Resilience to economic shrinking: reinterpreting the Asian economic miracle in a comparative perspective, 1964–2018

ABSTRACT

The successful economic performance of Pacific Asia is often held as a source of inspiration for aspiring catching-up countries. Notwithstanding, recent literature suggests that a key to the understanding of successful long-term economic performance lies not only in the ability to generate economic growth, but in limiting incidences of economic shrinking. Such analyses to explain the Asian economic miracle are however in short supply. This study highlights the significance of economic shrinking in Asia in a comparative perspective to demonstrate how and why resilience to economic shrinking was a significant aspect of the successful development of the region. To approach the question of why some countries became more resilient than others we propose a social capability framework and apply it to a sample of 26 developing countries between 1964 and 2018. We construct a social capability index on which we develop a set of simple OLS regressions to estimate the impact of social capabilities on shrinking patterns. We demonstrate that poorly endowed countries do not lack the ability to generate growth, but their limited resilience prevents them from catching up.

KEYWORDS:

• Social capabilities
• economic shrinking
• Asian economic miracle
• resilience

Introduction

In the post-World War II period only a few developing countries have so far managed to catch up with developed industrialized economies. To address why, attention has typically been directed towards how an economy can achieve economic growth (see, e.g. Collier 2007; Easterly 2002; Koyama and Rubin 2022; Nayyar 2013). Over the last decade, however, it has been argued that successful long-term economic performance is predominately determined by the reduction in the incidence of economic shrinking – that is, when the per capita income level contracts from one year to another (Andersson 2018; Andersson, Axelsson, and Palacio 2021; Broadberry and Gardner 2022; Broadberry and Wallis 2017; North, Wallis, and Weingast 2009; World Bank 2017). While all countries throughout history have had the ability to occasionally achieve growth, just a few have managed to systematically reduce the magnitude and frequency of economic shrinking. This is what happened in large parts of Western Europe, where the declining frequency of shrinking since the end of the eighteenth century improved long-run economic performance. After 1960s, a similar reduction of economic shrinking took place among rising East Asian economies. While the frequency of shrinking remained high among countries in other developing regions, such as Latin America and Sub-Saharan Africa, lower rates in East Asian countries clearly promoted the long-term catching-up process in the region. It is noteworthy that in the literature dominating the discussion of how Asian rose to prominence, the role of shrinking is seldom or never part of the explanations. Recent articles have highlighted the importance of resilience for some individual Asian cases (Andersson, Axelsson, and Palacio 2021; Axelsson and Martins 2023), but region-wide explanations remain lacking.

Despite the importance of economic shrinking and a vast literature on economic crises, efforts to understand what generates resilience to this phenomenon from a wider social perspective have been rare. Classical production functions and conventional growth theories do not provide any explanations. Therefore, there is scant understanding of why some countries shrink more often than others, and what countries can do to change their social and institutional arrangements to overcome frequent shrinking. By proposing an explanation of how resilience to economic shrinking can be strengthened, this study aims to start filling this gap. As a concept, resilience is used in a multitude of disciplinary contexts and can be defined in a variety of ways without any clear consensus (see Béné et al. 2014; Martin 2012; Martin and Sunley 2020). In the context of this study, resilience manifests the ability of an economy to absorb or dodge a disturbance that can lead to shrinking as well as containing the flexibility to adapt to changing circumstances. Concretely, our study highlights why East Asian economies have managed to develop resilience while countries in Latin America and Sub-Saharan Africa have not.

To identify and understand the determinants of resilience to economic shrinking, we develop a framework based on the concept of ‘social capability’ (Abramovitz 1986; 1995; Andersson and Andersson 2019; Temple and Johnson 1998). Our main hypothesis is that improvement in social capabilities helps to build up resilience to shrinking. To approach this issue empirically, we construct a social capability index, that we regard as a proxy of resilience, and apply it to 26 developing countries for the period from 1964 to 2018. The index is based on five broad aspects that we argue can be considered to be directly associated with resilience to economic shrinking: (i) transformation of the economic structure, (ii) market inclusion, (iii) social stability, (iv) accountability and (v) autonomy of the state. Thus, countries with better capabilities may sustain their economic performance, based on less shrinking rather than more growth. This is, we argue, also to a significant, although varied, extent what many countries in Asia were able to achieve that enabled catching up with developed nations over the last half a century.

Given the inadequate attention the role of shrinkage has received in the literature on economic growth in general and in the discussion of Asia’s post WW II development in particular, the objective of the present article is to fill this research gap by making a, to our knowledge, first and explorative attempt to analyze plausible causes of resilience to economic shrinking in a comparative perspective in the developing world since the 1960s. The period is chosen for both analytical and practical reasons. The start of the period roughly corresponds with the beginning of de-colonialization in parts of Asia and in sub-Saharan Africa and also when patterns of divergence in the world economy became more apparent as many economies in Western Europe had entered into their ‘Golden Age’ of high, stable and almost shrinking-free economic performance from 1950-73 (e.g. Crafts 2018; Temin 1997; Toniolo 1998). Furthermore, the availability of comparable data for a broad cross section of developing countries is scarce prior to 1960s. To explore shrinking patterns, the initial sample consists of as many countries as possible in the three regions of Asia, sub-Saharan Africa and Latin America in order to get the broadest possible coverage (in total 83). As we calculate the social capability index, we are forced to slim the sample to 26 countries from the three regions due to data limitations. We ascertained that the remaining countries in the final sample were economically and population-wise sizable in order to foremost include developing countries that carry some weight in those regards.

Our findings are as follows: The successful economic performance in Asia in relation to Latin America and sub-Saharan Africa is to a large degree based on the ability to reduce economic shrinking. Asian economies became over the period increasingly less likely to suffer from frequent and severe occurrences of economic shrinking. The empirical association between shrinking and social capability scoring is also clear. Countries with well-developed social capabilities, predominately found in Asia, shrink less over time, and when they do shrink, the shrinkage is usually not markedly deep, making recovery less troublesome. Conversely, both frequency and magnitude of shrinking are high in countries with poor capabilities. Not only do social capabilities deter shrinking in the long run, but the probability of recording an additional year of shrinking in the year following a shrink year is higher in countries with poor capabilities.

The main implication is that a shrinking perspective puts the Asian catching-up experience in a new light that complements a sometimes one-eyed focus on growth performance alone.

The article is organized as follows: Section 2 surveys the literature on economic catch-up and reveals the infrequency of studies on economic shrinking. Section 3 details the theoretical framework based on social capabilities. Section 4 provides methods and empirical strategy concluded with the construction of the social capability index. In this section, the paper presents a variety of empirical factual and counter-factual illustrations comparing Asian, African and Latin American shrinking patterns. Section 5 presents our results, while Section 6 concludes with a summary of our findings and their implications for future research on this topic.

Literature review

Since the middle of the twentieth century, sustained economic growth has spread to a small set of formerly developing countries, implying that only a few countries have been able to consistently narrow the gap with advanced economies (see e.g. Amsden 2001; Andersson and Axelsson 2016; Rodrik 2017). Only some East Asian nations (Japan, South Korea, Taiwan, Hong Kong and Singapore) have managed to transform their economic structures and experience a strong industrial transformation to reach the status of developed nations (see e.g. Lee 2019; Oqubay and Ohno 2019). Conversely, most Latin American and African economies have not experienced such a transformative process, leading to divergence both globally and within the developing world. Unsurprisingly, the drivers of catching up dynamics or the convergence/divergence process have attracted attention both in mainstream economics and among economic historians. The extensive literature on the Asian ‘economic miracle’ was originally divided into two exclusive, and opposite, camps stressing either openness to market forces or the interventionist role of the development state as the key to the Asian success (e.g. Amsden 1989; Johnson 1982; Kreuger 1981; Little 1981; Wade 1990). The struggle between these two explanatory perspectives might not be as fierce today as the discussion can be said to be situated in a less polarized space where a combination of effective institutions, good governance, well-constructed industrial policies and demand responsiveness of Asian governments and manufacturers are seen as the reasons behind much of Asia’s development experience (e.g. Feenstra and Hamilton 2006; Hausmann and Rodrik 2003; Lin 2012; WB 1993). An area of relative consensus regarding the region’s rise to prosperity relates to the superior technology adoption and innovation policies employed in Asia compared to other economies (Arezki, Fan, and Nguyen 2021; Lee 2019. See also Chopra 2015 for its continuing importance for the region). Nowhere in the miracle literature, however, the successful catching up is attributed to changes in the patterns of economic shrinking.

