VoxEU Column Poverty and Income Inequality

Recessions and size of the middle class

A sizeable middle class is essential to protect societies against socioeconomic and political instability. This column examines the effect of economic recessions on the size of the middle class using different income-based and self-perception definitions. It finds that anticipated recessions do not produce an overall middle-class squeeze, but unanticipated shocks such as the Great Recession do. It also finds that recessions increase the share of the population that regards itself as middle class.

A sizeable middle class is essential to protect societies against socioeconomic and political instability (Thurow 1984). A larger middle class size implies greater support for democracy (Barro 1999), more innovation, entrepreneurship, and productivity (Acemoglu and Zilibotti 1997), human capital accumulation (Alesina and Perotti 1996, Doepke and Zilibotti 2005), and upward mobility (Easterly 2001). 

Macroecomic conditions and the middle-class squeeze

Unexpected ‘economic shocks’, such as those associated with the Great Recession,  can produce a ‘middle-class squeeze’. This is especially the case for employment shocks, some of which are not anticipated. However, the effect of a recession is dependent on country-specific social protection institutions that help to accommodate the effects of unemployment shocks. Adverse income shocks from unemployment during recessions can be absorbed by unemployment insurance and subsidies alongside other welfare benefit spending, or via wage bargaining and labour regulations. However, the effect of a recession also depends on its duration and depth. While short-lasting recessions might have negligible effects that can easily be alleviated by credit mechanisms to counteract liquidity constraints, the mitigating effect of such mechanisms might not persist when recessions are long-lasting. Finally,  some recessions give rise to specific policy interventions, such as austerity cuts, which reduce the potential mitigating effect of employment insurance, as was in the case in the Great Recession.

Measuring the middle class 

Measures of the size of the middle class are divided between income- or consumption-based measures (for example, defined by an income cut-off point relative to a poverty measure or by the income distribution in a given population) and those based on broader social definitions (for example, involving occupational or class status, and income self-perception). Both definitions capture different features of what we mean by ‘middle class’ (Atkinson and Brandolini 2013). Unsurprisingly, sociological definitions that reflect individual identity as middle class are generally more stable measures of occupational or self-perceived status, while income-based definitions are typically time-varying and heterogeneous depending on the definition of ‘income’ adopted. 


In a new paper (Batinti and Costa-Font 2019), we use the Luxembourg Income Study (LIS) data to estimate the size of the middle class based on overall disposable income, labour income, and market income, and explore if and how our estimates change depending on the definition of income.  We use two measures:

  • an ‘inequality-based’ definition, which is based on whether an individual’s income falls (i) between the 2ndand 4th quintile (2to4) or (ii) within the 3rd quintile of the income distribution (3); and 
  • ‘size-based” measures, which estimate the proportion of the observations (population-weighted) of the sample within predetermined income brackets (e.g. within 75% and 125% of the median income) (Atkinson and Brandolini 2013).

We also employ a measure self-perception of middle-class status by using both the World Value Survey (longitudinal files covering Waves 1 to 6 for the  1981-2014 period) and the European Values Study (longitudinal data covering waves 1 to 4 for the period 1981 – 2008), referred henceforth as an integrated values study (IVS).


We do not find evidence of a ‘middle-class squeeze’ resulting from an unexpected employment shock (in this case, an economic recession interacting with the unemployment rate). However, when we explore the effect of a largely unanticipated shock such as the Great Recession we do find evidence of squeezing, but only on labourincome. This is robust to alternative specifications controlling for the depth and duration of the recession. Nonetheless, the effect is heterogeneous across income-based middle-class definitions. 

Moreover, we find that the middle-class squeeze is mitigated if the recession hits at higher levels of unemployment. A similar effect is detected for the countries with the highest spending on unemployment insurance, which is consistent with the expected impact of partial unemployment insurance. Importantly, we also find that unemployment shocks during recessions increase the share of the population that defines itself as middle class.

Importantly, we find that a larger share of the population regards itself as middle class after a recession. Our interpretation of this is that employment shocks might increase the likelihood of those individuals at the top end of the income distribution viewing themselves as middle class, given the higher income threshold income for belonging to the middle class.

Policy implications

If governments wish to attain desirable policy objectives traditionally associated with the size of the middle class  – such as innovation, democratic stability, and human capital accumulation – only those income losses resulting from unanticipated employment shocks, such as those emerging from the Great Recession, are a cause for concern. This is especially true for policies targeting labour income losses as opposed to overall income. In particular, we observe that such effects that occurred during the Great Recession were generally accompanied by welfare spending cuts.

Another significant result for policymaking is that mechanisms of income protection in the form of unemployment insurance seem to achieve their goal of protecting individuals against unexpected unemployment shocks, and hence ensure both the socioeconomic stability and consumption smoothing of the middle class overtime. Further studies might wish to explore in more depth why ‘this time was different’, i.e. why only recessions within the wave of the Great Recession’s produced a squeezing effect, as our evidence does not support this to be the case in general. 

Our results suggest that recessions exert heterogeneous effects across the income distribution, driving people from the middle class to the lower income groups while at the same time driving people with higher income towards the middle.


Acemoglu, D and F Zilibotti(1997), “Was Prometheus Unbound by Chance? Risk, Diversification and Growth”, Journal of Political Economy 105(4): 709-51 

Alesina, A and R Perotti (1996), “Income Distribution, Political Instability and Investment”, European Economic Review 40(6): 1203-1228

Atkinson, A B and A Brandolini (2013), “On the identification of the middle-class”, in J C Gornick and M Jäntti (eds), Income inequality: Economic disparities and the middle-class in affluent countries,  Stanford University Press, pp.  77-100.

Batinti, A and J Costa-Font (2019), “Do economic recessions ‘squeeze the middle-class’?”, forthcoming in Economics and Politics.

Banerjee, A V and E Duflo. (2008), “What Is Middle-class about the Middle-classes around the World?”, The Journal of Economic Perspectives 22(2): 3.

Barro, R J (1999), “Determinants of Democracy”, Journal of Political Economy 107(6): 158-83.

Doepke, M and F Zilibotti (2005), “Social Class and the Spirit of Capitalism”, Journal of the European Economic Association 3(2-3): 516-524. 

Easterly, W (2001), “The Middle-Class Consensus and Economic Development”, Journal of Economic Growth 6(4): 317–35.

Thurow, L (1984), “The Disappearance of the Middle-class”, New York Times, 5 February. 

1,050 Reads