Tag Archives: Development

The Middle Income Trap

Development State Evolving: Japan’s Graduation from a Middle Income Country

By Tetsuji Okazaki (University of Tokyo)

Abstract: This paper reexamines the industrial policy in postwar Japan from perspectives of the literature on a “development state” and a “middle income trap”. Japan transited from a middle income country to a high income country in the period from the 1950s to the 1970s. This process was characterized by a large structural change, such as resource reallocation from the primary industry to the secondary and the tertiary industries as well as resource reallocation within the secondary industry. Transition to a high income country is a challenging task for a middle income country. With respect to Japan, the industrial policy played a positive role in the transition. This was achieved by interactions between MITI and other related actors, who constrained and corrected MITI’s attempts of excess intervention.

URL: https://EconPapers.repec.org/RePEc:tky:fseres:2017cf1063

Distributed by NEP-HIS on 2017‒09‒03

Review by: Joyman Lee (University College London)


Students of modern Japanese economic history are familiar with the work of Chalmers Johnson (1982) on the Ministry of International Trade and Industry (MITI). In that work Johnson argued that MITI was the leading state actor in Japan’s economic miracle, playing a vital coordinating role between policymakers and the private sector. Johnson’s emphasis on the role of the state in the East Asian experience has triggered similar studies on the development state in Korea (Alice Amsden) and Taiwan (Robert Wade).

As Okazaki notes, the emergence of newly industrialising economies facing the challenges of globalisation and democratisation has led to a renewed interest in the development state. Okazaki argues that rather than constituting a static set of policies, Japan’s developmental state was highly dynamic and adaptive, echoing Douglass North’s idea of “adaptive efficiency” (North 2005). Significantly, this ceased to be the case in Japan after the 1990s. A second strand of literature that informs the paper is the idea of the “middle income trap” (Gill and Kharas 2007), which highlights a particularly challenging transition which middle-income economies face, as the policies that have fueled the initial stages of growth are no longer appropriate for continued growth. The idea has gained considerable traction among commentators in China.

china middle income

The fear of the “middle income trap” has been particularly acute in China.

Okazaki’s paper shows that Japan’s successful voyage through the “trap” was partly facilitated by its success in resource allocation across industries, in addition to well-known increases in the intra-sector productivity. Between 1955 and 1975, Okazaki attributes 29% of the increases in labour productivity to resource allocation, which he stresses was “substantial” (p. 4).

Okazaki traces the evolution of policies from the American occupation period, when U.S. advisor Joseph Dodge initiated the abolition of strict wartime controls. A 1953 government report was followed by the Five Year Plan of 1955, which highlighted the need to transition from light to heavy industries. MITI was formed in 1949 to pursue the policy of “industrial rationalization”. Formal economic controls were replaced by a portfolio of public financial institutions, including the Japan Development Bank (1951), tax relief, and foreign exchange allocation, and a central coordinating Council for Industrial Reorganisationolic . The government promoted new sectors, particularly the machinery and the automobile industries within it, which included the use of cultural strategies such as a campaign to promote the purchase of domestic cars at the same time as regulating foreign direct investment (1952) and curtailing the foreign exchange available for car imports (1954). The government also actively implemented policies concerning the automobile parts industry, which was quite atypical given the miscellaneous and low tech nature of that sector.

japan car industry

A Toyota factory in 1948. MITI’s policy in supporting the automobile parts industry which supplied major manufacturers such as Toyota was particularly distinctive.

At the same time as developing the domestic economy, MITI also foresaw foreign pressure on trade liberalisation, and formed a committee to formulate its strategy in 1959. While the ministry remained ambivalent with respect to its effects, it nonetheless adopted a sequential programme of liberalisation that was intertwined with plans to upgrade the industrial infrastructure. The high level of alert to likely external treasures had a direct effect on the government’s sector-specific strategies, e.g. to focus on passenger cars in the automobile sector. However, MITI’s more radical plans to consolidate the industry by policy intervention were not adopted, and instead the government aided the industry through JDB loans and low interest loans to small and medium-sized suppliers. MITI also successfully resisted IMF pressures to remove the industry from the foreign exchange system until Japan was well established in the world market (1963). Meanwhile, the government conceded that the coal industry would be uncompetitive and adopted a programme of gradual phasing out.