As far as the discussions of convergence are concerned, standard theory suggests that differences in productivity levels between countries tend to vary inversely with productivity growth rates (e.g. Solow 1956). According to this theory, developing countries should be able to achieve higher growth rates and catch up through the ‘advantage of backwardness’ and actualize that potential by accessing already existing technology and knowledge. The hypothesis of converging productivity levels seems to have been confirmed by the economic growth experience of the Western world during the twentieth century (see, for example, Barro and Sala-i-Martin 1992; Baumol 1986). Though many economists believe that this theory is applicable to East Asian growth as well, it has been argued that the convergence of some East Asian economies was fueled by the accumulation of resources rather than rising productivity (see for instance Krugman 1994). At any rate, when considering the global economy over the last half a century, it seems that divergence has been its dominant feature (Milanovic 2016; Pritchett 1997; Rodrik 2011).

A more historically oriented literature potentially provides a way forward to better understand the catch-up process. A standard reference should be Gerschenkron’s (1962) work, which suggests that the potential advantage of backwardness can overcome the so-called necessary prerequisites by acts of substitution. Following a related idea, Ohkawa and Rosovsky (1973) and Abramovitz (1986) elaborated upon the concept of social capability as the basis of catching up, largely influenced by the post-II War Japanese development experience.

The social capability approach holds that a country has stronger potential for catch-up growth when ‘it is technologically backward but socially advanced’ (Abramovitz 1986, 388). Under globalization, the potential to catch up would be strongest for countries in which ‘social capabilities are sufficiently developed to permit successful exploitation of technologies already employed by the technological leaders’ (Abramovitz 1986, 390). The realization of this potential involves a number of structural and institutional determinants, such as education level, social stability and state capacity. Also, social capabilities are associated with both the ‘ability to exploit modern technology’ and ‘people’s basic social attitudes and political institutions’ (Abramovitz 1995, 29).

In general, the literature on the nature and causes of economic growth in the post-war era has devoted little attention to the role of resilience to economic shrinking. In order to understand the improvement of long-run economic performance, studies have focused on the importance of concepts such as volatility and instability of growth rates, growth reversals and growth collapses. Negative growth rates as a frequent phenomenon in developing countries have been analyzed by Pritchett (2000) who pointed out that the standard growth literature is of little help to understand this issue. A related analytical approach aims to measure and understand the ‘episodic’ nature of economic growth (e.g. Pritchett et al. 2016). Contributions to these discussions have also been made by Easterly et al. (1993) and Rodrik (1999), who have highlighted and explained growth ‘collapse’ and ‘reversals’ through the occurrence of economic shocks and social conflict. Research has also advanced on finding ways to empirically capture distinct episodes of growth dynamics and to associate a number of correlates with either growth spurts or growth stops (Berg and Ostry 2011; Hausmann, Rodriguez, and Wagner 2006; Jones and Olken 2008; Kar et al. 2013). Though these studies may help to understand periods of growth, they do not analyze the relative importance of economic shrinking and its role in the catch-up process of developing economies, nor have they provided a theoretical model that explains why some developing countries are more resilient to shrinking than others.

In an exception, Broadberry and Wallis (2017) explicitly discuss long-run perspectives on economic shrinking by looking to historical data and offer possible explanations of why the industrialized West has managed to overcome it. Based on an analysis of the growth and shrinking trajectories of four industrialized Western economies (the UK, the Netherlands, Italy and Spain), they argue that institutional change and the movement towards ‘impersonal rule’ are the reasons for the reduced incidence of economic shrinking and the fostering of modern economic growth. They note the importance of changes in the rules that govern societies and ‘open access orders’ to limit the influence of powerful elites (see also North, Wallis, and Weingast 2009). Although Broadberry and Wallis provide an explanation of how economic shrinking divided the world into developed and developing countries, it does not document the different shrinking experiences in the developing world and the driving forces of this phenomenon. Notably, on the extensive discussion of how and why countries in Asia managed to escape under-development and, in a few cases, also succeeded in closing the gap to the economically developed world, attention to changing shrinking patterns is largely missing (with the exception of the case of Indonesia, see Andersson, Axelsson, and Palacio 2021; Axelsson and Martins 2023).

Theoretical framework

We take Abramovitz’ classic work on social capabilities as a point of departure to arrive at a modified and empirically more functional framework for capturing resilience to shrinking. Abramovitz famously discussed the pivotal role of social capabilities in catching up dynamics and as such the concept has hitherto been restricted to discussing determinants of growth (e.g. Andersson and Andersson 2019; Perkins and Koo 1995; Putterman 2013; Rohne Till 2022; Temple and Johnson 1998). Although Abramovitz did not distinguish between growth and shrinking, it is clear from both his argument and hypothesis that the capability approach was intended for understanding economic performance over time and not an economy’s prospects for short-term growth. As identified in the next section, economic performance over time constitutes the net effect of the magnitude and frequency of both growth and shrinking. This implies that the social capability-hypothesis of Abramovitz principally applies to both growth and shrinking. In extension, we argue, the hypothesis provides a framework for the understanding of resilience to shrinking. However, a complication with making empirical use of social capabilities has been that ‘no one knows just what it means or how to measure it’ (Abramovitz 1986, 388). Although Abramovitz might have underrated his own contribution on the matter, it is true that preciseness in on how to empirically capture the capabilities is lacking.

In our approach, we follow closely the ideas of Abramovitz as we single out our choice of indicators. Inspired by Kuznets, Abramovitz divided capabilities into two categories where the first is related to the set-up of egalitarian incentives and effective political institutions while the other is associated with the ability of society to make use of new technologies (Abramovitz 1995). To make these categories tangible, resilience to shrinking depends in our framework on five interrelated, but distinct, elements of social capabilities. These are derived from Abramovitz’ two broad categories just mentioned. The latter corresponds straightforwardly to (i) transformation of economic structures. The first, more complex, category we capture by the following four capabilities: (ii) broad-based inclusion of the population in the market, (iii) social stability, (iv) accountability and (v) the autonomy of the state.

Based on these, we develop a composite index to accommodate the different dimensions of social capabilities. These capabilities should be regarded as elements which reflect the forces that strengthen resilience to economic shrinking (Andersson, Axelsson, and Palacio 2021; Axelsson and Martins 2023). Importantly, the capabilities that generate resilience to shrinking, may in and of itself provide further boosting of the same set of social capabilities similar to a process of cumulative causation (cf Abramovitz 1995, 39–40; and Myrdal 1957, illustrated in Figure 1). Below, the capabilities, all of which are grounded in the development literature, are presented in more detail.

Figure 1. Social capabilities setting.

Transformation of the economy from Agrarian to industrial activities

The growth path is marked by a process of structural transformation. Structural transformation entails changes in the composition of output and employment as an economy develops (Kuznets 1973). The transformation of the agricultural sector leads the economy out of poverty by providing cheaper food to urban areas and releasing labor and capital that can be reallocated to the industrial and service sectors.

As this process of structural transformation takes place, the complexity of the economy changes. Economic complexity is understood by Hausmann et al. (2013) as the amount of productive knowledge that the economy contains. Economies dependent on a single resource, usually non-renewable natural resources, are more exposed and vulnerable to a price drop or a slump in demand. Undiversified economies may therefore have a more volatile aggregate output, making potential investors less willing to venture into superior technologies (Acemoglu and Zilibotti 1997). Furthermore, more diversified economies have been shown to be associated with more autonomous institutional settings (Olander 2019). Hence, economies able to produce and export a wide range of diversified, sophisticated and knowledge-intensive products are better prepared to overcome shrinking and avoid bottlenecks in their development process.

Inclusion of the population in the market

While structural change and the release of human resources from agriculture is an important avenue of growth in developing countries, there is no guarantee that labor will automatically be transitioned to higher value-added employment in industry and services (McMillan, Rodrik, and Verduzco-Gallo 2014). It is also critical that losers of the transformation (people employed in sectors that are diminishing in size, typically agricultural or blue-collar workers) be connected to the growth process for it to be inclusive. This is even more important in contexts where strong social protection networks and competitive financial markets are absent, as labor market outcomes are the main determinant of economic welfare for most households.

