VoxEU Column Macroeconomic policy

Precautionary strategies and household saving

An implication of the ‘precautionary saving’ hypothesis is that in countries faced with more macroeconomic volatility and risk, private saving should be higher. In the observable data, however, there is a negative correlation, particularly in developing countries. This column offers a plausible explanation for the disconnect between the precautionary theory and the empirical evidence,  based on a model with a richer account for the various modes of ‘precautionary’ behaviour by private agents, in cases where institutions are weaker and labour informality is prevalent.

Why do people save? This basic question has occupied economists for decades. One influential answer has emphasised the role of ‘precautionary’ motives; i.e., private agents save in order to mitigate unexpected future income shocks (e.g. see a recent discussion in Cronqvist and Siegel 2015).

An implication of the emphasis on precautionary motives is that in countries faced with more macroeconomic volatility and risk, private saving rates should be higher (ceteris paribus). It is, however, commonly observed that saving rates in Latin America are low, and more specifically are significantly lower than in East Asia, even though Latin America’s macroeconomies are generally much more volatile. A careful examination of the global data confirms that, instead of the hypothesised positive correlation between volatility and private saving rates, we observe a negative correlation. In the research project summarised here (Aizenman, Cavallo and Noy 2015), we provide a plausible explanation for the disconnect between ‘precautionary saving’ theory and the empirical evidence that is based on a model with a richer account for the various modes of precautionary behaviour by private agents, particularly in developing economies.

Common features of developing countries are the under-development of the financial system, weaker institutional frameworks, relatively high volatility of macroeconomic aggregates, and the absence or shallowness of safety nets mitigating households’ exposure to risk. These factors have profound implications for the functioning of firms and the choices made by households. The institutional environment in which firms operate in developing countries induces informality in labour markets. This informality, together with a difficulty in establishing collateral, prevents entrepreneurs from accessing formal credit. Thus, entrepreneurs wishing to operate businesses resort to self-financing, as this is regularly the only feasible option available to them. Both self-financing and small-scale, common features of informal firms in developing countries, connect the firms to households’ saving and investment decisions. This in turn suggests that the neoclassical discussion on precautionary saving that disconnects firms from households and investment from saving is too narrow.  

The empirical evidence

We use private saving data from the World Economic Outlook –162 countries in the period 1980-2007 in decadal averages – and, for a smaller subsample, data on Household saving rates from Bebczuk and Cavallo (2014). Consider Figure 1, which plots the partial correlation between saving and risk (partial in the sense that we control for other determinants of saving). Country codes are included to give the reader a sense of where different countries stand; the number next to the three-digit country code indexes the decade (i.e., 1 = 1980s, 2 = 1990s, and 3 = 2000s). The plot shows a negative correlation between private saving and real GDP per capita volatility that is not driven by outliers.

Figure 1. Partial correlation – private saving and realised past volatility

Informality with individual shocks

Our explanation for this puzzle is demonstrated in a stylised overlapping generation model focusing on the impact of financial market imperfections on informality, a household’s education investment, and income decisions in the presence of macroeconomic uncertainty.  The model extends the Galor and Zeira (1993) framework. We assume a two period life span for each member of the household, where consumption and bequest to the household’s children take place in adulthood, the second and last period of life.  Each generation lives for two periods.

In the first period, youth, children have the opportunity to enter the labour force as unskilled workers, earning in both periods of their lives the low unskilled wage. Alternatively, they may acquire human capital, thereby becoming skilled workers in the second period – providing them the capacity to run and manage the family firm. In the second period, adulthood, unskilled workers will keep earning the low unskilled wage. The educated workers are employed running the family firm.  They either take on an existing family business (replacing their parent), or start a ‘greenfield’ family firm. The second period is also the time when children are raised, and bequest and children’s education decisions are made.

We think of bequests as the adults’ decision to save in this setting. Bequests help to finance investment in human capital (adding physical capital to the model does not change the results). The model yields some relevant thresholds (which are in turn functions of interest rates, productivity, etc.); such that if the bequest in a period is above a threshold, over time the saving rate will converge to a relatively high steady state equilibrium. Instead, if the bequest is below the threshold, then the saving rate will converge overtime to a lower steady state. As in Galor and Zeira (1993), we find that differential bequests across otherwise identical households may lead to divergent long run outcomes, where the long run equilibrium is characterised by a bipolar distribution of households converging to a low-saving equilibrium and more affluent households converging to a high-saving equilibrium.

Macroeconomic shocks

Our principal contribution is to focus on the impact of the realised macro shocks taking place in the second period on the actual bequest determined in the first period, drawing on the association between the bequest motive in period 1, and the actual bequest in period 2, determined after the realisation of macro shocks. 

In short, the impact of recurrent negative shocks is to move the relevant thresholds to the right; meaning that for any given bequest (i.e., saving over the period) it is more likely that a household will end up in the low-saving equilibrium. This will push a larger mass of individuals to the low-saving equilibrium; thereby resulting in a lower aggregate private saving rate in the economy.  Therefore, a key insight of this model is that higher economic volatility may lead to lower aggregate saving rates. This is not at odds with precautionary saving theory; it simply expands the theory to encompass other modes of ‘precautionary’ behaviour by private agents (see Aizenman and Noy 2013, for more long-run analysis).

Concluding remarks

We have focused on the observed but counter-intuitive negative correlation between volatility and private sector saving. The standard view hypothesises a positive correlation, in that private actors living in countries with a more volatile macroeconomy or higher aggregate risk are expected to engage in more precautionary behaviour and save more. In fact, we observe that the statistical association is reversed, and find lower private saving rates observed in more volatile places. We suggest a theoretical model that incorporates the limiting institutional environment that characterises middle- and low-income countries, and explore the implications of these constraints to private sector saving behaviour.

Two important issues merit further consideration. First, our theoretical and empirical analysis assumed that the degree of aggregate risk is not itself dependent on the private sector’s saving choices. There are good reasons to motivate a more complete investigation of the feedback mechanisms between these two aggregates. Second, we expect that some of the considerations outlined above will change in more open economies. Our full model assumed a form of international labour mobility, but we did not examine the implications of more open capital markets. Further capital market openness may ameliorate some of the problems inherent in the shallow credit markets of lower income countries, but may also generate further sources of volatility that needs to be accounted for.

To summarise, our central message is that the interaction between saving behaviour, broadly construed, and aggregate risk and uncertainty, may be more complex than is frequently assumed, and we aimed to demonstrate that using both the observed empirical regularities and a formal model.


Aizenman J and I Noy (2013) “Public and private saving and the long shadow of macroeconomic shocks”, VoxEU.org, 29 May.

Aizenman J, E Cavallo, and I Noy (2015) “Precautionary strategies and household saving”, NBER Working Paper #21019.

Bebczuk R andE Cavallo (2014) “Is business saving really none of our business?”, IDB Working Paper No. 523, Inter-American Development Bank.

Cronqvist H and S Siegel (2015) “The origins of savings behavior”, Journal of Political Economy 123(1): 123-169.

Galor O and Y Zeira (1993) “Income distribution and macroeconomics”, Review of Economic Studies 60: 35-52.

525 Reads