Okazaki’s study provides a timely, quantitative and authoritative review on an important and relatively understudied topic (given the acceptance of Johnson’s view as orthodoxy among historians) by one of Japan’s leading economic historians, whose trans-war perspective is particularly useful in teasing out more subtle changes amidst MITI’s strong posture towards industrial policy. As Okazaki observes,  the difficulties that middle-income economies face are acute, as “one of the difficulties that middle income countries face is that they should compete with low income countries in the markets of labor-intensive industries as well as with high income countries in the markets of capital and technology intensive industries” (Bulman 2017). In this context, Japan’s success appears remarkable, perhaps no less than the historiographically well-recognised significance of Japan’s Meiji-period Westernisation.

However, the complexity of policies required for breaking the “middle-income trap” in Japan’s case may not provide much comfort for middle-income economies currently facing the challenge. Although Japan rejected centralised state controls, the Japanese example appears to require a complex set of policies that presupposes a high degree of political cohesion and long-range economic planning, which is often difficult in many middle income economies given various political and social challenges. It also requires a state that is highly persuasive to the populace with respect to its vision for economic development. These factors appear to mark Japan out as an exception rather than an example that can be easily perceived as immediately relevant by many developing countries.

Perhaps the most avid student of Japan’s experiences will be China, which possesses a similar state capacity for a coordinated industrial policy and a qualified commitment to the market, even if it may not enjoy the same degree of social cohesion. This likely Chinese interest may explain the timing of Okazaki’s paper. However, the requirement of a strong state may produce perverse incentives for middle-income countries to maintain authoritarian systems of government (even though Japan was not classically authoritarian in that period in its history), and reminds us of unresolved tensions between economic development and democratisation.

Additional References

Alice, A, 1992. Asia’s Next Giant: South Korea and Late Industrialization. New York, NY: Oxford University Press.

Bulman, D, Eden, M, Nguyen, H, 2017. “Transition from Low-Income Growth to High-Income Growth: Is there a Middle-Income Trap ?” Journal of the Asian Pacific Economy, 22(1): 5-28.

Gill, I, Kharas, H, 2007. An East Asian Renaissance: Idea for Economic Growth. Washington DC: The World Bank.

Johnson, C, 1982. MITI and the Japanese Miracle: The Growth of Industrial Policy, 1925-1975. Stanford, CA: Stanford University Press.

North, D, 2005. Understanding the Process of Economic Change. Princeton, NJ: Princeton University Press.

Wade, R, 2003. Governing the Market: Economic Theory and the Role of Government in East Asian Industrialization. Princeton, NJ: Princeton University Press.


Assessing the Determinants of Economic Growth in South East Asia

The Historical State, Local Collective Action, and Economic Development in Vietnam

By Melissa Dell (Harvard University), Nathaniel Lane (Stockholm University), Pablo Querubin (New York University)

Abstract – This study examines how the historical state conditions long-run development, using Vietnam as a laboratory. Northern Vietnam (Dai Viet) was ruled by a strong centralized state in which the village was the fundamental administrative unit. Southern Vietnam was a peripheral tributary of the Khmer (Cambodian) Empire, which followed a patron-client model with weaker, more personalized power relations and no village intermediation. Using a regression discontinuity design across the Dai Viet-Khmer boundary, the study shows that areas historically under a strong state have higher living standards today and better economic outcomes over the past 150 years. Rich historical data document that in villages with a strong historical state, citizens have been better able to organize for public goods and redistribution through civil society and local government. This suggests that the strong historical state crowded in village-level collective action and that these norms persisted long after the original state disappeared.

URL: http://econpapers.repec.org/paper/nbrnberwo/23208.htm

Circulated by nep-his on 2017/03/19

Review by Fernando Arteaga (George Mason University)

What was the impact of the ancient Vietnamese Dai Viet empire in promoting long-term economic development? That is the main question the authors try to assess. Their inquiry is embedded within the now large literature on the importance of culture and institutions, as deep determinants of growth. The contribution the paper makes is, however, not restricted to adding one more piece of evidence in favor of it, but, more importantly, in providing empirical support for a specific transmission channel: how state capacity can be built through time via the fostering of local self-organization capabilities.

The paper’s main story builds on the idea that two distinct meta-societies existed within East Asia, and idea around which, by the way, there is general agreement. One of these societies based on Chinese precepts, prevalent in the Northeastern region; and other spread in the Southeast throughout the Indian Ocean.  Societies of the former category were historically constituted around a sort of Weberian professional bureaucracy that consolidated the working of a central state. The latter depended more on informal networking mechanisms among local elites to survive, and hence, tended to promote hierarchical patriarchal relationships.