The inclusion capability is characterized by broad-based economic participation of the population in the market, changing income distribution in favour of poorer households. This provides a more vital and competitive domestic market, with less risk of experiencing supply-side bottlenecks and less fluctuation in domestic prices. High inequality is potentially detrimental to sustained growth in many ways: it prevents the economy from making full productive use of human capacities, fosters growth-inhibiting social conflict and policies, and shortens growth spells (Alesina and Rodrik 1994; Berg and Ostry 2011; Bourgignon 2003; Ostry, Berg, and Tsangarides 2015; Persson and Tabellini 1994). Pro-poor growth, on the other hand, raises the incomes of workers at the bottom of the distribution (Ravallion 2004) and denotes a growth process that would be able to lift poor households above the poverty line (Dercon and Shapiro 2007). This growth process in low-income countries is likely to be labor-intensive, typically engaging rural and relatively less educated labor. By fostering the participation of the majority of the population in economic activity, thus making the most of available human resources, more inclusive societies are less likely to incur shrinking and more likely to enjoy dynamic, cohesive internal markets. In particular, poverty reduction processes seem to be a particularly potent building block for resilience to economic shrinking (Smythe, Martins, and Andersson 2023).

Autonomy of the state

Autonomy means the ability of the state to keep vested interests at bay. This implies the ability to impose direct and progressive taxation on the non-poor, while at the same time remaining sufficiently aligned with powerful actors (aristocrats, entrepreneurs, politicians, journalists, trade unions and other social organizations) to ensure a shared commitment to development policies and goals. Such autonomy resembles the concept of ‘embedded autonomy’ (Evans 1995) and connotes a fine balance of simultaneous cooperating with and disciplining of powerful actors of society (see Acemoglu and Robinson 2019). Autonomy ensures credible commitment to investors or special interest groups and provides opportunities for the creation of consensual and representative government through ‘revenue bargaining’ between states and organized citizens (Brautigam, Fjeldstad, and Moore 2008).

This capability can be revealed in the monetary area of developed and developing countries. The end of the Bretton Woods in 1971 put an end to the convertibility of the US dollar to gold and made inflation a key policy area at the national level. The evidence between inflation and economic performance is mixed, but suggests that high inflation has a negative effect on long-term growth and can be seen as regressive taxation for those at the bottom of the income distribution (Barro 1995; Erosa and Ventura 2002). In the 1990s, many developing countries adopted a clear target for the inflation rate as a response to the loss of inflation tax revenue (Lucotte 2012). The process of implementing inflation targeting is a gradual process of economic and institutional reforms, which allows central banks to deal with difficulties in conducting their monetary policy, such as seigniorage and exchange rate pegs. Hence, a generally accepted bureaucracy of technocrats designs and executes the policy, while other branches of political power are not expected to dictate policy.

In sum, emerging economies that are able to control inflation will experience higher growth rates in the long run as states may improve their performance on tax administration and public provision. If this happens, building up such a capability may help to avoid recurrent shrinking behavior by smoothing the downsides of the economic cycle.

Accountability

While the autonomy of the state is necessary, it may not be enough to avoid arbitrary governance, abuses, waste and persistent inequality. Hence institutional quality also needs to be measured through accountability. This is understood as the quality of governance and provision of public goods (Besley and Persson 2013). Accountability can be summarized as the ratio of social spending and social subsidies to GDP or to total government spending. In low-income countries, education and health investments are used as a measure of the state’s ‘collective’ capacity (Besley and Persson 2014).

Although the provision of public goods is central and may foster catching up and encourage political stability in developing countries, the social and political reach of such spending matters. Thus, the accountability capability can be captured by looking at real outcomes in population health, educational attainment, or infrastructure, which are the bulk of people´s demand in (non)democratic regimes (see Savoia and Sen 2015). An appropriate measure would be life expectancy since it can be considered a good general indicator of comparative success.

Life expectancy captures the various dimensions of the formation of human capital, such as access to health care, education, and the orderliness of urban living (Sen 1998). There is also a strong relationship between per capita income and life expectancy on the aggregate level, but closer inspection reveals that life expectancy in Costa Rica, for instance, is similar to that of the United States despite the differences in income level (Daniels 2007). This implies that the social and political reach of government policies do influence the health–income relationship, and therefore countries that lack an open discussion of how spending is done in public health, education or infrastructure may be clear examples of low levels of accountability.

Social stability and democracy

Lastly, in recent decades the role of the state in ensuring law and order, dealing with social conflicts, guaranteeing the enforcement of contracts and supporting the functioning of markets has been emphasized (Bardhan 2016; Lin 2012; North, Wallis, and Weingast 2009; Rodrik 1999; World Bank 1997). Thus, the capability of social stability centres on success in conflict resolution (Collier et al. 2003; North, Wallis, and Weingast 2009; Rodrik 1999).

In societies where social unrest is high, the willingness to invest might be deterred (Jones and Olken 2008) and even progressive government’s actions may be curtailed as attention to conflict-solving might crowd out the promotion of efficient economic and social policies, leading to a higher likelihood of shrinking. Another source of social instability is the volatility of food prices, particularly in low-income countries (Dawe and Timmer 2012), suggesting that food price stability might also be related to resilience to shrinking.

Method and empirical strategy

To document the shrinking patterns in the developing world, we use data from the Penn World Table (v. 9.1) on GDP per capita growth rates for 45 countries in Sub-Saharan Africa, 17 in Latin America and 21 in Asia. We included all countries with consistent data to represent as much of the three regions as possible.

We calculate the economic performance (EP) over the long run by considering the net effect of the contributions of economic growth and economic shrinking. The contribution of economic growth is equal to the product of its frequency f(g) and its magnitude m(g); analogously, the contribution of economic shrinking equals the product of its frequency f(s) and its magnitude m(s). Frequency refers to the percentage share of growing and shrinking years respectively over a period of time. Thus, EP can be expressed algebraically as follows: (1) $\mathrm{EP}=f(g)m(g)+f(s)m(s)$(1) As f(g) + f(s) equals 1, the equation can be reduced to three independent variables: (2) $\mathrm{EP}=[1\text{–}f(s)]m(g)+f(s)m(s)$(2) The above identity was developed and used by Broadberry and Wallis (2017) to show that the declining impact of f(s) triggered long-run economic growth in industrialized Western economies since the nineteenth century. Additionally, it was shown that the reduction in the incidence of this component, f(s), has a greater impact on long-term economic growth than its magnitude, m(s). Figure 2 shows that it is consistently the poorer countries in the world that have a higher susceptibility to experience economic shrinking (see also World Bank 2017, 5–6). Furthermore, the magnitude of growth is stable and relatively universal among developing regions, while the magnitude of shrinking and its overall impact are highly volatile. This indicates that it is resilience to economic shrinking, not growth, that makes the difference in the catching-up process when looking at the performance of Asian economies in comparison with Latin American and Sub-Saharan African countries.

Figure 2. Concentration of shrinking episodes, by year per quintile (according to GDP per Capita). Source: Shrinking of GDP per Capita from PWT data. Quintiles were made in relation to the country with highest GDP per Capita every year. Interpretation: 28% means that on average, Q1 countries concentrated a 28% of shrinking episodes every year. Period: 1964 to 2018.

Next, we look at the trends and patterns of economic shrinkage in developing countries. Figure 3 shows the development of the frequency of shrinking and the magnitudes of both growth and shrinking in Asian economies. Here, frequency rates ranged around 10–15 percent (which means that, on average, Asian countries shrank 1–1.5 years per decade). The region’s frequency peaked during the 1970s; after that it has continuously declined. The magnitude of growth remains constant, varying between four and six percent of the annual GDP per capita growth rate. The shrinking magnitude remained constant at around three percent during the first three decades, then peaked in the 1990s when the Asian financial crisis occurred; after that, it experienced a significant reduction. Here we find highly resilient economies like the Republic of Korea (which shrank twice in 56 years), Thailand (three times in 56 years) and Malaysia (six times in 54 years). Also, the decline in average frequency was driven partly by countries that were high ‘shrinkers’ in the 1960s and 1970s but barely experienced economic shrinking since then (as India, Indonesia, Myanmar or China).

Figure 3. Frequency of shrinking and magnitudes: Asia. Data from Penn World Tables version 9.1. Left axis: frequency of shrinking by decade. Right axis: magnitude of growth and shrinking. Countries: Bangladesh; Cambodia; China; Hong Kong; India; Indonesia; Japan; Republic of Korea; Laos; Malaysia; Maldives; Mongolia; Myanmar; Nepal; Pakistan; Philippines; Singapore; Sri Lanka; Thailand; Vietnam; Taiwan.

