Inequality, poverty, and resilience to economic shrinking

1. Introduction

How can societies become more resilient to economic shrinking? The vast majority of literature on economic catch-up dynamics has focused on the growth imperative. However, research by Broadberry and Wallis (2017) introduced the concept of economic “shrinking” as an alternative focus when understanding the determinants of long-term economic performance. Shrinking is when a country’s output per capita is less compared to the preceding period [1], usually measured per year, as is the case in this study. Economic shrinking manifests itself in two ways: in the frequency of episodes and the magnitude of rates. When it comes to developing countries’ prospects for catching up with more prosperous countries, the issue is not that they do not experience growth episodes; it is that such episodes are unsustained as they are frequently interrupted by shrinking episodes (Broadberry and Wallis, 2017; Andersson, 2018). Moreover, it has even been hypothesised that successful long-term economic development has primarily been due to economies, at a certain moment, beginning to experience a decline in the frequency of economic shrinking (Broadberry and Wallis, 2017; Andersson, 2018). Shrinking magnitude, on the other hand, is thought to be more related to how a country can handle short-term shocks rather than systemic developmental weaknesses (Andersson, 2018). As such, past research is ill-placed to explain how an economy internally might build resilience to economic shrinking.

In parallel to these developments, a large body of literature has argued for the importance of “inclusive growth” or “shared prosperity” for long-term development (see Ianchovichina and Lundström, 2009; World Bank, 2020). The burgeoning research on economic shrinking also suggests that inclusive societies are more resilient. However, less attention has been devoted to what and how particular aspects of inclusiveness are associated with the frequency of economic shrinking over time. This study investigates two central features of inclusiveness – income inequality and poverty – and discusses how these characteristics relate to patterns of shrinking in the developing world.

Income inequality within countries and absolute poverty are increasing in many places globally, with 60% of the global population still below the $4.16 (purchasing power parity [PPP]) per day poverty line (Jayadev et al., 2015). Moreover, rising inequality is fuelling growing societal dissatisfaction and destabilising democracies (United Nations, 2020). Still, inequality trends are not all bad news. Global relative income inequality between countries has been declining steadily for the past three decades as Gini coefficients have been falling in East Asia, Sub-Saharan Africa and Latin American countries (Gradin et al., 2021).

Similarly, global poverty trends show some success, albeit a mixed one. At the World Bank’s extreme poverty line benchmark of $1.25 (2005 PPP) per day, the world has seen substantial poverty reduction over the past 40 years, with the most dramatic changes felt in the East Asia and Pacific region (Jayadev et al., 2015). Poverty reduction in Latin America has also seen success though Sub-Saharan Africa has experienced little decline. However, this measure has come under increased criticism with developing nations often setting themselves far higher poverty benchmarks (Sharma, 2018; Roser, 2021). Considering these higher and perhaps more realistic measures of poverty, the world has seen only small reductions in global poverty levels. By using the higher measure of $4.16 (2005 PPP) per day, for example, 80% of the Sub-Saharan Africa population would still be considered poor (Jayadev et al., 2015).

Mixed success in achieving both poverty and inequality reduction could serve as a call for more growth. Countries need to produce higher growth rates to eliminate such developmental hurdles, yet many countries have already made the journey from lower- to middle-income countries. However, between 1951 and 2016, the major difference in economic performance between Asia, Latin America and Sub-Saharan Africa was a lower frequency of economic shrinking rather than comparatively higher growth rates in years of positive growth (Andersson, 2018). By giving Sub-Saharan Africa the shrinking frequency pattern of Asia over this period, Andersson’s (2018) counterfactual simulation resulted in Sub-Saharan Africa having a GDP per capita more than three times higher in 2016 than this measure in reality. Furthermore, poorer countries experience higher shrinking frequencies on average, whereas average growth and shrinking magnitude rates do not fluctuate considerably in the long run. Table 1 shows the average shrinking frequencies for 183 countries between 1964 and 2018, split across GDP per capita quintiles, Q1 representing the poorest countries and Q5 the richest. Table 1 shows the average shrinking magnitudes and growth rates for the same countries and period, suggesting that these rates are relatively comparable though shrinking magnitude is slightly higher for poorer countries. Interestingly, it is only the poorest of countries that have lower average growth rates than shrinking magnitude, further suggesting that growth alone is not the panacea for developmental success.

