Given the secular rise in income inequality, economists increasingly worry about its macroeconomic implications. A link between inequality and saving lies at the heart of these considerations. For instance, the debate about secular stagnation has revived the Keynesian idea that rising inequality may exert a drag on demand by increasing the aggregate propensity to save (e.g. Summers 2015).

## Inequality and saving: An important transmission channel

By the same token, a positive relationship between inequality, saving, and capital accumulation is a transmission channel in the neoclassical growth literature (Bourguignon 1981). However, with regard to global current account imbalances, recent studies suggest a negative link between inequality and saving, which could be one reason behind the current account deficits of countries like the US or the UK (Ranciere et al. 2012, Behringer and van Treek 2013).

Theoretically and empirically, the effect of income distribution on aggregate saving is ambiguous, due to at least two opposing effects on the microeconomic level that might be offsetting in the macroeconomic aggregate. On the one hand, richer households tend to have a higher propensity to save than households at the lower end of the income distribution (e.g. Dynan et al. 2004). An increase in income inequality may thus cause a rise in aggregate saving (Keynes 1936). On the other hand, several micro-econometric studies find that middle- and low-income earners lower their saving rate in response to rising top incomes (Drechsel-Grau and Schmid 2014, Bertrand and Morse 2016). By triggering expenditure cascades, an increase in inequality could thus lead to a decline in aggregate saving (Alvarez-Cuadrado and El-Attar Vilalta 2012, Frank et al. 2014).

Given that both mechanisms are backed by evidence on the household level, the overall impact of income distribution has to be assessed by country-level evidence. In a recent paper, we estimate the effect of inequality on aggregate saving based on an unbalanced panel of 29 high-income OECD countries, which we observe between 1961 and 2013 (Bofinger and Scheuermeyer 2016). While preceding studies had to struggle with inequality data that is either full of gaps or hardly comparable in terms of income definitions or household units, recent advances in data availability (Solt 2016) allow us to exploit a harmonised set of net income Gini coefficients for a broad sample of countries and years.

## Results

In line with the inconclusiveness of earlier cross-country and panel-data studies (e.g. Schmidt-Hebbel and Serven 2000, Leigh and Posso 2009), we do not find a linear effect of inequality on aggregate saving. However, we reveal a highly significant hump-shaped relationship that is robust to a large set of control variables, including asset prices and credit availability. As pictured in Figure 1, the impact of inequality on the household-sector saving rate is positive at low levels of inequality, whereas it becomes negative if the Gini coefficient rises above a level of roughly 30, which is the level of net income inequality that Germany is currently approaching and the US always exceeded. This hump-shaped pattern is robust to different estimation techniques and sample compositions, endogeneity concerns, and data uncertainty.

**Figure 1** The marginal effect of inequality on saving at different levels of inequality

*Notes*: Dependent variable is the saving rate of the household sector. Values are calculated from the results of Table 2, Column (4) in Bofinger and Scheuermeyer (2016). The downwards sloping line plots the marginal effect of inequality. Surrounding dotted lines represent the 90% confidence intervals.

## Intuition behind a hump-shaped relationship

An explanation for the decreasing marginal effect of inequality could be given by a non-linear adaption in household consumption behaviour. If inequality only becomes gradually visible, saving rates of poor- and middle-class households possibly remain unchanged, while inequality is still rising from a low level. Thus aggregate saving would be initially dominated by an increasing income share of households with a high propensity to save. As inequality rises further, the former positive effect could be increasingly compensated by a changing behaviour of households from the middle and lower ranks of the income distribution. When inequality becomes more and more visible, the incentive to engage in conspicuous consumption rises until the decrease in saving of poorer households dominates in aggregate.