In Figure 4, the patterns in Sub-Saharan Africa look quite different than those of Asian countries. Until the 2000s, when the region experienced a strong improvement (the commodities ‘boom’ took place), the frequency of shrinking was above 30 percent. Both magnitudes show similar numbers in most decades; the main difference between the two was achieved in the first decade of the twenty-first century. The situation in the region was turbulent in the 1980s, with a frequency higher than 50 percent and the magnitude of shrinking higher than the one of growth, meaning that the region’s GDP per capita actually contracted over the entire decade. The frequency of shrinking in the 1970s and 1990s was also quite noteworthy. Nevertheless, these aggregate numbers show poor attainment and high heterogeneity within the region. On the one hand, there are countries like Nigeria, the Republic of Congo and Gabon, amongst others, that continue to experience shrinking frequencies around 40–50 percent each decade. On the other hand, there are relative champions with small shrinking rates, that is, countries that experienced an overall strong improvement during the entire period of analysis (e.g. Mauritius, Botswana and Ethiopia).

Figure 4. Frequency of shrinking and magnitudes: Sub-Saharan Africa. Data from Penn World Tables version 9.1. Left axis: frequency of shrinking by decade. Right axis: magnitude of growth and shrinking. Countries: Angola; Congo, Republic of; Equatorial Guinea; Gabon; Nigeria; Benin; Botswana; Burkina Faso; Burundi; Cabo Verde; Cameroon; Central African Republic; Chad; Comoros; Congo, Democratic Republic of; Cote d’Ivoire; Ethiopia; Gambia, The; Ghana; Guinea; Guinea-Bissau; Kenya; Lesotho; Liberia; Madagascar; Malawi; Mali; Mauritania; Mauritius; Mozambique; Namibia; Niger; Rwanda; Sao Tome and Principe; Senegal; Seychelles; Sierra Leone; South Africa; Sudan; Swaziland; Tanzania; Togo; Uganda; Zambia; Zimbabwe.

In Latin America shrinking patterns (Figure 5) look similar to Sub-Saharan Africa, although with lower levels. The situation in the region reached its worst point in the 1980s, with a lost decade in absolute terms of growth (the region shrank half of the years and with greater magnitude than the magnitude of growth in growing years). The region slowly improved thereafter, recovering to the levels of the 1960s during the last two decades. The trend here is typical of a region that is highly dependent on commodity exports. Again, there is a wide range of experiences within the region. Some countries have a high tendency to shrink (e.g. Venezuela, Argentina and Brazil) while others shrink more infrequently, with rates similar to those of East Asian countries (e.g. Colombia and Costa Rica). Also, some countries have improved significantly during the last 30–40 years (e.g. Chile, and Peru).

Figure 5. Frequency of shrinking and magnitudes: Latin America. Data from Penn World Tables version 9.1. Left axis: frequency of shrinking by decade. Right axis: magnitude of growth and shrinking. Countries: Argentina; Bolivia; Brazil; Chile; Colombia; Costa Rica; Ecuador; El Salvador; Guatemala; Honduras; Mexico; Nicaragua; Panama; Paraguay; Peru; Uruguay; Venezuela.

In developing countries, we are able to discern a variety of shrinking trajectories. Looking at regions in aggregate, we see that Asia performed much better than Sub-Saharan Africa and Latin America– the latter two exhibiting similar patterns –by reducing the prevalence of economic shrinking rather than by achieving stronger growth. Figures 6 and 7 illustrate simulations of what GDP/cap levels and income convergence with the economically developed world respectively would have looked like for countries in sub-Saharan Africa and Latin America had they mimicked the resilience to shrinking of Asia while keeping the factual magnitudes of growth and shrinking intact. Such scenario shows that poorer countries’ income levels would be significantly higher and that the catching up of poor countries to the rich would not only be possible but also under progress. Although merely a simulation, it indicates the significant impact of resilience to shrinking.

Figure 6. GDP per Capita by region and simulations of LA and SSA having Asian shrinking performance. Source: GDP per Capita from PWT 9.1. Simulations of Sub-Saharan African countries and Latin American countries were made taking Asian economies frequency of shrinking and keeping magnitudes of growth and shrinking of each region. Countries: Asia (Countries: Bangladesh; Cambodia; China; Hong Kong; India; Indonesia; Japan; Republic of Korea; Laos; Malaysia; Maldives; Mongolia; Myanmar; Nepal; Pakistan; Philippines; Singapore; Sri Lanka; Thailand; Vietnam; Taiwan), Sub-Saharan Africa (Countries: Angola; Congo, Republic of; Equatorial Guinea; Gabon; Nigeria; Benin; Botswana; Burkina Faso; Burundi; Cabo Verde; Cameroon; Central African Republic; Chad; Comoros; Congo, Democratic Republic of; Cote d’Ivoire; Ethiopia; Gambia, The; Ghana; Guinea; Guinea-Bissau; Kenya; Lesotho; Liberia; Madagascar; Malawi; Mali; Mauritania; Mauritius; Mozambique; Namibia; Niger; Rwanda; Sao Tome and Principe; Senegal; Seychelles; Sierra Leone; South Africa; Sudan; Swaziland; Tanzania; Togo; Uganda; Zambia; Zimbabwe), Latin America (Countries: Argentina; Bolivia; Brazil; Chile; Colombia; Costa Rica; Ecuador; El Salvador; Guatemala; Honduras; Mexico; Nicaragua; Panama; Paraguay; Peru; Uruguay; Venezuela).

Figure 7. GDP per Capita by region, relative to Developed countries, including simulations of LA and SSA having Asian shrinking performance. Source: Relative GDP per Capita from PWT 9.1. (1 = GDP per Capita in Developed economies). Simulations of Sub-Saharan African countries and Latin American countries were made taking Asian economies frequency of shrinking and keeping magnitudes of growth and shrinking of each region. Developed (Countries: Austria; Belgium; Cyprus; Denmark; Finland; France; Germany; Greece; Iceland; Ireland; Italy; Luxembourg; Netherlands; Norway; Portugal; Spain; Sweden; Switzerland; United Kingdom; United States; Canada; New Zealand; Australia).

In order to analyze resilience in developing countries, we selected a sample from the original larger set of 26 countries from Africa, Asia and Latin America that were all considered developing countries at the start of the investigating period (1964).¹ Our selection was based on two additional criteria: that the included countries should be relatively large in terms of population and that data availability should be consistent enough to take account of all capabilities. The sample represents over 40 percent of the global GDP and 60 percent of the world’s population. The period of analysis covers the years from 1964 to 2018, which is as far back as available data allows us to reach. The following data sources were used to create the Social Capability Index: the Economic Complexity Index from the Economic Complexity Observatory (transformative capability); the disposable income GINI coefficient from Solt’s (2020) Harvard database (inclusion); the rate of inflation from the IMF (autonomy); life expectancy from the World Bank’s World Development Index 2020a (accountability); and the Polity 5 Index from the Center for Systemic Peace (2020) (social stability).² We believe the selected indicators satisfactorily capture the capabilities we theoretically argue increase resilience to shrinking. To be able to cover all countries over time, we rely on consistent, comparable and widely used data. In order to meet these requirements, indicators that are perfectly congruent with each capability are lamentably not available. For our purposes, however, they make possible empirical testing of the suggested associations.