Understanding and exploring the dynamics of how an economy might build resilience to economic shrinking episodes can give fresh insight into the development process. Simply focusing on how to increase growth rates potentially misses the other side of the coin when it comes to long-term development and societal prosperity. As a concept, resilience is used in many disciplinary contexts and can be defined in a variety of ways without any clear consensus (Martin, 2012; Béné et al., 2014). In this study, resilience expresses 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, enabling the economy to reduce the risk of experience subsequent shrinking. Inspired by the social capability approach developed by Abramovitz (1995), we argue that economic inclusion – measured by both inequality and poverty indexes – can be understood as a social capability that decreases the chance of a country to experience episodes of economic shrinking. An economic shrinking episode in this paper is measured annually and thus defined as a dummy for negative output from the preceding year. While this focus does not capture shrinking magnitudes, we believe that it is an appropriate measure for investigation as shrinking frequency arguably takes logical precedent over magnitude. Should a country build resilience to shrinking frequency, then concerns for magnitude are effectively “solved”. Furthermore, shrinking frequency is hypothesised to be more related to a country’s systemic weaknesses rather than short-term shock response, and the social capabilities framework used in this study was developed specifically with this in mind.

This study builds a longitudinal data set including 23 developing countries throughout 42 years [2] to test the extent to which poverty and inequality can explain countries’ economic resilience, proxied by a rolling shrinking ratio. Our findings suggest that all measures of inclusiveness are correlated with lower shrinking frequency rates, yet it is the changes in poverty – under many different thresholds – which correlate with countries’ resilience building. Inequality, while still an important element to explain countries’ growth potential as an initial condition, does not seem to make our sample more resilient to shrinking. We conclude that the mechanisms in which poverty and inequality are correlated with the catch-up process must run through different channels. Ultimately, processes that explain growth may intersect but not always overlap with the ones that explain resilience to shrinking.

The rest of this article is as follows: Section 2 discusses previous literature on convergence, building to the relatively novel concept of economic shrinking before presenting the social capabilities theoretical framework. Section 3 highlights the regional inequality and poverty trends over the past 50 years before discussing previous studies highlighting their relationships with development. Sections 4 and 5 outline the data and empirical strategy, respectively, used in this study. Section 6 presents and analyses the article’s main results. Section 7 concludes with the observation that the processes that govern growth and resilience to shrinking are different but may overlap and, as such, poverty eradication is perhaps of greater importance for building resilience to shrinking episodes.

2. Literature review and the analytical framework

Most of the post-Second World War era is defined by economic divergence, and catch-up is seldomly observed (Collier, 2007; Rodrik, 2011; Milanovic, 2016). The convergence hypothesis in growth theory proposes that developing economies tend to grow faster, per capita, than developed economies due to the law of diminishing returns to investment. Developing economies can also adopt existing technologies without the pressures and costs associated with innovation, with the developed nations subject to different mechanics by being at the forefront of the technological frontier. Pritchett (1997), in his now-famous essay “Divergence, big time”, noted the disadvantages for developing countries that such “backwardness” has resulted. He acknowledges the notable exceptions of some East Asian economies, commonly referred to as the Asian Tigers, but highlights how surprisingly few researchers have tried to tackle the economic volatility and negative growth rate phenomenon (Pritchett, 2000).

Economic volatility has since received more attention but has mainly been viewed through a growth lens, with much of the literature focused on the variations of growth rates. Output volatility has been documented to be higher in lesser developed nations and causes fragilities that negatively impact growth rates. The reasons for the fragile and increased cyclical nature of output in some developing countries have been diverse, though they have focused on the lack of export diversification and poor institutional environments (Calderon and Yeyati, 2009).

Broadberry and Wallis (2017) investigate economic volatility by examining the long-run economic performance of selected European countries by defining a measurement of economic performance that includes instances of growth episodes and growth rates alongside shrinking episodes and shrink rates. They determined that economic performance was primarily improved by decreases in the magnitude and frequency of shrinking episodes rather than any increases in growth rates. By taking into account the role of shrinking, they also find that increases in economic performance over the long run are also associated with a general decline in both shrinking frequency and short-term economic growth rates. Moreover, they go on to hypothesise institutional change to be the key reason for such shrinking reductions.

