduc- ple, translating the estimates into the units tion and to long-term economic growth are used in the discussion of the Preston curve powerful, much stronger than is generally above, a health improvement that raised understood. The burden of disease in some life expectancy by five years would raise low-income regions, especially sub-Saharan Africa, stands as a stark barrier to economic 3 See Ashraf, Lester, and Weil (2009) for an evaluation growth and therefore must be addressed of this channel. Weil: A Review of Angus Deaton’s The Great Escape 111 labor productivity by 3.6 percent and output list institutions, geography, history, and so per capita in the steady state by the same on. With so many common elements, we amount. Recall that along the Preston curve might not be surprised that the error terms in 2010, an increase in life expectancy of five are so highly correlated. years is associated with a doubling of output However, there is an additional constraint per capita. imposed by Deaton’s theory. Consider a set Ashraf, Lester, and Weil (2009) go beyond of factors, X, that affect both error terms, the static analysis of Weil (2007) to examine and for simplicity let these effects be addi- the dynamic effect of health shocks. Their tive and linear. The equations for the two simulation model allows not only for direct error terms are then effect of health on productivity, but also for several other channels, including the effect of (3) X ϵi = ∑ γ j i, j better health on human capital investment, j the change in population growth triggered by (4) i j X i, j . increased survival (stressed by Acemoglu and μ = ∑j δ Johnson), and the negative response of fertility The correlation between and will be ϵ μ to increased child survival. The effect that they determined by the variances and covariances find is again relatively modest: an increase in of the X terms, as well as the two sets of life expectancy from forty to sixty years would parameters and . Roughly speaking, there γ δ raise GDP per capita in the long run by only are two possible situations under which ϵ 15 percent, and for the first thirty years after and will be very highly correlated. The first μ such an increase, output per capita would be possibility is that a single one of the X vari- lower than if health had not improved. In the ables explains most of the variance in both ϵ context of the two-equation model presented and . In this case, it is easy to see why the μ above, these results imply that causation from two will then be correlated. The other pos- health to income could not be driving much sibility is that there are several X variables of the observed cross-country correlation that contribute to the variances of both and ϵ between the two variables. , and that ratios of the parameters relevant μ to these variables in equation (3) and (4) are 5. Common Determinants of Health roughly equal, that is __γ1 __γ2 . and Income 1 ≈ 2 ≈ ⋯ While I can’t bring δany particularδ data to If the structural effects of income on bear against the second theory, it strikes me health ( in equation 1) and of health on as unlikely. What, other than coincidence, α income ( in equation 2) are both small, then would explain two or more of the important β the observed correlation between health and omitted factors affecting health and income income must result from the errors in these with the same ratio? By contrast, the first the- two equation being highly correlated. This is ory does not require any great coincidence. hardly a radical idea. The first error term ( ) It only requires that a single factor be very μ contains everything other than income that important.4 affects health, while the second ( ) contains ϵ everything other than health that affects 4 Of course, one can imagine alternative stories: for income. Not only is that, in both cases, a lot example, there might be one X variable that explains most of the variance in while a different X variable that explains of things, but there are many of them that ϵ most of the variation in , and these two X variables hap- μ we would expect to enter both equations. pen to be very highly correlated. But possibilities like this Among these omitted variables, one could are not very plausible either. 112 Journal of Economic Literature, Vol. LIII (March 2015)

The natural candidate for that factor governments provide. Finally, there is “the is institutions, and indeed, the reader of irritating but frequently encountered prob- Deaton’s book will already have been primed lem that projects do much better as experi- to accept this view. In the case of health, ments than when rolled out for real” (p. 292). much of his discussion of differences in More significantly, as currently conceived, health outcomes across countries, as well aid is more likely to hurt than help institu- as improvements over time, is centered on tional quality. Aid places a heavy burden on government capacity. Government capacity already stretched government resources. shows up in many dimensions: in the ability And, by eliminating the need to ask citizens to organize large public health projects such for funds to carry out its activities, foreign as clean water and sanitation; in the ability aid eliminates the need for governments to to effectively mount public information cam- obtain the consent or approval of those they paigns to encourage private health behav- rule, leading to long-run damage to the qual- iors such as hand washing and condom use; ity of institutions. “[L]arge inflows of foreign in the regulation of harmful behavior such aid change local politics for the worse and as smoking; in the provision of