Results

In order to construct the Social Capability Index, we rank countries according to their relative positions within the sample of 26 developing countries each year.³ Each capability provides a simple and transparent indicator. In each category, the Index has a value between 1 (best performer in each capability) and 26 (worst performer).⁴ By obtaining a yearly ranking of each social capability variable, we produce an Index by year. Thus, our Index can be represented as follows: (3) $\begin{array}{rl}& \mathrm{SocialCapabilityInde}{x}_{\mathrm{it}}=\mathrm{Rankin}{g}_{\mathrm{it}}[(\mathrm{TransformativeRankin}{g}_{\mathrm{it}}\\ & +\mathrm{InclusionRankin}{g}_{\mathrm{it}}+\mathrm{AutonomyRankin}{g}_{\mathrm{it}}\\ & +\mathrm{AccountabilityRankin}{g}_{\mathrm{it}}+\mathrm{SocialStabilityRankin}{g}_{\mathrm{it}})/5]\end{array}$(3) Table 1 shows the evolution of the Index by decade (based on the results of annual data), sorted by the score during the last decade. The table allows us to distinguish which countries improved overtime and which ones lagged behind. Over the whole period, the country with the best social capabilities as measured by the Index was the Republic of Korea, which had the highest level of economic complexity and lowest inequality in the sample, moderate inflation, high life expectancy and social stability that improved significantly since the 1970s. Furthermore, 6 out of the top 10 countries are in East Asia. At the other end of the spectrum, Nigeria was the most vulnerable economy. Looking at the evolution of the Index (Table 1), there is a group of countries that clearly improved (Chile, China, Peru, Brazil and Thailand); others experienced a slight improvement (Ghana, Senegal, Indonesia, Philippines and Mexico); and some countries’ capabilities worsened slightly (Argentina, Colombia, Kenya, Nigeria and Costa Rica). Finally, there is a group of countries that experienced sharp deteriorations (Madagascar, South Africa, Venezuela, Tanzania and Zambia).

Table 1. Social Capability Index (1964–2018), by decade.

Countries	1960s	1970s	1980s	1990s	2000s	2010s
Korea, Republic	5	9	2	1	1	1
Singapore	2	1	1	2	2	2
Costa Rica	1	2	3	4	10	3
Chile	6	15	16	11	6	4
Thailand	15	12	8	3	7	5
Malaysia	3	6	6	6	4	6
Mexico	13	10	14	12	3	7
Philippines	11	18	11	10	5	8
China	17	7	7	9	9	9
Argentina	14	11	9	7	8	10
Brazil	20	19	17	15	13	11
Peru	19	23	20	21	11	12
India	10	3	5	5	12	13
Indonesia	23	22	19	20	15	14
Colombia	12	13	13	19	16	15
Kenya	16	20	21	24	23	16
Pakistan	9	8	10	13	19	17
Senegal	21	16	15	16	14	18
South Africa	8	14	18	14	17	19
Sri Lanka	7	4	12	17	22	20
Venezuela	4	5	4	8	18	21
Ghana	26	24	22	22	20	22
Madagascar	18	17	24	18	21	23
Zambia	22	26	26	26	26	24
Tanzania	24	21	23	23	24	25
Nigeria	25	25	25	25	25	26

Note: Data from the five categories from which our index is based. Time period (1964–2018). The results here shown are an average of the annual score of each country by decade.

In sum, the Index allows us to track the relative position of countries within the sample, and how they improved/worsened their social capabilities. Below we proceed to analyze the relationship between social capabilities and resilience to economic shrinking between 1964 and 2018 in developing economies. According to our hypothesis, countries with better capabilities display greater resilience to economic shrinking. Such countries may therefore sustain their economic performance to make possible a process of catching up.

In order to test this hypothesis, we estimate the impact of having better social capabilities on economic performance in its most basic form (GDP per capita growth), to have an overview of its effect; on the shrinking trajectory (in terms of frequency), to understand how its decline is triggered by better capabilities; and lastly on the magnitudes (of both shrinking and growing), to see whether the severity of crises is lower in countries endowed with better social capabilities.

To further investigate resilience to economic shrinking, we run two types of OLS regressions: one for the relative Social Capability Index and another for the five separate capability indicators that make up the Index. This is intended to distinguish any additional pattern or information that may be missed by evaluating the Index in an aggregate manner. An important caveat is that the estimated associations are correlations. Making causal inferences would require detailed case studies, which is beyond the scope of the present study. Despite the limitations of this method, the use of correlations provides at this stage sufficient insights to establish a rough, yet generally applicable, sketch of the relation between social capabilities and resilience to economic shrinking.

Economic performance

We start with simple linear regressions to understand the relationship between economic performance and social capabilities (measured by the Index we developed in Section 5). The data we use comes from the Penn World Table (PWT), version 9.1 (Feenstra, Inklaar, and Timmer 2015). The data are annual for both variables (the Index and GDP per capita growth) and available for most countries between 1964 and 2018.

In Figure 8 we plot economic performance (from PWT data) and our Index in relative terms (a score close to 0 identifies the best performer, while a value of 1 is the worst performer). We distinguish that countries with low social capabilities have a much weaker economic performance than those with high social capabilities, between 1964 and 2018. Countries at the bottom of the social capabilities distribution are almost stagnant, while those at the top achieve annual average growth rates in the range of three to four percent. Table 2 confirms the association between social capabilities and economic performance: the better the score in the index, the more successful the economic performance.⁵ Furthermore, the R-squared between values are close to 40 percent, suggesting that the model helps to explain differences in economic performance across countries rather than what happened within each country.

Figure 8. Relationship between economic performance and social capabilities (1964–2018). Data: Social Capabilities Index, own creation; Economic performance based on GDP per capita growth, PWT 9.1. Average between years 1963 and 2018.

Table 2. OLS estimation, economic performance and S.C.Index.

	(1)	(2)
Variables	Performance	Performance
Social Capability Index (0 top–1 bottom)	−0.0295***	−0.0315***
	(0.0107)	(0.0106)
1960s		0.00853**
		(0.00420)
1970s		0.00195
		(0.00383)
1980s		−0.0102***
		(0.00383)
1990s		−0.000791
		(0.00384)
2000s		0.0106***
		(0.00384)
2010s		-
Constant	0.0401***	0.0399***
	(0.00630)	(0.00654)
Observations	1396	1396
R-squared (within)	0.006	0.036
R-squared (between)	0.3998	0.3971
Number of countries	26	26

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

The results here shown are based on Fixed Effects (within groups estimators).

From the previous discussion, we can say that there is strong evidence that countries with worse social capabilities display a weaker economic performance than those with better capabilities. However, this does not help us distinguish which of the five capabilities that compose the Index may be driving this higher growth. Table 3 shows the OLS regression using the five variables separately in relative terms (0 means best performer and 1 worst performer). By looking at the coefficients, we can see that lower transformation (measured by the Economic Complexity Index) and lower inclusion (GINI) can push up annual growth rates for this set of countries. Conversely, lower state autonomy reduces growth, virtually offsetting the ‘positive effect of inequality’. These results show that less complex and unequal societies can experience higher overall growth. It makes sense that in less diversified, developing economies dominated by primary goods producers and commodity exporters, and where assets and resources are concentrated in the hands of a small elite, rent-seeking behaviors may lead to short-term economic growth. However, this does not provide any insight about the stability of such growth in the long run.

Table 3. OLS estimation, GDP per capita growth rate and 5 social capabilities.

	(1)	(2)
Variables	Growth Annual	Growth Annual
ECI (ranking relative, (0 top–1 bottom))	0.0317***	0.0317***
	(0.01000)	(0.00982)
GINI (ranking relative, (0 top–1 bottom))	0.0307***	0.0257***
	(0.00924)	(0.00913)
Inflation (ranking relative, (0 top–1 bottom))	−0.0249***	−0.0259***
	(0.00513)	(0.00504)
Life Expectancy (ranking relative, (0 top–1 bottom))	0.00968	0.00924
	(0.0176)	(0.0173)
Polity 5 (ranking relative, (0 top–1 bottom))	−0.00330	−0.00296
	(0.00556)	(0.00548)
1960s		0.0111*
		(0.00605)
1970s		−0.000509
		(0.00481)
1980s		−0.0155***
		(0.00461)
1990s		−0.00950**
		(0.00455)
2000s		0.00143
		(0.00452)
2010s		-
Constant	0.00385	0.0115
	(0.0118)	(0.0120)
Observations	1,052	1,052
R-squared	0.044	0.087
Number of Countries	26	26

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

The results here shown are based on Fixed Effects (within groups estimators).

Therefore, we need to check whether unequal and less complex economies can sustain growth in the long run. To do so, we run the same OLS regression, but taking the average growth rate by decade as the dependent variable instead of the annual growth rate.⁶ Table 4 shows that the positive effects of complexity on growth disappear while the effect of inequality remains, but with a diminished level of statistical significance. This confirms that starting a process of economic growth and sustaining it in the long run are not the same (Hausmann, Rodriguez, and Wagner 2006; Jones and Olken 2008). Countries may achieve short-term growth, but downturns can be recurrent, offsetting previous gains and to understand the sustainability of growth we need to take the role of resilience to shrinking into account.

Table 4. OLS estimation, GDP per capita growth (Decade) and five social capabilities.