Economic development can thus be thought of in terms of both the growth and shrinking episodes that economies might face. Andersson (2018) added his perspective to the argument of Broadberry and Wallis (2017) by analysing the economic performance of a select group of Sub-Saharan Africa, Latin American and Asian countries from 1951 to 2016. His conclusions concurred that the success of Asian countries has been primarily driven by their resilience to economic shrinking, as opposed to the often exclusively considered high growth rates.

With regard to the frequency of shrinking episodes, Figure 1 shows the total shrinking frequencies per year for this study’s sample countries. Immediately noticeable is the general early volatility, which steadily fell over the given period. In 1982, the highest measure of 74% of the sample countries experiencing a shrinking episode, yet this was 0% nearly two decades later, in 2004, 2005 and 2007. Between 1974 and 1990, the average yearly shrinking episode frequency was 37%. This had fallen to 29% from 1990 to 2000, and still further to 12% from 2000 to 2015. Moreover, Figure 1 presents the relative weighting that each region contributed to the total shrinking period. In line with conjectures by Andersson (2018), East Asian economies have had a relatively strong resilience to shrinking over the entire period, with their most frequent shrinking years stemming from the 1998 Asian Financial Crisis and the 2009 Great Recession. On the other hand, both Latin American and Sub-Saharan Africa countries experienced frequent bouts of shrinking episodes. Even in the more resilient post-2000 period, the majority of shrinking episodes are still found in these two regions.

Returning briefly to the theory of economic convergence; the theoretical “advantages of backwardness” may be present, but the ability for countries to take full advantage of them may not be. Developing Gerschenkron’s (1962) classic work further, Abramovitz (1986, 1995) introduced his concept of “social capability” in that an economy’s potential to converge is stronger if technologically backward but socially advanced. The premise is that while new technology might be there for adoption, the actual process of adoption might be reliant on societal factors that inhibit such developments. For example, low skill levels might make certain technologies too difficult to use en masse. Some structural and institutional environments would thus need to be in place for developing countries to successfully exploit new technologies and move up their respective technological ladders (Abramovitz, 1995). As such, it is through this framework that Andersson and Palacio (2017) and Andersson (2018) developed a social capabilities theory in building resilience to economic shrinking. They present five broad categories that encompass the social and institutional aspects considered important to the development process: inclusion, structural transformation, state autonomy, state accountability and social stability and conflict resolution.

Inclusion refers to the access to economic opportunities and their distribution so that the population at large can broadly participate and reap the benefits of economic activity (Andersson, 2018). Growth episodes that are pro-poor are more likely to be sustained (Pritchett and Werker, 2012), with higher inequality and poverty also tending to limit resource allocation with underutilised human capital and productive capacities. The beneficial entities of inclusiveness and equity pertain to society at large, for example, when it comes to general well-being and a levelled playing field of opportunities, increasing trust, sense of fairness and a stronger social contract (Wilkinson and Pickett, 2010; Ferreira and Peragine, 2015). With regard to resilience building specifically, Andersson and Andersson (2019) found that a key detriment to sustainable growth patterns in Côte d’Ivoire and Senegal between 1930 and 1980 was the lack of broad-based economic participation.

Structural transformation of an economy has traditionally been seen as a vital part of any growth process, both economic and social. An agricultural transformation is seen as a key determinant of structural change as this can release labour and capital to higher productivity sectors while helping to support sustainable growth patterns (Andersson, 2018). Transformation also indicates a greater diversification and complexity in what an economy can produce giving less global market vulnerabilities through, for example, price or demand slumps (Hausmann et al., 2013). On the other hand, Indonesia, between 1950 and 2015, was able to generate strong pro-poor growth and substantially reduce poverty through farmer investment that increased production and food security (Andersson et al., 2021a). However, faced with capital and labour constraints, industrialisation led to the formation of powerful special interest groups, which only changed after the emergence of democracy (Andersson et al., 2021a). Furthermore, Andersson and Andersson (2019) showed that strong agricultural growth in Côte d’Ivoire and Senegal was not enough to drive long-term development, and societal transformation lagged behind (Andersson and Andersson, 2019). Both of these examples highlight how developments in one social capability alone would not be enough to sustain development patterns.

