Measuring ‘what?’ and ‘how?’ Welfare development and the dependent variable problem within East Asian social policy analysis | Melbourne Asia Review

Melbourne Asia Review is an initiative of the Asia Institute. Any inquiries about Melbourne Asia Review should be directed to the Managing Editor, Cathy Harper.

Email Address

Since the 1990s, there has been a surge in interest in the development of welfare policies across East Asian nations and territories. Critical questions for East Asian scholars have become what welfare policy areas to focus on and how to conceptualise and empirically measure East Asian welfare development in response to shared challenges, such as globalisation, ageing, and unpaid family caring responsibilities.

I argue that in order to address these questions what is needed is a way to address the so-called dependent variable problem within East Asian social policy analysis.

This article introduces the notion of the ‘dependent variable problem’ within social policy analysis (Part 1), presents empirical findings of selected social spending, social rights and social outcome indicators to demonstrate the extent to which it aggravates macro-level analyses of East Asian welfare development (Part 2), and makes several suggestions on how scholars can potentially address its main implications for East Asian social policy analysis (Part 3). 

1. The dependent variable problem

The ‘dependent variable problem’ primarily refers to the lack of conceptual clarity when multidimensional concepts such as ‘welfare transformation’, ‘welfare development’, and ‘welfare state change’ are used to analyse the nature of a nation’s social policy development typically including the historical pillars of the welfare state namely social security, employment, housing, health care, and education. Across all of these social policy areas, the lack of clarity results in different findings on the key trends and drivers of welfare development.

This theoretical ambiguity led political scientist Christoffer Green-Pedersen to conclude that, ultimately, the ‘dependent variable problem [within social policy analysis] is a problem of theoretical conceptualisation rather than a problem of data’.

In a previous article focusing on rich OECD/EU countries, I argued that limitations of widely-used indicators in mainstream comparative social policy analysis have also contributed to the ‘dependent variable problem’. For instance, expenditure-based summary measures of welfare development produce inconsistent findings at the country level and are often hampered by conceptual limitations and imprecise operational definitions.

Advances in the widespread availability of social rights-based summary measures that focus  on adjustments in benefit levels, eligibility criteria, amongst others, have enabled a seemingly preferable perspective on welfare development. But the proliferation of these measures has largely failed to end long-fought debates about best and second-best measures in comparative social policy analysis.

In addition, a growing interest in social divisions, investment, sustainability, and the ability of welfare policies to alleviate growing income inequality after the 2008 Global Financial Crisis and, more recently, the COVID-19 pandemic has re-balanced the focus of much analysis on the dynamics of social outcomes.

It is worth contrasting the findings of several approaches to measure ‘welfare state development’ in East Asia. This will demonstrate that the ‘dependent variable problem’ has not only afflicted Western scholarship but that the ‘what’ and ‘how’ to measure welfare development continues to present a very real challenge for East Asian social policy scholars too. 

2. Capturing the multidimensionality of welfare development in East Asia

Considering the abovementioned issues, a suitable starting point for analysing East Asian welfare development remains social spending figures, typically including expenditure on social security, employment and health care policy, published by national statistical bureaus or official census and statistics departments (Figure 1).

The figures presented here lack international standardisation and, therefore, must be treated cautiously. Nevertheless, a picture of stability and slow upward expansion emerges, with Japan outspending all other East Asian economies roughly by a factor of three (South Korea and Taiwan) or four (mainland China, Hong Kong SAR – hereafter: Hong Kong). However, Japan remains the only economy in East Asia with social spending close to the OECD average of 22 per cent of GDP in 2022. South Korea and Taiwan have reached levels of social expenditure of 11-12 percent and Hong Kong and mainland China continue to lag in relative terms in relation to advancing levels of social spending of 5-6 percent of GDP by 2019. Moreover, Taiwan (after 2013) and Hong Kong (after 2013 and 2015, respectively) are the only two economies with some fluctuations in social spending, as their short spells of cost containment/retrenchment were followed by subsequent increases in social spending.