	(1)	(2)
Variables	Growth Decade	Growth Decade
ECI (ranking relative, (0 top–1 bottom)), by decade	0.0249	0.0207
	(0.0160)	(0.0146)
GINI (ranking relative, (0 top–1 bottom)), by decade	0.0235*	0.0247**
	(0.0127)	(0.0116)
Inflation (ranking relative, (0 top–1 bottom)), by decade	−0.0257**	−0.0261***
	(0.0106)	(0.00983)
Life Expectancy (ranking relative, (0 top–1 bottom)), by decade	0.0301	0.0325
	(0.0262)	(0.0238)
Polity 5 (ranking relative, (0 top–1 bottom)), by decade	0.00167	0.00242
	(0.00937)	(0.00854)
1960s		0.0129**
		(0.00606)
1970s		0.00411
		(0.00528)
1980s		−0.0116**
		(0.00486)
1990s		−0.00302
		(0.00486)
2000s		0.00845*
		(0.00476)
2010s		-
Constant	−0.00208	−0.00269
	(0.0180)	(0.0164)
Observations	132	132
R-squared	0.104	0.302
Number of country	26	26

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

The results here shown are based on Fixed Effects (within groups estimators).

Frequency of shrinking

In this section, we assess whether better social capabilities may drive long-term catch-up by reducing the frequency of shrinking. The measure used for frequency of shrinking is the number of times that each country registered economic shrinking by decade.

To graphically illustrate this relation, Figure 9 displays the share of frequency of shrinking by decade and the Social Capability Index by country between 1963 and 2018. This scatter plot shows that there is a positive correlation between worsening social capabilities and experiencing higher frequencies of shrinking: countries with worse social capabilities score an average frequency of around 40 percent while the frequency for those with better capabilities is lower than 10 percent (see Table 5 for regression outputs). The Asian countries are clustered at the southwest corner, combining relatively low frequency of shrinking and relatively high score on social capabilities. The case of Indonesia might be somewhat of an outlier in terms of social capabilities, but this is mainly due to the fact that the average low score 1963-2018 does not reveal the continuous improvements of the Index score taking place in Indonesia over time.

Figure 9. Relationship between Frequency of Shrinking by decade and Social Capability Index. Data: Social Capabilities Index, own creation; Frequency of shrinking by decade, PWT 9.1. Based on decade data of both variables between 1963 and 2018.

Table 5. OLS estimation, with Fixed Effects. Frequency of Shrinking (Decade) and S.C. Index.

	(1)	(2)
Variables	Frequency Decade FE	Frequency Decade FE
S.C. Index (ranking relative, (0 top–1 bottom))	0.370**	0.393***
	(0.160)	(0.150)
1960s		0.0357
		(0.0485)
1970s		0.140***
		(0.0471)
1980s		0.227***
		(0.0468)
1990s		0.128***
		(0.0473)
2000s		0.0204
		(0.0471)
2010s		-
Constant	−0.00407	−0.110
	(0.0970)	(0.0903)
Observations	156	156
R-squared	0.040	0.253
Number of countries	26	26

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

The results here shown are based on Fixed Effects (within groups estimators).

Results in Table 5 reaffirm the role of capabilities in reducing the incidence of shrinking and promoting favourable economic performance. The coefficients here can be interpreted as follows: taking the results of column (2) countries with the best social capabilities such as South Korea (relative Social Capability Index = 0.04) have a predicted frequency of shrinking of 1.57 percent (0.393*0.04 = 0.01572); on the other hand, countries with the worst social capabilities (near to 1) have a predicted frequency of shrinking close to 40 percent by decade (0.393*1 = 0.393). Hence, having relative high level of social capabilities makes a significant difference in terms of shrinking, and consequently affects long-term economic performance.⁷ Dummy variables for each decade show, as expected, that shrinking was much higher in the 1980s in comparison with the 2010s, and to a lesser extent in the 1990s, 1970s and 1960s. The fact that the dummy variable for the 2000s is not statistically significant may be a sign that shrinking was not very different than in the 2010s.

Differences in the frequency of shrinking are significant due to social capabilities in the long run, but this is also relevant on a yearly basis. As a way to understand how resilience is shaped in the short term, we decided to adopt an alternative approach – a Probit model with data based on an annual basis. Here, the measure of shrinking is all the times that a country went below 0 percent in GDP per capita growth rate. Thus, the dependent variable has a value of 1 if the country incurs shrinking and 0 otherwise. Running the Probit model with this variable and the yearly relative Social Capability Index (Table 6) shows that being at the bottom of the distribution (Index = 1) significantly increases the probability of shrinking by 152.7 percent in that year. Conversely, the probability of shrinking for a country with an Index value of 0.1 (top 10 percent) increases by just 15.27 percent.

Table 6. Probit model, annual shrinking.

	(1)	(2)
Variables	Annual Shrinking	Annual Shrinking
S.C. Index (ranking relative, (0 top −1 bottom))	1.527***	1.561***
	(0.185)	(0.191)
1960s		0.212
		(0.164)
1970s		0.605***
		(0.148)
1980s		0.844***
		(0.146)
1990s		0.557***
		(0.148)
2000s		0.174
		(0.154)
2010s		–
Constant	−1.692***	−2.167***
	(0.120)	(0.171)
Observations	1447	1447

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

The results here shown are based on Fixed Effects (within groups estimators).

To sum up the results of this section, social capabilities can help to configure the resilience to economic shrinking in developing countries. Improvements in the set of capabilities considered here seemingly enable emerging economies to suffer less from recurrent shrinking. The differences in shrinking –between being a highly endowed country with rich social capabilities and being socially backward in terms of capabilities – are noteworthy. Differences are outstanding in the long term, but the probability of shrinking when a deterioration of capabilities takes place in the short run is also clear.

Magnitudes of growth and shrinking

Additionally, to further comprehend how resilience to shrinking is shaped, we include an analysis of the role of social capabilities in the severity of economic contraction (magnitude of shrinking) and in the size of growth periods (magnitude of growth). When talking about the magnitude of shrinking we mean the ‘size’ of the negative rate (below 0 percent of GDP per capita annual growth) that a country experienced. On the other hand, the magnitude of growth indicates the positive rate (above 0 percent of GDP per capita) that a country experienced. We argue that resilience to economic shrinking, measured by social capabilities, also smooths the severity of crises. Additionally, as developing countries also have the capacity to grow, at least in the short term, the magnitude of growth between developed and developing nations would not be as high as the difference in shrinking. We test this set of hypotheses through similar OLS models as we did in the previous two subsections, but this time the dependent variables are both magnitudes.

Table 7 shows in the first two columns that a lower Social Capability Index leads to deeper or ‘more negative’ shrinking rates. Columns 3 and 4 show that a higher Index value increases the size of growth rates (‘more positive’), but its contribution is much lower than for shrinking (by looking at the size of its coefficients, comparing columns 1–2 versus 3–4). In sum, countries with lower social capabilities (in comparison with better-endowed countries) have greater shrinking rates and greater growth rates, although the effect of the magnitude of shrinking is greater than the effect of the magnitude of growth in their overall economic performance. We also know that countries with lower social capability have a greater frequency of shrinking. This confirms that countries with lower social capability do not lack the ability to generate growth, rather the contrary. However, their resilience to shrinking is very low. This implies that social capabilities, by flattening the severity of shrinking both in terms of magnitude and frequency, rather than merely increasing growth, should be acknowledged when understanding the prospects of catching up.⁸

Table 7. OLS models with magnitudes of shrinking and growth.

	(1)	(2)	(3)	(4)
Variables	Negative rates	Negative rates	Positive rates	Positive rates
S.C. Index (ranking relative, (0 top–1 bottom))	−0.0457**	−0.0505***	0.0245***	0.0188**
	(0.0180)	(0.0185)	(0.00818)	(0.00821)
1960's		−0.0121		0.0146***
		(0.00971)		(0.00298)
1970s		−0.00962		0.0114***
		(0.00858)		(0.00278)
1980s		−0.0156*		0.00546*
		(0.00809)		(0.00284)
1990s		−0.00593		0.00506*
		(0.00837)		(0.00278)
2000s		−0.00664		0.0113***
		(0.00920)		(0.00269)
2010s		–		–
Constant	−0.00405	0.00887	0.0264***	0.0215***
	(0.0121)	(0.0153)	(0.00462)	(0.00470)
Observations	319	319	1,077	1,077
R-squared	0.022	0.041	0.008	0.041
Number of countries	26	26	26	26

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

The results here shown are based on Fixed Effects (within groups estimators).