The aspect of state autonomy is the ability of the central government to keep vested interests in check. This capability ensures states are insulated from outside influence and, as such, are likely to credibly represent a consensual and exemplary government for promoting general development policies (Andersson, 2018). For Indonesia, Andersson et al. (2021a) interestingly showed how the initial phases of industrialization, in some sense, required a “worsening” of autonomy to address inclusion and transformation capabilities, which later on became detrimental and was only effectively addressed with the fall of the Suharto regime.

The accountability capability is the ability of the state to provide quality governance and public goods provisions; in essence, the legitimacy of the governing among the governed (Andersson, 2018). For example, under-taxing and tax evasion are common features of lower-income countries, which can restrain virtuous cycles of development through constrained fiscal space and limited investment. Broadly speaking, the Latin American region has managed to build an increased resilience to economic shrinking through democratic and liberal economic transitions, though this was aided by favourable terms of trade since the 2000s (Andersson et al., 2022).

The final social capability under this theoretical framework is that of social stability and conflict resolution. The ability of countries to peacefully resolve conflict within their respective societies has been seen as a key aspect of the long-term development process. Here, the state is seen to play an integral part by providing such institutions as law and order and encouraging “good” business environments (North et al., 2009; Rodrik, 2000). The potential destructive nature of conflict is an obvious impediment to development, but a state can also be fiscally constrained by needing to commit more resources to conflict resolution rather than other investments.

Combined, the framework suggests that advancing these social capabilities will increase the ability of an economy to absorb or circumvent both internal and external disturbances that would otherwise increase the risk of an economy to shrink. As such, the capabilities generate resilience to economic shrinking. Andersson et al. (2021b) built and investigated a social capabilities index for 26 developing economies between 1964 and 2018. They found strong support for the notion that the more advanced the capabilities, the greater the resilience to shrinking in severity, frequency and recovery. Moreover, Axelsson and Martins (2023) argue that Indonesia’s capabilities, when compared with Brazil, are only more advanced due to Brazil starting its development journey earlier. Indonesia’s seemingly greater resilience to shrinking is due to it having relatively more advanced capabilities while also suggesting there is an ideal timing for developing capabilities (Axelsson and Martins, 2023).

3. Inclusive economic development

The inclusion aspect of the theoretical framework is the focal point emphasising more equal access to economic opportunities and broad-based participation. A steady decline in poverty rates over the past decades in East Asian and Latin American countries gives hope that economic development has been more broadly shared in these regions. Figure 2 shows the $4.16 (2005 PPP) per day poverty headcount for the three regions under investigation. Stagnant levels can be seen in Sub-Saharan Africa, with the regional average beginning and ending the period with around 85% of the population below this line. This contrasts sharply with the East Asian region, whose regional poverty headcount average declined from around 70% to 30%, and Latin America, whose regional average declined from around 57% to 27%. While the overall decline was sharper in East Asia, Latin America can be said to have made monumental progress over the past decades, particularly in light of their relatively higher economic shrinking frequencies.

Income inequality trends over time are shown in Figure 3, and following reasoning by Milanovic (2016), wave patterns in each region can be observed. Both the East Asian and Latin American regions ended the period with their regional Gini coefficient averages lower than when they started. East Asia’s Gini declined from 48 to 45, while Latin America’s Gini declined from 50 to 47. Sub-Saharan Africa’s Gini, on the other hand, remained stagnant at around 0.58. Therefore, Sub-Saharan Africa and Latin America begin and end the study period with higher levels of income inequality than the East Asian region. In this way, they can be said to be less egalitarian than their East Asian counterparts, with Sub-Saharan Africa’s income inequality perhaps being too high initially, limiting the regions’ ability to reduce inequality in the first instance (Adams, 2004; Fosu, 2017).

Previous research has found that the more unstable outlooks of volatile developing economy environments follow close associations with increased poverty and income inequality, with causality likely flowing in both directions. For example, Gavin and Hausmann (1998) find that the terms of trade for Latin American countries were largely exogenous to domestic market macroeconomic developments, implying output volatility causally limited long-term growth rates by a full percentage point per year in their sample. Moreover, they show that real GDP volatility was the only macroeconomic variable in their study that explained income inequality, even proposing it accounts for around a quarter of the sizeable inequality differences between Latin American countries and their higher-income industrialised counterparts. The relatively affluent in society can handle transitory shocks better through higher savings and better access to formal and informal channels of financial support, whereas an unemployment spell for the poor can be a mechanism of intergenerational poverty transfer.