Social spending-based measurements of ‘welfare development’ have been subject to criticism on several fronts—too many to mention here in great detail. Suffice it to say that the so-called ‘denominator problem’ is particularly pertinent in East Asian scholarship as GDP growth in mainland China has consistently outperformed economic development in the high-income Tiger economies and Japan after the 2000s. This means that ‘welfare development’ expressed by social spending as a percentage of GDP creates a bias against the Chinese case and that other ways to standardise expenditure figures in international dollars and purchasing power parities, such as per capita, by shares of unemployed and elderly, etc., may be more appropriate.   

Figure 1: Social spending in five East Asian economies, % of GDP, 2010–2020

Notes: China: public expenditure on social security, employment, and health. Hong Kong: public spending on health and social welfare. Sources: China: National Bureau of Statistics, 2021; Ministry of Finance, PRC, 2021; Hong Kong: Census and Statistics Department, 2021. Japan and Korea: OECD, 2021; Taiwan: Directorate-General of Budget, Accounting and Statistics, Executive Yuan, ROC, 2021. Adapted from Kühner & Shi (2023).

When comparing the aggregate social spending data with social rights data on benefit replacement levels, qualifying criteria, duration limits, and coverage rates of unemployment benefits, sickness benefits, and public pensions, Japan appears to be a much less exceptional case (Table 1). Indeed, following sociolologist Gøsta Esping-Andersen’s approach in his ground-breaking book The Three Worlds of Welfare Capitalism to capture the overall generosity of welfare policy, South Korea’s public pensions index score is positioned much closer to Japan, and in the cases of Taiwanese unemployment and sickness benefits even surpassed those in Japan by 2010. The disaggregated data also shows a much more apparent expansion of unemployment protection in South Korea (since 1996) and Taiwan (since 2000) compared to sickness benefits and public pensions, which have largely remained stable or even experienced mild degrees of retrenchment (see, e.g., Taiwan’s public pensions since the 1970s).

Table 1: Welfare generosity in three East Asian economies, 2010-2011

Note: Welfare generosity is computed as the sum of unemployment benefit, sickness benefit, and public pensions generosity index scores. Mainland China and Hong Kong are not included in the Comparative Welfare Entitlements Dataset II (CWEDII).

In short, welfare generosity statistics suggest a much more fluctuating picture of welfare development compared to social spending statistics particularly when considering individual policy areas rather than total aggregate scores.

Japan retains the most extensive overall welfare generosity among the three economies—at least until more complete data is available for other high-income East Asian economies. However, welfare generosity statistics have also been subject to criticism. For instance, social rights tend to be much less accessible than social spending statistics and struggles with sub-national/state-level variation in social rights. Furthermore, the calculation of benefit replacements levels typically depends on the average production worker or other model assumptions, which becomes less meaningful in the context of substantial informal/precarious labour markets and internal migration, for instance dividing those with and without local Hukou. There is, then, a danger that analysis of East Asian welfare development that focusses on social rights biases the analyses on ‘paper freedoms’ over the lived experiences of many working families.

In our analysis of East Asian welfare development based on a mix of social spending and social rights data across six policy domains, including old age income protection, housing, unemployment, education, health care, and family, and spanning 16 indicators, a colleague and I found a noticeable overall expansion of East Asian welfare development since the 1990s (Figure 2).  

However, this general trend masks substantial variation between development trajectories for (most) East Asian economies. According to our quantitative and qualitative data, the most significant welfare development is apparent in South Korea, which has seen significant increases in all six policy domains. Likewise, Taiwan managed to increase welfare development in three policy domains: education, old age pensions, and unemployment benefits.