Today’s Socialist Republic of Vietnam (henceforth Vietnam) is an interesting case study precisely because it arose out of the union of those two distinct cultures. The northern part, the Dai Viet, is an example of a Sino-style state, while the southern part of Vietnam (initially part of the Champa State and later as part of the larger Khmer Empire) resulted from a Indo-style society.  Figure 1 below offers map of present day Vietnam aligned with the size of the historical Dai Viet empire. Figure 1 suggests the Dai Viet expanded southwards through time but ended up establishing its final frontier in 1698 (orange color). It is this border the authors think provides a natural experiment that allows a clean regression discontinuity (RD) strategy that permits the disentanglement of the effect of being part of a bureaucratized state vis a vis a patriarchal state.


Figure 1: Dai Viet Historical Boundaries (Dell et al., 2017)

The use of the RD design is appropriate, the authors argue, because the chosen border resulted from exogenous contingencies that do not reflect any difference in future economic potential. The 1698 demarcation was settled on the ridges of a river, but there was nothing else particular to it that made that boundary preferable to other potential borders. The Dai Viet stopped its expansion because of constrains imposed by a local civil war (something that has nothing to do with the river itself). Moreover, the environmental characteristics of both sides of the river are almost identical (or vary smoothly), so there is no important geographical difference either. The only thing that changes abruptly is that on the east shore of the 1698 border, Dai Viet settlers occupied and controlled the land, while Khmer villagers occupied and controlled the land to the west of the river. Another possible counterargument to the use of the 1698 border as a natural experiment is the relevance of migration: if settlers moved across villages (at any time after the establishment of the original border), then the boundary becomes inconsequential. The authors argue that, even though they do not have historical data to control for it, there is qualitative evidence that refers to negative attitudes towards outsiders within the villages, which constitutes an important constraint to any major migratory flow. Today, both sides are part of Vietnam. It is then possible to assess if Die Viet institutions still exert some type of effect in current economic outcomes.

Figure 2 portraits the main outcome of the paper. Using household expenditure data from recent censuses (2002-2012), the authors find that today, villages situated along the historical Die Viet side of the border earn a third more than those communities that are situated on the historical Khmer side (Within the figure, the darker the zone depict lower earnings).


Figure 2a: Household Consumption, RD Graph (Dell et al., 2017)

The authors, however, not content with establishing the effects on current outcomes, look for historical evidence too. They collect data from different periods of Vietnamese history: 1878-1921 for the French Colonization, 1969-1973 for the South Vietnam State, and 1975-1985 for the early Communist Period; and find that the pattern is persistent through time: The Diet Viet zone is, in general, more developed than the Khmer side.

How can these results be interpreted?  The income differences must be due to the Die Viet heritage of greater state capacity that acted through local community self-organization that made them more co-operative and facilitated the resolution of local collective action problems. To test whether this transmission channel matters, the authors looked for data on social capital. Their main sources were the surveys and census of the South Vietnamese period. What they find corroborates their story: villagers on the Diet Viet side were more prone to participate in community activities, to collect more taxes (that at the time were local responsibility, not provincial), to have greater access to public goods (health, school and law enforcement), to be skeptical of central government in favor of local, and to give more to charity.


All in all, the authors do a thorough job in assessing the robustness of their main story. They control for several of potential alternative stories and/or possible variables that could affect the results and mechanisms.  Any critique of it may sound redundant or unreachable.  Yet, I would point to three different aspects that may be important.

First, and perhaps most importantly, I would stress that although the argument makes sense, the narrative is unclear as to how specifically the Dai Viet, which supposedly was a centralized bureaucratized state, fostered local governance. As the authors mention in the introduction, the literature on social capital is ambivalent on its effects on economic outcomes. As it is, the paper’s contribution is the finding of empirical evidence on the presence of a particular transmission channel (from state to local governance), but without a clear model and/or an analytical narrative, we are left in the dark about how explicitly this mechanism worked its way throughout society.

Second, and pushing the level of pickiness even further, one can always speak of a potential omitted variable bias. I must ask then: what about genes? The authors minimize ethnic fragmentation as a problem because they find the studied area is cataloged as being almost entirely composed of homogeneously ethnic Vietnamese. The problem is that censuses and surveys may under-report true ethnicity, and cannot capture genetic differences at all. By the authors’ own account, we are told the Diet Viet State originated as, and remained for a long time, Chinese. Moreover, as Tran (1993) attests, Chinese ethnicity may conflate the results of the paper in other several ways:

  • the largest Chinese migration occurred between the late 17th century and early 19th century, just at the time that the Dai Viet-Khmer border was being established;
  • The Chinese settled mostly in southern Vietnam, the part that the authors use as study case;
  • Chinese early importance resided precisely in that they helped establish new villages and trade outposts. They (not merely the Diet Viet heritage) helped to build local governance structures.

