The mystery of Chinese savings

Shang-Jin Wei, 6 February 2010



Much attention has been directed toward China’s high savings rate (Broda et al. 2009, Prasad 2009, Reisen 2009). Not only is the savings rate disproportionately high compared to virtually any other country, but it directly impacts China’s current account surplus and the US consumer debt and trade deficit. When national savings exceeds investment, the excess savings becomes China’s current account surplus.

Given its far-reaching effects, both private sector analysts and policymakers have attempted to trace the causes of China’s high savings rate and to predict how long it will last. Some have attributed the savings primarily to Chinese corporations. Others point to a precautionary savings motive – as Chinese are worried about costs of healthcare, education, and old-age pensions and are unsure about how much these costs might change over time, they save more.

But these explanations may not be the most important part of the story. For example, while the Chinese corporate savings rate is high, the pattern is consistent with many countries. For example, Korea and Japan always have an even higher corporate savings rate than China. In fact, corporate savings rates in most countries have experienced a steady rise in the recent decades. In research with Tam Bayoumi and Hui Tong (2009), we show that to understand why China’s national savings rate is so high, the corporate sector is the wrong place to start.

A marriage proposal

It is the high Chinese household savings that has no equal among major economies. As to explanations related to poor social safety net, they have trouble being reconciled with the improvements in the pension system, health insurance, and other government programs in recent years. Yes, the Chinese social safety net is still poor. Yes, part of the Chinese savings is to make up for the inadequacy of the social safety net. Yes, there may be some increase in the cost of healthcare. But, as the social safety has improved and the insurance coverage has expanded over the last decade or so, we should expect household savings to decline, or at least not increase. Yet, household savings as a share of disposable income almost doubled from 16% in 1990 to over 30% today.

In my recent research paper with Xiaobo Zhang (Wei and Zhang 2009), we hypothesised that a social phenomenon is the primary driver of the high savings rate. For the last few decades China has experienced a significant rise in the imbalance between the number of male and female children born to its citizens.

There are approximately 122 boys born for every 100 girls today, a ratio that means about one in five Chinese men will be cut out of the marriage market when this generation of children grows up. A variety of factors conspire to produce the imbalance. For example, Chinese parents often prefer sons. Ultra-sound makes it easy for parents to detect the gender of a foetus and abort the child that’s not the “right” sex for them, especially as China’s stringent family-planning policy allows most couples to have only one or two children.

Our study compared savings data across regions and in households with sons versus those with daughters. We found that not only did households with sons save more than households with daughters on average, but that households with sons tend to raise their savings rate if they also happen to live in a region with a more skewed gender ratio. Even those not competing in the marriage market must compete to buy housing and make other significant purchases, pushing up the savings rate for all households.

Cross-regional evidence

In Figure 1, we have plotted the time series of the (standardised) Chinese private sector savings rate (in red), defined as GDP less private and government consumption over GDP, from 1975 to 2005 against the (standardised) gender ratio at birth lagged by twenty years (i.e., 1955-1985) (in blue). There is a strikingly close connection between the two variables. The savings rate started to shoot up around 2002 just as the gender ratio for the marriage-age cohort began to be seriously out of balance. We show that this is more than a coincidence.

Figure 1. Chinese private sector savings rate vs private savings 1955-2005

First, in panel regressions across 30 Chinese provinces during 1990-2007, the local savings rate tends to be higher in regions and years in which the local gender ratio is higher. This continues to be true after we control for local income level, income inequality, the enrolment in the social security system, the age profile of the local population, and the province and year fixed effects.

Second, we recognise possible endogeneity of, and measurement error in, local gender ratios, and employ an instrumental approach. Gender ratio imbalance comes primarily from sex-selective abortions. This, in turn, results from a combination of parental preference for sons, and some limit to the number of children a couple is allowed or wants to have (which for the Chinese is a strict family-planning policy).

As suggested by other scholars, we use two measures of local financial penalties for violating family-planning policies – set more than a decade earlier – as instrumental variables for local gender ratio imbalance for the pre-marriage age cohort. As an extension, we also add the fraction of the local population that is legally exempted from the family-planning policy as an additional instrument. With either set of instruments, the effect of local gender ratios on local savings rates remains positive and statistically significant. In fact, the point estimate becomes larger. This suggests that an increasing imbalance in the gender ratio causes a rise in the savings rate.

Based on the point estimate in the instrumental regression, a rise in the gender ratio for the pre-marital age cohort from 1.05 to 1.14 (which is the mean increase across the provinces from 1990-2007) would lead to a rise in the savings rate by 6.7 percentage points, which is about 42% of the actual increase in the savings rate. If we run separate regressions for rural and urban areas, we find that the elasticity of the local savings rate with respect to the local gender ratio is larger in rural areas than in urban areas. An increase in the local gender ratio from 1.05 to 1.14 in the rural areas (the actual mean increase from 1990 to 2007) could account for about 49% of the actual increase in the savings rate.

Household-level evidence

We then examine household data using household surveys that cover 122 rural counties and 70 cities in 2002. While households with a son typically save more than households with a daughter, we do not regard this per se as supportive evidence of our hypothesis, since other channels could account for this difference. Instead, the evidence that we find more compelling is that savings by otherwise identical households with a son are greater in regions with a higher local gender ratio. This is something clearly predicted by our hypothesis, but not by any other existing explanations. In addition, we find that savings by households with a daughter do not decline in regions with a high gender ratio. This is consistent with the possibility pointed out in the model that the two opposing effects approximately cancel each other out. It is also consistent with the possibility that the savings pressure on households with a son spills over to other types of households.