communi- undercut the institutions needed to foster ty-level health services; and in the ability to long run growth” (p. 294). supervise private health providers. Though More surprisingly, the institutional per- Deaton spends much less time discussing the spective also underlies much of the Deaton’s determinants of income than those of health, discussion of measurement, a topic that when he does address the issue, he places suffuses the book. Recording data—births, institutions in the foreground. And of course, deaths, treatments applied, and so on— the argument that institutions are the dom- requires much the same state capacity as inant determinant of income has recently delivering services. Thus, the same insti- been forcefully laid out by Acemoglu and tutional characteristics that lead to bet- Robinson (2012) in their recent book. ter health and economic outcomes lead to This institution-centric view of the world better data by which these outcomes can finds expression in a number of other places be charted. Measurement is also a channel in The Great Escape. It is present most affecting outcomes: when things are mea- forcefully in the final chapter, which dis- sured, they become, politically and prac- cusses the effects of foreign aid on both tically, easier to act upon. What is counted economic growth in general and on health is a good indicator of how political power is in particular. While Deaton allows that aid distributed. It is no coincidence that the fifty targeted at health outcomes has had good years over which the U.S. government has effects (though he thinks that raising it fur- made no adjustment to its poverty measure ther would not), his view of aid targeted at (beyond indexing to inflation) correspond to achieving economic development can only a period in which the political importance of be described as dismal. The recent vogue the poor and near poor has been declining. for project evaluation and randomized con- The insight that measurement is a useful trolled trials does not impress him. External indicator of institutional quality and an input validity is a serious problem, in his view, into institutional change gives Deaton’s book along with the general equilibrium effects a pleasing unity. It is like a mystery story in on prices that arise when small, successful which we discover at the end that the seem- projects are scaled up. Similarly, there is ingly detached narrator—in this case, the the problem of successful projects divert- available data —has in fact played a major ing resources and undercutting services that part in the underlying action. Weil: A Review of Angus Deaton’s The Great Escape 113

6. Conclusion knowledge and health knowledge advanced together, driven by the underlying advance In Leviathan, Thomas Hobbes character- of science and the spirit of experimentation ized the life of man in the state of nature as born of the Enlightenment. Among individu- “nasty, brutish, and short.” The fact that so als within a country, the correlation between many of us can now lead lives that are both income and health is strongly influenced by long and suffused with material comfort is a human capital, which allows people to apply blessing of which we should be ever mind- available health knowledge to their own ful. And the fact that so many people in the lives, and also, of course, to earn more in the world today still cannot lead lives like this labor market. raises a puzzle deserving of our best intellec- The most compelling correlation between tual efforts. income and health is that observed in coun- The most compelling questions addressed try averages, the so-called Preston curve. in Angus Deaton’s book are about health: Whatever omitted factor drives this correla- Why are poor people—both poor people tion must be extremely powerful, because who live in rich countries and almost every- the correlation is very high. The natural sus- one who lives in poor countries—so much pect, and the one that Deaton points to, is more likely to die than rich people? And, the quality of institutions. Applying available what do the gaps in health have to do with knowledge to stop people from dying from gaps in income? diarrhea and lower respiratory infections The idea that the answer to these ques- requires a government that is responsive and tions might simply be “because higher accountable to all of its people, and that has income makes people healthier” hangs over the capacity to achieve its goals. The same much of The Great Escape. Deaton devotes institutional characteristics that make coun- much of his considerable intellectual fire- tries good at producing output make them power to demonstrating—convincingly, in good at organizing clean water and access my view—that this is not the case. While to medical care. Thus, having started in a income surely has some effect on health out- very different place—with germs and mal- comes via nutrition, access to medical care, nutrition, rather than with parliaments and and the ability of governments to afford pub- contracts—Deaton’s book ends up mak- lic health spending, the salient fact is that ing a powerful contribution to economists’ there are many health improvements that evolving understanding of the importance could be made that are very cheap. of institutions. Though the former focused If causation running from income to his attention on violence committed by health is modest, and if the same is true of men, and the latter on violence commit- causation running in the other direction ted by microbes, Hobbes would approve of (though Deaton pays much less attention to Deaton’s conclusion that a well-functioning this second channel), then the observed cor- government is the actor that allows mankind relation between health and wealth must be to escape the state of misery. due to other factors. Which omitted variables are most important varies with the setting References examined. The contemporaneous advance of Acemoglu, Daron, and Simon Johnson. 2007. “Disease income and health in the leading countries and Development: The Effect of Life Expectancy over the last several centuries (after millennia on Economic Growth.” Journal of Political Economy 115 (6): 925–85. of stagnation) is explained largely by changes Acemoglu, Daron, and James A. Robinson. 2012. Why in knowledge. More specifically, productive Nations Fail: The Origins of Power, Prosperity, and 114 Journal of Economic Literature, Vol. LIII (March 2015)