Discussion of results

Our results highlight the importance of resilience to economic shrinking for long-term catching up. Most countries can achieve short-term growth, but few are those that manage to restrict the frequency and severity of economic downturns. The importance of low shrinking is displayed by the successful experience of the East Asian economies since the 1960s. Despite low initial GDP per capita level in comparison to other developing regions, high resilience of Asian countries is shown by the low incidence and frequency of shrinking episodes. Thus, according to our analysis, resilience to shrinking in the long run became a key determinant in putting East Asian countries on the catching-up track while most of the other emerging regions could not do so.

Social capabilities seem to have been decisive in this process. Strong capabilities of Asian economies consolidated the high resilience of the region. Lower inequality levels, increasing economic transformation, and higher state autonomy were particularly relevant. The inclusion of larger segments of the population in market activities and lower concentration of productive factors consolidated domestic markets by avoiding supply-side bottlenecks and stabilizing internal demand. The more cohesive internal markets in combination with a fuller use of human capacities may also have reduced the incidence of economic downturns. In the region, this was further consolidated by the dynamic process of structural transformation experienced over time. This process guaranteed the reallocation of former agricultural workers into industrial and service sectors, preventing them from falling back into poverty. In addition, the reduced reliance on few commodity exports limited the incidence of volatile external markets.

Our finding’s limited relevance of social stability measured by the autocracy-democracy indicator is not surprising. The relation between democracy and economic development is not straightforward (Przeworski et al. 2000). China is the most evident case of this when looking at shrinking patterns. Despite the lack of an open political system, the country managed to place itself on a low-shrinking path. Chinese low inequality, structural transformation, and improvements in human development were pivotal. Although limited gains in some of the single capabilities, our analysis nevertheless shows that the overall improvement in social capabilities is a key indicator for building resilience to shrinking.

The importance of our findings becomes stronger in the case of trade and capital markets turmoil. The role of capital market liberalization has been highlighted in making crises less likely and less steep (Stiglitz 2000). For the East Asian countries, this became evident during the 1990s (Furman et al. 1998). The low incidence of that shock in China and India is explained by the interaction of local policies and high capabilities which displayed the high resilience of these two economies. Because of that not only the role of trade shocks should be added in future extensions investigating shrinking patterns, but the inclusion of capital inflows/outflows and the construction of sound financial systems in the analysis will strengthen the understanding of economic shrinking in developing countries.

It is beyond the scope of this paper to provide specific policy implications emanating from our findings other than it is advisable to strengthen the social capabilities proposed. However, a sound precaution should be that in order for development policy to shape resilience to shrinking it is required to also understand the institutional underpinnings of each society. As argued by North (2005), what sets successful development examples apart is adaptive efficiency, meaning institutions that are stable yet allow for policymakers to make reforms and base policy priorities on changing conditions. The policymaker needs what UNCTAD refers to as a suitable policy space that allows for this flexibility to choose appropriate policy for each specific societal setting (Jackson 2021) and a research agenda devoted to such tasks would certainly push forward our knowledge on how to design development policy for long term, sustainable development.

Conclusion

This study suggests that economic shrinking prevents developing economies from catching up with developed nations, while resilience to shrinking, as observed in many Asian countries, is a decisive factor for sustained economic progress. The regions and countries that have managed to catch up did so by achieving long-term successful economic performance based on developing resilience to economic shrinking rather than achieving growth alone. Hence, according to our analysis, the prevalence of shrinking clearly restricts the potential of developing countries to close the gap with developed countries. Although standard growth and catching-up literature acknowledges that achieving growth is not the same as sustaining it, the role of resilience to economic shrinking has still received very limited attention. As so far as shrinking patterns have been acknowledged, the varied experiences among developing countries over the last half a century per se have so far not been recognized.

This paper aimed to fill these gaps by analyzing possible determinants of resilience to economic shrinking in the developing world, of which we considered a set of 26 emerging countries from Asia, Latin America and sub-Saharan Africa between 1964 and 2018. To this end we constructed an index based on social capabilities, inspired by Abramovitz (1986), serving as our proxy of resilience.

The Asian countries clearly distinguish themselves from other countries and demonstrate features that contain hitherto ignored aspects behind the post WW II economic miracle of the region. Our analysis suggests that countries developing stronger social capabilities demonstrate better economic performance, predominately because they tend to shrink less over time. And when they do, the magnitudes are usually not very large, making economic recovery easier and catching up more probable. Conversely, frequency and magnitude of shrinking are higher in countries with limited social capabilities, which prevents them from catching up. While countries characterized by high social inequality and simple economic structures indeed have the ability to achieve growth in the short term, such growth is unlikely to be sustained. Countries in Asia corroborate these patterns. Regardless of social capabilities, countries can grow, but it tends only to be the well-endowed countries that develop resilience to shrinking. Hence, social capabilities tend to have a stronger effect on resilience to shrinking than on growth per se.

Finally, it is important to emphasize that this study is one of the first attempts to analyze long-term development from this perspective; so far, we have only scratched the surface. Even if our main hypothesis of the link between social capabilities and resilience to shrinking still stands, we are still far from reaching any conclusive evidence of the dynamics involved. Further studies may be pursued to better understand specific events in transitions from high- to low-frequency shrinking. The importance of the export structure in promoting resilience to shrinking would be a fruitful avenue for further exploration, as would the questions of the driving forces behind the improvement in social capabilities and how resilience to shrinking can be incorporated into strategies for sustainable development. Also, the role of capital markets should be included in future extensions investigating shrinking patterns. Looking at capital inflows/outflows patterns and the construction of sound financial systems will be key to fully understand the interaction of external forces with social capabilities in developing countries. Although this article analyses developing countries over the last half-a-century, future studies should also examine the development of resilience to shrinking during earlier periods. Such studies should emphasize the role of colonial legacies in the construction of social capabilities and their implications on long-term shrinking patterns.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Riksbankens Jubileumsfond under Grant [P18-0603:1]; Marianne and Marcus Wallenberg Foundation under Grant [MMW 2014.0151]; and Jan Wallanders och Tom Hedelius Stiftelse samt Tore Browaldhs Stiftelse under Grant [P21-0044].

Notes

1 Argentina, Brazil, Chile, China, Colombia, Costa Rica, Ghana, India, Indonesia, Kenya, Madagascar, Malaysia, Mexico, Nigeria, Pakistan, Peru, Philippines, Senegal, Singapore, South Africa, Sri Lanka, South Korea, Tanzania, Thailand, Venezuela and Zambia.

2 Summary statistics (Table A1 in the appendix).

3 Table A2 in the Appendix shows how the Index is elaborated, by disaggregating the average of five categories for each country for the full period of analysis.

4 Best-performing would mean to have the highest economic complexity (EC), lowest income inequality (lowest GINI coefficient), lowest inflation, highest life expectancy and highest social stability.

5 In order to avoid comparability problems of the index through time, we transformed it to range from 0 (top performer) to 1 (worst performer). In practice, best performers achieve a value of around 0.037 and worst performers a value of 1 (see summary statistics).

6 Note also that for the five independent variables (capabilities) we are considering their averages by decade.

7 Dummy variables for each decade show that shrinking was much higher in the 1980s in comparison to the 2010s, and to a lesser extent in the 1990s, 1970s and 1960s. The fact that the 2000s dummy is not statistically significant may be a sign that shrinking was not very different than the numbers from the 2010s.

8 The robustness of our models was tested in the section included in the appendix for that purpose. There, by switching some of the proxies in which the index is based on and extending the sample of countries (to 49 developing countries), we confirm that our results hold, economic performance and resilience to shrinking is shaped by social capabilities.