If ethnicity has been underreported and/or Chinese genetics matter in fostering economic development in any way (as suggested by Ashraf-Galor, 20013a, 2013b) then the interpretation of the paper could dramatically change: the importance of the Dai Viet state would be downplayed in favor of just being more ethnic/genetic Chinese. After all, it is known that there is a correlation between having larger ethnic Chinese minority and larger economic growth (Priebe and Rudulf, 2015).

Third, related to the last point: one would expect that given the importance of the result – the long-term reach of Diet Viet institutions–, its impact would feel more broadly across all the territory, not only in the immediate zones of the frontier which were the last to be incorporated into the state.  Figure 3, for example, shows the level of poverty in Vietnam (Epprecht-Heinmann,2004). It is visible that the area under study (along the last border of the historical Diet Viet) has the lowest share of poverty in the whole country. The immediate area to the left (which coincides with the area that historically belonged to the Khmer Empire) is poorer indeed. But the differences are minor if we compare them to the rest of current Vietnam, which belonged almost entirely to the Diet Viet, and has the largest poorer areas.  The RD design may be identifying a non-observable variable that is concentrated in the southern part (like ethnicity or/and genes) and is not broadly distributed across the rest of Vietnam.


Figure 3: Incidence of Poverty in Vietnam (Epprecht-Heinmann, 2004: 155).

Additional References

Ashraf, Q., Galor, O., 2013a. Genetic Diversity and the Origins of Cultural Fragmentation. The American Economic Review: Papers on Proceedings 103, 528–533.

Ashraf, Q., Galor, O., 2013b. The “Out of Africa” Hypothesis, Human Genetic Diversity, and Comparative Economic Development. American Economic Review 103, 1–46.

Epprecht, M., Heinemann, A., 2004. Socioeconomic Atlas of Vietnam: A depiction of the 1999 Population and Housing Census. Swiss National Centre of Competence in Research, Bern.

Priebe, J., Rudolf, R., 2015. Does the Chinese Diaspora Speed Up Growth in Host Countries? World Development 76, 249–262.

Trần, K., 1993. The Ethnic Chinese and Economic Development in Vietnam. Institute of Southeast Asian Studies, Singapore.

(Spoiler Alert) Game of Science: Higher life expectancy does not cause Economic Growth

Disease and Development: A Reply to Bloom, Canning, and Fink

By Daron Acemoglu and Simon Johnson (both MIT)

URL: http://d.repec.org/n?u=RePEc:nbr:nberwo:20064&r=his


Bloom, Canning, and Fink (2014) argue that the results in Acemoglu and Johnson (2006, 2007) are not robust because initial level of life expectancy (in 1940) should be included in our regressions of changes in GDP per capita on changes in life expectancy. We assess their claims controlling for potential lagged effects of initial life expectancy using data from 1900, employing a nonlinear estimator suggested by their framework, and using information from microeconomic estimates on the effects of improving health. There is no evidence for a positive effect of life expectancy on GDP per capita in this important historical episode.

Reviewed by Sebastian Fleitas

 “The game of science is, in principle, without end.   He who decides one day that scientific statements do not call for any further test, and that they can be regarded as finally verified, retires from the game.” 

The Logic of Scientific Discovery, Karl Popper, 1934.

Bill Gates' Infographics

Bill Gates’s Infographic.

Not a long time ago, on April 25, Bill Gates posted an infographic on his blog revealing which is the world’s deadliest animal. Sharks, bugs, snakes and many very scary animals are not even close. The mosquito has the first place by far. They carry terrible diseases, including malaria, which kills more than 600,000 people every year. This infographic is just a reminder of how important it is to improve health around the world. Better health conditions could make millions of people live longer and better lives. But will these better health conditions (and a longer life expectancy) actually cause economic growth? Cross-country regression studies show a strong correlation between measures of health and both the level of economic development and recent economic growth. But, as we know, correlation does not imply causation.

What Acemoglu and Johnson (AJ hereafter) do in their 2014 paper (NEP-HIS 2014-05-17) is just to play the Game of Science. AJ (2007) argue that life expectancy does not cause economic growth and that previous studies had not established a causal effect of health and disease environments on economic growth. Since countries suffering from short life expectancy are also disadvantaged in other ways that are correlated with their poor health outcomes, previous macro studies may be capturing the negative effects of these other unobservable disadvantages. To address this identification problem, AJ (2007) used an instrument for the life expectancy: medical advances that occur at the health frontier, interacted with variation in the prevalence of diseases across the world, used together to construct a predicted mortality variable. The adoption of new medical practices is clearly endogenous, but the authors argue that the technology at the frontier is potentially exogenous. Since there was variation across countries in the prevalence of different diseases, the timing of new medicine advances has a different effect on the predicted mortality for different countries. In other words, the predicted mortality variable satisfies the requirements of a good instrument: it is correlated with the life expectancy in the country, but it is arguably not correlated with other unobservables that determine growth that may be changing at the same time in a country.