## Interactions with financial development

Facing a decline in relative income, poorer households may need credit financing to keep up with the rising consumption of their richer neighbours. By introducing interaction terms into our regression models, we test whether the impact of inequality on saving depends on the extent of financial development. Indeed, we find that inequality increases saving, when credit is scarce, while it reduces saving at high levels of credit , as shown in Figure 2. (We find a similar interaction effect between inequality and financial liberalisation, which is measured by the financial reform index from Abiad et al. 2010). Nonetheless, in both a low-credit and high-credit environment, the hump-shaped relationship between inequality and saving prevails.

**Figure 2**. The marginal effect of inequality on saving across different levels of credit availability

*Notes*: Dependent variable is the saving rate of the household sector. Values are calculated using the results of Column (2) of Table (8) in Bofinger and Scheuermeyer (2016). The downwards sloping line plots the marginal effect of inequality. Surrounding dashed lines represent the 90% confidence intervals. Vertical lines indicate the distribution of the credit to GDP ratio in the sample: dotted lines mark the first and 99th percentiles, the dashed line marks the median value.

## Implications for current account imbalances

Expenditure cascades and Keynesian consumption theory relate to household behaviour, which is why we primarily focus on the saving rate of the household sector. Yet we also find some evidence for a non-monotonic effect of inequality on private saving, national saving, and the current account balance. In fact, the current pattern of global imbalances corresponds quite well to the hump-shaped relationship between inequality and saving – surplus countries like Germany or Sweden have experienced an increase in inequality starting from a low level in the 1980s, while in deficit countries like the US or the UK, inequality has risen from a higher level.

## Conclusion and outlook

Drawing on data from 1961 to 2013, we find that the relationship between inequality and aggregate saving is non-monotonic and hump-shaped. Moreover, the marginal effect of inequality on saving depends on credit availability and financial liberalisation.

Yet, for the future, a permanent compensation of income losses via credit financing is hardly conceivable. Consequently, as soon as saving rates of low and middle income households have reached a floor at zero, it is likely that the Keynesian effect of rising income concentration will dominate. We thus may have to worry about a lack of aggregate demand.

## References

Alvarez-Cuadrado, F, and M El-Attar Vilalta (2012), “Income inequality and saving”, IZA Discussion Paper 7083, Institute for the Study of Labor.

Behringer, J, and T van Treeck (2013), “Income distribution and current account: A sectoral perspective”, IMK Working Paper 125, Macroeconomic Policy Institute.

Bertrand, M, and A Morse (2016), “Trickle-down consumption”, *Review of Economics and Statistics*, forthcoming.

Bourguignon, F (1981), “Pareto superiority of unegalitarian equilibria in Stiglitz' model of wealth distribution with convex saving function”, *Econometrica*, 49 (6), 1469-1475.

Bofinger, P, and P Scheuermeyer (2016), “Income Distribution and Aggregate Saving: A Non-Monotonic Relationship [5]”, CEPR Discussion Paper No. DP11435.

Drechsel-Grau, M, and K D Schmid (2014), “Consumption savings decisions under upward-looking comparisons.” *Journal of Economic Behavior and Organization,* 106, 254-268.

Dynan, K E, J Skinner, and S P Zeldes (2004), “Do the rich save more?”,* Journal of Political Economy*, 112 (2), 397-444.

Frank, R H, A S Levine, and O Dijk (2014), “Expenditure cascades”, *Review of Behavioral Economics*, 1 (12), 55-73.

Leigh, A, and A Posso (2009), “Top incomes and national savings”, *Review of Income and Wealth*, 55 (1), 57-74.

Ranciere, R, N A Throckmorton, M Kumhof, C Lebarz, and A W Richter (2012), “Income inequality and current account imbalances”, IMF Working Paper 12/8, International Monetary Fund.

Schmidt-Hebbel, K, and L Serven (2000), “Does income inequality raise aggregate saving?”, *Journal of Development Economics,* 61 (2), 417-446.

Solt, F (2016), “The standardized world income inequality database”, *Social Science Quarterly,* forthcoming.

Summers, L H (2015), “Demand Side Secular Stagnation”, *American Economic Review*, 105 (5), 60-65.