With household-level data, we can control for a variety of household features that allow us to measure the importance of lifecycle factors (e.g., the age structure of the family members) and the precautionary savings motive (education level of the household head, whether any member of the household works in a government sector, has lost a job, or has experienced major illness). We find no robust support for the lifecycle hypothesis and some support for the precautionary savings motive. Indeed, the quantitative effect of a rise in the gender ratio on the savings rate is unaffected by controlling for these factors.

Still, could the local gender ratio reflect some omitted or unobserved variables that also affect the household savings decision? One may imagine that a region with more intrinsic income uncertainty, or a greater local aversion to a given uncertainty, may simultaneously exhibit a higher local gender ratio imbalance and a higher local savings rate. Can we rule this out? Yes. A pure location-specific shock should affect savings by all households in the same region in the same way. But that is not what we find. Instead, only the savings by those households with a son react strongly and positively to a rise in the local gender ratio, while savings by households with a daughter do not.

The next possibility is far more challenging. Could a gender ratio imbalance reflect something that is both location and household specific? For example, a region may have an unusually high level of income uncertainty that is common to all households, but some households care about this more than others. Those with a stronger aversion to uncertainty may engage in a sex-selective abortion more aggressively and save more at the same time. By construction, selection at both household and location level is much harder to rule out since our unit of observation is at the same level. But there are good reasons to think that if we focus on households with a single child, such selection is unlikely to be quantitatively significant. Ebenstein (2009) shows that gender ratio imbalance is overwhelmingly a result of sex-selective abortions at higher orders of birth. That is, the gender ratio for first-born children is close to normal. This is particularly true in rural areas. Since a second child is officially permitted if the first child is a girl, and since many families exhibit a preference for a balanced gender ratio (one boy and one girl) over having two boys, there is very little reason to perform sex-selective abortions on the first pregnancy. However, the gender ratio at birth goes up substantially over time for the second-born children and becomes even more skewed for higher order births. This pattern is also pointed out by Zhu, Lu and Hesketh (2009) in an article in the British Medical Journal. This suggests that the first son (or daughter) is unlikely to result from a sex-selective abortion. In our empirical examination, when we restrict attention to households with only one child, we still see that those with a son exhibit a strongly positive elasticity of savings with respect to the local gender ratio, but those with a daughter do not. Furthermore, savings by a household with a son are more sensitive to the gender ratio in rural areas.

Policy implications

While the conventional explanations for the high savings rate all play a role, this new research indicates those explanations are not as important as people previously thought. While sociologists and other social scientists have looked at the gender ratio imbalance as a social problem, they have not looked at it in relation to the high Chinese savings rate. Similarly, as economists and policymakers have looked with concern to the large Chinese current account surplus and large US current account deficit, or global imbalances, much of their discussion has focused on changing exchange rate policy.

None of the discussion about global imbalances has brought family-planning policy or women’s rights to the table, because many do not see these issues as related to economic policy. Our research suggests that this is a serious omission. You can only implement the right policy when you have the appropriate diagnosis, and fruitful policy dialogue has to include discussion on these issues.

Editor’s note: This piece is an expanded version of the author’s byline publication on, which in turn is based on his research paper with Xiaobo Zhang (2009)


Bayoumi, Tam, Hui Tong, and Shang-Jin Wei (2009), “The Chinese Corporate Savings Puzzle: A Little International Comparison Can Go a Long Way”, Unpublished working paper, IMF and Columbia University.

Broad, Christian, Piero Ghezzi, and Eduardo Levy-Yeyati (2009), “The new global balance: Financial de-globalisation, savings drain, and the US dollar”,, 22 May.

Ebenstein, Avraham (2009), “Estimating a dynamic model of sex selection in China,” Harvard University mimeo.

Prasad, Eswar (2009), "Getting past the blame game”,, 28 January.

Reisen, Helmut (2009), “On the renminbi and economic convergence”,, 17 December.

Wei, Shang-Jin and Xiaobo Zhang (2009), “The Competitive Saving Motive: Evidence from Rising Sex Ratios and Savings Rates in China”, NBER Working Paper 15093.

Zhu, Wei Xing, Li Lu, and Therese Hesketh (2009), “China’s excess males, sex selective abortion, and one child policy: analysis of data from 2005 national intercensus survey,” British Medical Journal, (338):b1211.


Topics: Frontiers of economic research
Tags: China, global imbalances, savings glut


The mystery of Chinese savings

This article, though interesting, does no more than identify a correlation between hige savings rates and families with only one male child. What it does not do, or even attempt to do, is to provide any coherent explanation of the claimed linkage. Without knowing what it is, it is very difficult to see what useful policy lessons can be drawn from it.

Social Institutions and Gender Index (SIGI)
Please consult our innovative dataset (which puts China last in gender bias on the 'son preference index', e.g.) to underpin Shang-Jin Wei's innovative analysis.
Helmut Reisen, OECD Development Centre

Shang-Jin Wei

Professor of Finance and Economics, and of Chinese Business and Economy, Graduate School of Business, Columbia University