Hong Kong has traditionally been weak regarding its family policy and unemployment benefit provision. However, according to our analysis, it increased its overall welfare development by investing in old-age pension provisions. Mainland China has been re-examined in light of its recent shift toward a more statist orientation and rising regional differences. Nevertheless, due to decreases in health care and housing provision, the overall degree of welfare development slightly slipped according to our findings. Similarly, Japan maintained its exceptionally well-developed healthcare services and old age pension provision. Still, its decrease in public housing provision resulted in an overall reduction of its degree of welfare development.

Figure 2. Degrees of East Asian welfare development, 1990-2016

Note: Figures 0 (fully out) – 1 (fully in) in the X-axis depict degrees of membership of a ‘welfare development’ fuzzy set. For more details, please refer to: Yang & Kühner, 2020.

Besides considering social spending or social rights data to measure East Asian welfare development, welfare outcomes can also be analysed by making use of recent advancements in the standardisation of world income inequality databases.  

Disposable GINI index scores (a measure of the distribution of income across a population) suggest variations in income inequality in East Asia despite a common, increasing trend. Hong Kong and mainland China have ‘high’ inequality according to qualitative categories of the International Labour Organisation. GINI scores in South Korea, Japan, and Taiwan remained comparatively lower than those in mainland China and Hong Kong albeit with different trajectories, particularly since the 2008 Global Financial Crisis: South Korea and Japan have been on an upward trajectory, whereas the opposite has happened in Taiwan. Amongst other things, the consideration of disposable income inequality is useful since it combines the impacts of social security payments and variations in taxation levels, the latter of which are not systematically covered in the analysis of social spending and (most) social rights approaches.

More interestingly, however, is that income redistribution, measured by comparison of GINI scores based on market and disposable incomes, appear firmer in Hong Kong and China (and Japan) compared to Taiwan and South Korea. In other words, expansion of social spending and social rights in the latter cases seem to have had little bearing on the ability of welfare policies to alleviate income inequality in the two leading economies in East Asia. By contrast, it is mainland China and Japan, two cases which have been characterised by decreasing welfare development, that show the most significant decreases in income inequality after taxes and transfers. Lastly, redistribution has remained virtually unchanged in Hong Kong since 1985, but at a higher level than the portrayal of Hong Kong’s strongly means-tested, safety-net approach to welfare policy may have suggested (Figure 3).

Figure 3. Income redistribution in five East Asian economies, 1985-2020

Source: Standardized World Income Inequality Database, Version 9.4, November 2022.

Lastly, among the five East Asian economies, Japan (28 percent) and South Korea (32 percent) had the lowest total pre-tax labour income share of the female adult population in 2019, which the World Inequality Database presents as an indicator of gender inequality; here, it is Hong Kong (38 percent) and Taiwan (36 percent) that fared best. It is worth noting that none of the above distinctions in social outcomes would be visible if social expenditure were to be considered on its own.

3. Addressing the dependent variable problem within East Asian social policy analysis

The current state-of-the-art of East Asian social policy studies presents an abundance of theoretical conceptualisation and a scarcity of empirical investigation. Moreover, the challenge of identifying the elements of a distinct East Asian welfare model has been exacerbated by East Asian nations and territories embarking on various pathways of welfare development before and during the COVID-19 pandemic. I argue that addressing the ‘dependent variable problem’ within East Asian social policy analysis is crucial to managing several persistent biases in the macro-comparative analysis of East Asian welfare development.

More precisely, as this article demonstrates via concrete empirical examples, how macro-level social spending, social rights, and social outcomes data suggest different degrees and trajectories of welfare development across five East Asian economies. The picture becomes even more complicated when moving beyond aggregated analysis, considering the directions of change in specific welfare policy areas. For instance, Japan has the highest degree of social spending, welfare generosity, and redistribution, but fares worst in a more comprehensive analysis of ‘welfare development’ taking into account a broader range of policy domains and mixing quantitative and qualitative data. South Korea spends more on social security, employment, and health care policy than Taiwan. Still, it has a lower level of welfare generosity/decommodification of unemployment benefits, sickness benefits, and public pensions (and vice versa). Despite comparatively low spending levels, international data suggests that mainland China redistributes incomes more extensively than South Korea and Taiwan (Tables 2 and 3).