Dr. Jonas Salk and Dr. Albert Sabin developed two different polio vaccines that have pretty much  almost eradicated polio from the world.

Dr. Jonas Salk and Dr. Albert Sabin developed two different polio vaccines that have pretty much almost eradicated polio from the world.

Bloom et al. (2013, hereafter BCF) disagree with AJ’s strategy and conclusions. In their paper, which earlier appeared as an NBER working paper, they argue that the problem with AJ’s instrument is that it assumes the predicted mortality to be exogenous and not affected by contemporaneous income shocks. In other words, it implies that the initial mortality rate in 1940 should be unaffected by income levels in 1940, which is difficult to believe. As BCF explain very clearly, the “natural experiment” constructed by AJ is flawed. The “treatment group” that received large health gains from technological innovations is fundamentally different from the “control group” that received low health gains, since the “treatment group” had lower life expectancy initially. Therefore, if initial conditions are important for subsequent economic growth, the results will be biased if these initial conditions in 1940 are not considered. BCF included the level of life expectancy in their econometric specifications (a “partial adjustment model”) and they concluded that exogenous improvements in health due to technical advances associated with the epidemiological transition appear to have increased income levels.

In their reply to the reply, Acemoglu and Johnson (2014) address by different means the concern raised by BCF about their original work. First, in order to capture the long-run effects of the initial life expectancy, they include the level of life expectancy in 1900 interacted with time dummies in their decadal panel data set (which runs from 1940). Second, they estimate the “partial adjustment model” of BCF via non linear GMM, since the linear estimation of BCF’s specification will lead to a great deal of multicollinearity and the standard errors become very large. Finally, they use microeconomic estimates from another paper to calculate potential macroeconomic effects of current life expectancy on future growth and examine the implications of their baseline results. AJ conclude that all these approaches confirm that their main results are robust. There is no evidence that increases in life expectancy after 1940 had a positive effect on GDP per capita growth.

There are three issues in this Game of Science that I would like to comment on. First, the intent to quantify the contribution of health to economic growth is extremely relevant for both scientific and policy-related motivations. The general conclusion of the debate, at this stage of the game, is that health conditions were not a factor that shaped the differences in GDP per capita during the second half of the 20th century. Even more generally, the evidence casts doubts on the views that health has a first-order impact on economic growth. With this in mind, it is important to recognize the limitations in the study, especially to extract conclusions for today’s effect of health on economic growth. This is recognized by AJ, who warn that international epidemiological transition was a one-time event and that the diseases that take many lives in the poorer parts of the world today are not the same as those 60 years ago. Despite these considerations, it is important to notice that no author in this debate has questioned the crucial role of improving health conditions to save and improve the lives of millions of people.

Correlation and Causation

Correlation versus Causation

Second, it is important to highlight that the main contribution of AJ is that they provide a sound way to address the problem of endogeneity in order to answer this important question. It is not the first time that Acemoglu and Johnson find a way to design a natural experiment to address some fundamental development questions by using exogenous variation in a country-level panel data setting. In another famous paper, Acemoglu, Johnson and Robinson (2001, AJR hereafter) address the problem of endogeneity that raises in the study of the linkages between income and institutions with the famous instrument of mortality rates of European settlers in different colonies. In both occasions Acemoglu and co-author(s) show us in practice the nuts and bolts of economists’ empirical work, that is, to address the endogeneity concerns by doing good research designs and by finding exogenous sources of variation.

Finally, I see this debate as a privileged example of Popper´s quote. In this short reply to BCF, AJ (2014) present further tests for their results in AJ (2007), overcoming the important point that BCF raise. This is a fair game; both articles are forthcoming in the Journal of Political Economy and the database and programs for AJ papers can be downloaded from Daron Acemoglu’s webpage at MIT. Even more, this is not the first time these authors play the game in the same way. A similar, and also very illustrative debate about AJR (2001) and David Albouy’s critiques can be found in the American Economic Review, or in the NBER working paper. In both debates, Acemoglu and co-author(s) present more evidence on their results that are robust to additional tests, but in both episodes we gain from the debate. We just need to recall that our knowledge is always limited by the evidence we have at the moment, and that this evidence will change over time. After all, in the Game of Science, just like in another famous game, you do not know how it is going to end, even if you read all the books that have been published on the topic.