These associations are unsurprising as Ravallion (2012) argues that higher levels of poverty limit economic growth with the potential to form a consumption trap, limiting the poverty-reducing aspects of growth in the first instance. Such a relationship could be inferred from Figure 2 for Sub-Saharan Africa, for example. Human capital accumulation is also vulnerable to volatility through poverty. Ferguson et al. (2007) note that higher levels of poverty disadvantage individuals’ educational outcomes, with the outcomes being influenced by the depth, duration and timing of poverty instances. For example, the poor may need to take children out of school, which relatively disadvantages them further against higher-income families. Furthermore, economic volatility also increases the volatility of poverty itself, which, in turn, is consistently associated with increased overall poverty rates (Gnangnon, 2021). In this regard, Dabla-Norris et al. (2015) show that more sustainable growth patterns are mostly poor and middle-class-driven through several economic and social dynamics. For example, education has been one of the most important drivers of increasing higher income shares for poorer households, with educational access also lifting the poverty elasticity of economic growth and economic growth itself (Dabla-Norris et al., 2015). Increased financial inclusion, rather than simply capital deepening, also can play an important role by encouraging innovation and growth by giving poorer entrepreneurs the credit access they need to progress their ideas (Doering, 2016).

The potential negative effects of income inequality and poverty become evident on economic growth but, as of yet, are left unexplored from the shrinking perspective. Huang et al. (2015) find that income inequality is significantly and positively associated with higher economic volatility, but the significance disappears when growth rates are negative. Such a study touches on the magnitude of shrinking rates, without explicitly stating as much, but misses the frequency of shrinking perspective. Thus, their study only considers short-term shock dynamics with income inequality and does not consider poverty aspects at all. The novelty of shrinking, therefore, offers a new way of measuring and analysing vulnerabilities, particularly shrinking frequency as a measure of long-term patterns. While an egalitarian and inclusive society is more likely to demonstrate greater resilience to economic shrinking, the effect of changes in income inequality and poverty on economic shrinking frequency has yet to be explored.

4. Data

This study uses annual time series data for 23 countries across three distinct geographical regions from 1974 to 2015. The social capabilities, that are theorised to help build resilience to economic shrinking, are inclusion, transformation, state autonomy, state accountability and social stability (Andersson, 2018). No ideal measures of these capabilities exist and, following Palacio (2018), Andersson (2018), Andersson and Andersson (2019), Andersson et al. (2021) and Axelsson and Martins (2023), well-known proxy measures are used. As measures of inclusion, poverty headcount ratios, Gini coefficients and Palma ratios are taken from the Global Consumption and Income Project with measures of output from the PWT 10.0. Measures of accountability are taken from the PWT 10.0, human capital index and World Development Indicators, under 5’s mortality rate, with the autonomy proxy, inflation, also taken from the WDI. Following de Vries et al. (2021), transformation measures were taken from the GGDC’s Economic Transformation database with historic measures extrapolated backwards using growth rates from the 10-sector database [3].

The Global Consumption and Income Project (GCIP) standardises the Gini coefficient and Palma ratio for all countries to be income-based, and the main poverty measure used in this study is the poverty headcount ratio $4.16 (USD) (2005 PPP). This line is set by the US Thrifty Food Plan as the minimum cost to achieve a nutritious diet in the USA. As this measure is by PPP, it can be used as an international measure for a food poverty line (Jayadev et al., 2015). Alternative poverty lines are considered as robustness checks.

To arrive at a measure of economic shrinking consistent for time series analysis, an economic shrinking dummy variable was introduced with a value of 1, if the GDP per capita growth rate was negative, and 0 if the rate was positive. This was then transformed into a non-binary measurement by a cumulative moving average per country over the period in question. This measure represents our major dependent variable, named here as the Shrinking Frequency Ratio. As the cumulative rolling average of equally weighted yearly observations was taken, the Shrinking Frequency Ratio can be conceptualised as the percentage of shrinking vs non-shrinking years over a given period. This is formally expressed by:

While the Shrinking Frequency Ratio enables time series analysis, it notably does not take into account shrinking magnitude rates. While this measure does not capture shrinking magnitudes, we believe that it is an appropriate measure for investigation for three reasons. Firstly, shrinking frequency arguably takes logical precedent over magnitude, as should a country build resilience to shrinking frequency, then the magnitude becomes a moot point. That is not to say that shrinking magnitude is not important, only that it shifts considerations towards depth of crisis and recovery, which cannot be covered in this study alone. Secondly, shrinking frequency is hypothesised to be more related to a country’s systemic weaknesses rather than short-term shock response. The social capabilities framework used in this study was developed with this in mind. Thirdly, previous studies have investigated income inequality with negative growth rates before (Huang et al., 2015), yet shrinking frequency studies are lacking and provide a complimentary perspective.