Appendix

Table A1. Summary statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
GDP per Capita Growth	2601	2.23%	4.54%	−28.63%	51.86%
Frequency of Shrinking by decade	2657	23%	21%	0%	100%
Shrinking year	2699	22.71%	41.90%	0.00%	100.00%
Magnitude of Growth	1988	3.97%	3.22%	0.02%	51.86%
Magnitude of Shrinking	613	−3.42%	3.49%	−28.63%	0.00%
Inflation	2352	32.310	320.338	−8.423	11,749.640
Economic Complexity Index	2495	−0.389	0.677	−2.764	1.906
GINI, disposable	1773	43.029	6.312	26.7	60.1
PolityV	2480	0.947	6.644	−9	10
Life Expectancy	2503	61.674	10.368	18.907	82.495
Government Effectiveness Index	980	−0.233	0.646	−1.582	2.437
Legal System & Property Rights Index	1222	4.582	1.069	1.970	8.070
Terms of Trade	1004	114.834	44.712	43.878	458.575
Social Capability Index, relative ranking, year (26 countries)	1447	0.5793	0.2073	0.04	1
Social Capability Index, relative ranking, year (49 countries)	931	0.6083143	0.2126804	0.1058824	1

Table A2. Social Capability Index and Score by Capability, mean of 1964–2018.

	Transformation	Inclusion	State Autonomy	Accountability	Social Stability	Capability Index
Countries	ECI	GINI	Inflation	Life Expectancy	Polity IV	Index
Korea, Republic	1	1	6	5	11	1
Singapore	2	3	1	1	23	2
Costa Rica	6	13	14	2	1	3
India	5	10	9	18	2	4
Malaysia	8	16	2	6	14	5
China	4	4	5	10	26	6
Philippines	13	8	7	14	8	7
Thailand	12	11	4	11	12	8
Chile	14	20	10	4	4	9
Argentina	9	7	23	3	15	10
Mexico	3	21	16	7	13	11
Venezuela	16	6	26	8	7	12
Pakistan	15	2	13	17	17	13
South Africa	11	26	8	19	3	14
Sri Lanka	20	18	11	9	9	15
Colombia	9	23	19	12	5	16
Brazil	7	22	22	13	10	17
Senegal	18	14	3	22	19	18
Peru	17	24	17	15	6	19
Kenya	18	19	12	20	20	20
Madagascar	23	12	15	23	16	21
Indonesia	21	15	18	16	21	22
Ghana	25	5	24	21	18	23
Tanzania	24	9	21	24	25	24
Nigeria	26	17	20	26	24	25
Zambia	22	25	25	25	22	26

Note: Each category has a ranking. Last column ‘Capability Index’ is based on the average score of all five categories. The value of the Index is the result of ranking that average of the five categories.

Table A3. Frequency of shrinking by decade, 1960s–2010s.

Country	1960s	1970s	1980s	1990s	2000s	2010s	Ranking 1960–2010
Korea, Republic	0%	10%	0%	10%	0%	0%	1
Thailand	0%	0%	0%	20%	10%	0%	2
Sri Lanka	14%	10%	0%	0%	10%	0%	3
Malaysia	0%	10%	20%	10%	20%	0%	4
Colombia	14%	0%	30%	20%	0%	0%	5
China	29%	20%	20%	0%	0%	0%	6
Pakistan	0%	20%	0%	20%	30%	0%	7
Indonesia	43%	0%	10%	20%	0%	0%	8
Singapore	14%	0%	20%	10%	30%	0%	9
Costa Rica	0%	20%	30%	20%	10%	0%	10
Philippines	0%	0%	30%	40%	10%	0%	11
Chile	14%	40%	20%	10%	10%	0%	12
India	43%	50%	0%	10%	0%	0%	13
Mexico	0%	0%	50%	10%	40%	13%	14
Tanzania	20%	50%	40%	20%	0%	0%	15
Peru	14%	30%	60%	20%	20%	0%	16
Brazil	0%	0%	50%	40%	30%	38%	17
Ghana	71%	50%	30%	0%	0%	13%	18
South Africa	0%	40%	60%	40%	10%	25%	19
Kenya	29%	20%	40%	70%	20%	0%	20
Argentina	14%	40%	60%	30%	30%	38%	21
Senegal	43%	50%	40%	50%	30%	13%	22
Nigeria	57%	40%	60%	50%	0%	38%	23
Zambia	20%	80%	100%	40%	0%	13%	24
Venezuela	14%	30%	60%	50%	40%	63%	25
Madagascar	14%	80%	70%	60%	30%	38%	26

Note: Frequency of shrinking by decade (times that GDP per capita growth went below 0 percent by decade), based on data from Penn World Tables 9.1. Last column shows a ranking based on the mean of all decades.

Robustness checks:

To test the strength of our previous results, we run some robustness checks. We test proxies for autonomy and accountability. Inflation as a measure of autonomy and life expectancy as a measure of accountability may confound the results because multiple causes may affect their evolution in the long run. The alternative measure of Autonomy is the Legal System and Property Rights index from the Fraser Institutes’ (2020) Index of economic freedom. This indicator reflects the extent to which justice, legal systems and property rights are independent, unbiased, impartial and secured in different countries. This index ranges from 0 (lowest) to 10 (highest), from year 1970 to 2018. Additionally, the alternative measure of accountability is the Government Effectiveness measure from World Bank’s Worldwide Governance Indicators 2020b (WGI), which reports data for 200 countries over the period 1996 and 2019. This Government Effectiveness index captures the quality of public and civil services, the quality of policy formulation and implementation and the credibility of government’s commitment to such policies. This index ranges from −2.5 to 2.5.

The second aspect is sample size. Despite having full information between 1963 and 2018, the fact that our analysis is based on data from a relatively small number of countries (26 countries) could question the robustness of our findings. To address this concern, we extended our sample with 23 additional countries (Algeria, Bangladesh, Bolivia, Burkina Faso, Cambodia, Cameroon, Dominican Republic, Ecuador, Egypt, El Salvador, Guatemala, Honduras, Jamaica, Laos, Mali, Mauritania, Morocco, Mozambique, Nicaragua, Panama, Paraguay, Uruguay and Vietnam).

With these considerations, we created the new social capabilities index. Then we did two exercises: (i) we use the new index for the original 26 countries sample between 1996 and 2018 (note that our initial year is now 1996 as data for Government Effectiveness was not available before that). (ii) we extend the sample size to the 49 countries sample because of the loss of observations when reducing the period. Table A4 provides the results of the impact of the new index on economic performance. The first column displays the results for the original sample between 1996 and 2018. As we can see its significance has decreased, possibly caused by the fact that observations are fewer, but still showing that improvements in social capabilities improve the economic performance. In column 2, where we extended the sample, the relation between the two variables is reinforced.

Table A4. OLS estimation, Economic performance and S.C.Index.

	(1)	(2)
Variables	Performance Annual FE	Performance Annual FE
New Social Capability Index (0 top–1 bottom)	−0.0365*	−0.0416***
	(0.0207)	(0.0149)
1990s	−0.00933**	−0.00740**
	(0.00455)	(0.00315)
2000's	0.00498	0.00290
	(0.00389)	(0.00277)
2010's	–	–
Constant	0.0467***	0.0546***
	(0.0109)	(0.00882)
Observations	494	931
R-squared	0.046	0.030
Number of countries	26	49

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Additionally, Table A5 confirms the relationship between frequency of economic shrinking by decade and the social capability index. The first column (fixed effects for the 26 countries) is not significant. If we consider the fact that there we just have 78 observations (3 observations per country) this is not unexpected. Moreover, when we extend the sample in columns 3 and 4 the results are significant. Comparing the results of this table from the ones of Table 5, it can be appreciated that coefficients are smaller. This makes sense due to the fact that during the period 1996 and 2018 economic shrinking in developing countries was lower than between 1963 and 1996. The last two decades (2000s and 2010s) were particularly good for developing countries in terms of shrinking and, as we just have information since 1996, those two decades reduce the size of the coefficients.

Table A5. OLS estimation, Frequency of Shrinking (Decade) and S.C. Index.

	(1)	(2)	(3)	(4)
Variables	Frequency Decade FE	Frequency Decade RE	Frequency Decade FE	Frequency Decade RE
New S.C. Index (ranking relative, (0 top–1 bottom))	0.197	0.247**	0.302*	0.166**
	(0.202)	(0.103)	(0.171)	(0.0769)
1990s	0.132***	0.128***	0.145***	0.154***
	(0.0387)	(0.0360)	(0.0282)	(0.0263)
2000s	0.0207	0.0169	0.0458*	0.0516**
	(0.0385)	(0.0360)	(0.0269)	(0.0260)
2010s	–	–	–	–
Constant	−0.000362	−0.0283	−0.0905	−0.0117
	(0.117)	(0.0662)	(0.101)	(0.0500)
Observations	78	78	147	147
R-squared	0.288		0.321
Number of countries	26	26	49	49

Note: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.