Poverty. New York: Random House, Crown Business. Floud, Roderick, Robert W. Fogel, Bernard Harris, and Ahuja, Amrita, Brian Wendell, and Eric Werker. 2007. Sok Chul Hong. 2011. The Changing Body: Health, “Male Circumcision and AIDS: The Macroeconomic Nutrition, and Human Development in the Western Impact of a Health Crisis.” Harvard Business School World since 1700. Cambridge and New York: Cam- Working Paper 07-025. bridge University Press. Almond, Douglas. 2006. “Is the 1918 Influenza Pan- Fogel, Robert William. 1997. “New Findings on Secular demic Over? Long-Term Effects of In Utero Influ- Trends in Nutrition and Mortality: Some Implications enza Exposure in the Post-1940 U.S. Population.” for Population Theory.” In Handbook of Population Journal of Political Economy 114 (4): 672–712. and Family Economics, Volume 1A, edited by Mark Ashraf, Quamrul H., Ashley Lester, and David N. Weil. R. Rosenzweig and Oded Stark, 433–81. Amsterdam 2009. “When Does Improving Health Raise GDP?” and Boston: Elsevier, North-Holland. In NBER Macroeconomics Annual 2008, Volume 23, Jayachandran, Seema, Adriana Lleras-Muney, and edited by Daron Acemoglu, Kenneth Rogoff, and Kimberly V. Smith. 2009. “Modern Medicine and Michael Woodford, 157–204. Chicago and London: the 20th Century Decline in Mortality: Evidence on University of Chicago Press. the Impact of Sulfa Drugs.” National Bureau of Eco- Behrman, Jere R., and Mark R. Rosenzweig. 2004. nomic Research Working Paper 15089. “Returns to Birthweight.” Review of Economics and Lesch, John E. 2007. The First Miracle Drugs: How the Statistics 86 (2): 586–601. Sulfa Drugs Transformed Medicine. Oxford and New Bleakley, Hoyt. 2007. “Disease and Development: Evi- York: Oxford University Press. dence from Hookworm Eradication in the American Lucas, Robert E. 2000. “Some Macroeconomics for the South.” Quarterly Journal of Economics 122 (1): 21st Century.” Journal of Economic Perspectives 14 73–117. (1): 159–68. Caldwell, John C. 1986. “Routes to Low Mortality Oeppen, Jim, and James W. Vaupel. 2002. “Broken in Poor Countries.” Population and Development Limits to Life Expectancy.” Science 296 (5570): Review 12 (2): 171–220. 1029–31. Case, Anne, and Christina Paxson. 2010. “Causes and Preston, Samuel H. 1975. “The Changing Relation Consequences of Early-Life Health.” Demography between Mortality and Level of Economic Develop- 47 (1 Supplement): S65–85. ment.” Population Studies 29 (2): 231–48. Cristia, Julian P. 2007. “The Empirical Relationship Preston, Samuel H. 1996. “American Longevity: Past, between Lifetime Earnings and Mortality.” Congres- Present, and Future.” Syracuse University Center for sional Budget Office Working Paper 2007-11. Policy Research Paper 36. Cutler, David, Angus Deaton, and Adriana Lle- Preston, Samuel H., and Michael R. Haines. 1991. ras-Muney. 2006. “The Determinants of Mortality.” Fatal Years: Child Mortality in Late Nineteenth-Cen- Journal of Economic Perspectives 20 (3): 97–120. tury America. Princeton and Oxford: Princeton Uni- Cutler, David, and Grant Miller. 2005. “The Role of versity Press. Public Health Improvements in Health Advances: Sachs, Jeffrey D., ed. 2001. Macroeconomics and The Twentieth-Century United States.” Demogra- Health: Investing in Health for Economic Develop- phy 42 (1): 1–22. ment. Geneva: World Health Organization. Deaton, Angus. 2006. “The Great Escape: A Review of Soares, Rodrigo R. 2007. “On the Determinants of Robert Fogel’s The Escape from Hunger and Prema- Mortality Reductions in the Developing World.” ture Death, 1700–2100.” Journal of Economic Liter- Population and Development Review 33 (2): 247–87. ature 44 (1): 106–14. Thomas, Duncan, and Elizabeth Frankenberg. 2002. Depetris-Chauvin, Emilio, and David N. Weil. 2013. “Health, Nutrition, and Prosperity: A Microeco- “Malaria and Early African Development: Evidence nomic Perspective.” Bulletin of the World Health from the Sickle Cell Trait.” National Bureau of Eco- Organization 80 (2): 106–13. nomic Research Working Paper 19603. Vogl, Tom. 2012. “Height, Skills, and Labor Market Gwatkin, Davidson R., Shea Rutstein, Kiersten John- Outcomes in Mexico.” National Bureau of Economic son, Eldaw Suliman, Adam Wagstaff, and Agbessi Research Working Paper 18318. Amouzou. 2007. “Socio-economic Differences in Weil, David N. 2007. “Accounting for the Effect of Health, Nutrition, and Population within Develop- Health on Economic Growth.” Quarterly Journal of ing Countries.” World Bank Health, Nutrition, and Economics 122 (3): 1265–1306. Population Working Paper 48361. Weil, David N. 2014. “Health and Economic Growth.” Jones, Charles I. 2002. “Sources of U.S. Economic In Handbook of Economic Growth, Volume 2B, edited Growth in a World of Ideas.” American Economic by Philippe Aghion and Steven N. Durlauf, 623–82. Review 92 (1): 220–39. Amsterdam and Boston: Elsevier, North-Holland.