Table 2. Degrees of welfare development in five East Asian economies

Note: 1-5 refers to the ranking of East Asian economies according to different indicators from ‘highest’ (1) to ‘lowest’ (5).

Table 3. Trajectories of welfare development in five East Asian economies

Note: trajectories based on author’s qualitative categorisation: – negative; +- neither positive or negative; + positive; ++ strongly positive.

Variations in the findings across domains and indicators also aggravate the straightforward classification of welfare trajectories of the same five East Asian economies. Mainland China appears to have decreased its public welfare provision in housing and education, while significantly increasing its degree of income redistribution. Significant expansion of social spending in South Korea did not seem to have affected welfare outcomes. Taiwan appears to have considerably improved welfare generosity and income redistribution in the context of fluctuating social spending.

In short, the multidimensionality of welfare development and trajectories, which manifests the dependent variable problem, goes against the varying accounts that have focused on ideal rather than real types of East Asian welfare provision and aimed to discern the distinguishing features of East Asia as unique welfare geography at the conceptual level.

Conversely, it should be increasingly standard that researchers of East Asian social policy take the multidimensional nature of welfare development processes more seriously and, wherever possible, conduct studies of diverse social policy areas before reaching encompassing conclusions. In particular, this means, e.g., that the sole focus on protective forms of social policy, such as social security, should be avoided. Similarly, wherever possible, researchers should accompany analysis of social spending trends with those of social rights and social outcomes.

Ideally, research frameworks can be expanded beyond single-epistemology approaches to address the limitations of case-study and statistically-driven research designs. Qualitative case studies frequently focus on the theoretical notion of welfare development, but not on the exact measurement of critical theoretical concepts such as retrenchment/expansion, de/recommodification, and expansion/containment. By contrast, although quantitative large-N research (research with a large number of participants) is generally easier to duplicate, highly-aggregated comparative data can occasionally fail to adequately capture important nuances of welfare development, especially in policy areas that are harder to quantify, such as housing and social care.

All of the above point to the necessity of further data investments for comparative social policy analysis in East Asia since it can still be challenging to find large-scale, comprehensive, and easily accessible data repositories aiding quantitative research across time and space. For instance, the most important international data collections, such as the OECD Social Expenditure Database, Asian Development Bank Social Protection Index, or International Labour Organisation World Social Security Reports, have either included East Asian economies selectively, failed to reach sufficiently back in time to allow historical analyses, or are plagued by missing data (or combinations of all the above).

What is more, most of the issues this article outlines are not unique to East Asian scholarship. Indeed, many scholars are currently attempting to extend the range of comparisons in social policy analysis to include more cases across the Asia-Pacific. While this trend should be applauded without reservation, there are also reasons to cautious:

  • Scholars should always be cognisant to the fact that competing theories on welfare state development exist and often explain the same data across different welfare regimes equally well.
  • Against the backdrop of limited comparative historical data, conceptual ambiguity, and persistent measurement problems, endeavours to expand country samples to include Asia-Pacific cases are likely to accelerate the ‘dependent variable problem’ within comparative social policy analysis.
  • Comparative social policy scholars are well advised to consider the lessons from the surging interest in East Asian welfare transformations since the 1990s. More precisely, they should be conscious of the limitations of research agendas purely driven by concerns about ideal-typical classifications of diverse cases, avoid approaches that overly emphasise the historical particularities of Western (or, in fact, East Asian) countries or territories, and resist the tendency to place themselves into epistemological ‘silos’.

In short, as was argued elsewhere, the future development of theories explaining the welfare development in the Asia-Pacific will require researchers to be increasingly theoretically-versed, data-conscious, and humbler in their claims.


Image: Hong Kong apartments. Credit: Himbeerdoni/Flickr.


dependent variable problem East Asian social policy economy empirical data welfare development