5. Empirical strategy

Although both panel-data and cross-sectional models are common in income inequality and poverty studies, panel models are generally more appropriate, with cross-sectional parameter estimates more likely to face omitted-variables bias (Forbes, 2000; Konya and Mouratidis, 2006). Ultimately, this study uses country and period fixed-effects specifications to account for such unobservable characteristics as much as possible.

The Shrinking Frequency Ratio shall act as the dependent variable, with explanatory independent variables following. However, to arrive at the final specification pooled ordinary least squared (OLS) regressions are necessarily first estimated using a stepwise method. As such, an initial OLS model specification can be expressed as:

In this specification, µ and θ denote the time-invariant and time-variant components, respectively, with ε denoting the regression error term. Subscript i denotes individual countries, representing the spatial aspect captured by differing economies. Subscript t denotes the time dimension of the study, measured in years. Five-year moving averages for the independent variables were used. This strategy is used to dynamically capture the changing trends by eliminating the need for arbitrarily lagged variables. The logarithmic transformation of GDP per capita is also included to account for different country income levels, making results more comparable.

Model [2] represents regression numbers 1 through 4 in the Results section, with each subsequent estimate introducing a new social capability into the model. One problem with OLS specification is that there is a significant possibility of the independent variables also being correlated with country-specific factors or the error term, which can introduce omitted variable bias. To control for any potentially idiosyncratic confounding factors, a Hausman test was performed to establish the suitability of a fixed- or random-effects model for the final regression. The null hypothesis was rejected, indicating that a fixed-effects model would be more appropriate. This, along with the moving-average method, has the advantage of smoothing longitudinal data, helping to account for any omitted variable bias in the time-invariant component. Thus, Model [3] represents regression 5 estimate results and can be written as:

Here, α denotes the intercept concerning the individual countries now present in the fixed-effects model without the time-invariant component. Another consideration is around period fixed-effects; thus, a Wald test was run against including year fixed-effects. The null hypothesis was rejected, indicating that such a control should be included in the final fixed-effects models. The Model [4] for regression 6 of the Results section can thus be specified without the µ time-invariant and θ time-variant components:

One potential problem with fixed-effects methodology is it can have a “flattening out” effect on variables that change little over time. The chief variables under investigation, the Gini coefficient, Palma ratio and poverty headcount, have been known to exhibit such characteristics. It is for this reason that the stepwise OLS models are included in the regression output tables to allow for cross-comparisons.

One final model that is of interest is the pooled effect of interacting geographic regional dummies with the main independent variables of interest, i.e. inclusion. This is to gauge the different effects that each region may experience considering previous “clustering” observations while holding all other variables constant. This specification follows on from the fixed-effect method of equation (4), including testing, and can be expressed as:

In Model [5], χ donates the categorical interaction term for the different geographic regions. This model represents the most complete and comprehensive specification used in this study. Also, it is important to highlight the interpretation of these regional interactions. The effect of the inclusion variable for the geographic regional dummy would then be given by the sum of β₁ + β₂, while the effect of the baseline region would be given as β₂ only of Model [5].

Common with panel data is the presence of serial autocorrelation, which was an expected issue as using moving averages across time series can create temporal autocorrelation. When present, serial autocorrelation can lead to the overly optimistic significance of regressors due to the standard errors of the estimates being too low. As such, a Lagrange-multiplier test was used to confirm the presence of serial autocorrelation in all models, as the null hypothesis was rejected. The Breusch–Pagan and Cook–Weisberg test for heteroskedasticity was also performed on all models. All models rejected the null hypothesis, indicating the presence of heteroskedasticity of the error term. Therefore, all estimations apply Huber and White’s estimators clustering the standard errors to correct for autocorrelation and homoscedastic error terms.

A final consideration is needed around cross-sectional dependence, also known as spatial correlation, which is a potentially common issue in macroeconomic panels with relatively long time series, i.e. over 20- or 30-year periods (Torres-Reyna, 2007). Cross-sectional dependence can enter panels through unobserved but correlated factors, leading to biased standard errors (Hoechle, 2007). The Pesaran test was used for the fixed-effects models, and the null hypothesis was rejected. Thus, final fixed-effects specifications are presented with Driscoll and Kraay standard errors to correct for cross-sectional dependence, heteroskedasticity and autocorrelation.

Even with this empirical strategy, it is recognised that the issue of reverse causality still persists. Fixed-effects methodology, robust standard errors and moving averages are attempts to help mitigate this point as much as possible. It is, however, not fully accounted for, and reverse causality is not only possible but also probable. A lack of shrinking episodes leading to reductions in poverty or income inequality is just as plausible as the other way around or at least that causality would flow in both directions. As such, the logarithm of GDP per capita is included, alongside other social capabilities proxies, to try and account for potentially differing growth rates in between shrinking episodes by capturing country income levels. Alongside country-fixed effects, year fixed-effects are also included as a means of accounting for simultaneous shocks with the goal of trying to isolate the shrinking frequency and inclusion associations as much as possible. As this study represents, to the best of our knowledge, the first exploration of income inequality and poverty from a shrinking frequency perspective, endogeneity is a recognised limitation that must be kept in mind when interpretating the results.

6. Results

Previous argumentations suggest that more equal and less poverty-stricken economies should tend to shrink less often than societies with unequal income distribution and a high incidence of poverty. However, it is less clear whether there are continuous associations between inequality and poverty on the one hand and shrinking episodes on the other.

Below are the main empirical results presented using the shrinking frequency ratio as the dependent variable. All estimation results produced F-tests >0.000, rejecting the null hypothesis and indicating that every model holds some explanatory power. Within-group R² is reported for each model, which consistently improves with each subsequent specification.

7. Conclusion

With the recognition that generating economic growth is not the same as sustaining it, the challenge to catch-up and growth literature is discerning between these processes. In this paper, we set out to investigate how societies can become more resilient to economic shrinking as a means of sustaining the development journey. The need for inclusive growth in long-term development has been championed for decades, yet inclusion has seldom been explored from the shrinking perspective. Income inequality and poverty, as proxies for economic inclusiveness, are shown to not correlate uniformly to shrinking patterns in the developing world. As such, a stronger inclusion social capability is important in any developing process, but seemingly poverty eradication plays a more important role in building resilience to shrinking.

Our findings suggest that marginal changes in income inequality are not correlated with developing country shrinking frequency. Reductions in poverty, on the other hand, appear significantly correlated with shrinking frequency. As such, income inequality perhaps then represents an initial condition for resilience to shrinking. The more egalitarian East Asian economies displayed consistently lower shrinking frequencies, even in high-poverty societies. Yet, their poverty reduction efforts did not correlate with reduced shrinking frequencies when compared to Latin American and, especially, Sub-Saharan Africa economies. Thus, poverty alleviation can be hypothesised to directly impact resilience to shrinking. Income inequality can then be said to run through indirect channels but also its ability to speed up poverty eradication through the poverty-growth-inequality triangle (Bourguignon, 2004). These findings, coupled with strong empirical and theoretical reasons advocating for inclusive growth, give rise to the notion that processes explaining growth may intersect but not always overlap with processes explaining resilience to shrinking. Therefore, we argue that the mechanisms in which income inequality and poverty are correlated with developing country catch-up processes run through different channels.

The obvious policy implication of this study is the need for greater poverty reductions in developing countries. Though this is already an important mainstream political objective, this paper differentiates itself by providing an alternate reason why. Relying on inclusive growth as the vehicle of long-term development might underplay the importance of building resilience to shrinking alongside growth dynamics. Perhaps then income inequality has more of an economic growth limiting effect, while poverty reduction is required to build resilience to economic shrinking. Developing countries will need both growth and resilience to shrinking to catch up with higher-income economies, which policymakers might need to balance carefully. As such, future research should emphasise the need to uncover the different channels through which poverty and income inequality seem to operate. Both will play important roles in any development process but for seemingly different reasons.