Is there a link between social pensions and a decrease in the number of children born (technically referred to as a decreased “total fertility rate”) in sub-Saharan Africa? This Comment summarises the findings by Göran Holmqvist of the Nordic Africa Institute in a new paper which suggests there is.

People often worry that child-related benefits may increase the number of births; but they seldom consider the reverse side of the argument: that benefits aimed at the elderly might decrease the number of births. The hypothesis is that decisions about having children are influenced by the costs and benefits. People may choose to incur the additional costs associated with having more children in order to provide a better guarantee of care in their old age. But if they are confident that a pension will contribute to that care, then they will choose to have fewer children. Taking into account the assumption made in similar studies undertaken in Organisation for Economic Co-operation and Development (OECD) countries that “children may serve as a parental investment in old age care, and that the existence of public old age pensions reduces that motive”, Holmqvist set out to discover if the same is true in sub-Saharan Africa.

Although there is a large amount of literature available on the subject, which all seems to concur that pensions do lower the fertility rate (i.e. the average number of children born to a woman over her lifetime), Holmqvist found that African countries were less represented due to lack of data and to the fact that the measure of aggregate pension cost in relation to GDP used in these studies was low: “because data availability is scarce, but more importantly its variation is likely to reflect the public expenditure generosity towards a small fraction of the population (state and formal sector employees) rather than the extent to which the population in general can count on some form of public old-age security.”

While most sub-Saharan African countries have some kind of pension system for public and formal sector employees, the vast majority of the population are only active in the rural or informal urban sector and therefore these pension systems have little impact on aggregate fertility rates. But some sub-Saharan African countries have introduced non-contributory pension schemes that benefit the great majority of the elderly – which makes them ‘high-coverage’ systems – and it is these countries which Holmqvist examines in order to test the pension-fertility theory, namely: South Africa, Mauritius, Seychelles, Namibia, Botswana, Lesotho and Swaziland. These countries’ systems all have a comparable coverage rate (above 80 percent of age qualified population), pension benefit as share of GDP per capita (mostly in the range of 15-25 percent) and aggregate pension cost in relation to GDP (roughly 1-2 percent). They have also all stayed in place once introduced which “seems to illustrate the ‘stickiness’ of a social policy instrument such as pensions.”

Holmqvist has reviewed a handful of recent empirical studies that have shown that fertility rates are now declining all over sub-Saharan Africa and that there is a clear rural-urban divide in this development. He further points out that “countries with lower fertility rates also tend to be richer in terms of GDP/capita, more urbanised and with better social indicators.” He adds that “the hypothesis of a pension-fertility link is built on the assumption that fertility choices are influenced by the costs and benefits of having children... and that children do provide some form of old-age care to their parents; if not there could be no substitution effect as pensions are introduced. Furthermore it is a link that is assumed to work through peoples’ perception. For instance, introducing a pension scheme that people in reproductive ages do not trust will remain in place should not be expected to impact on fertility rates. This also implies that, to the extent it takes time for a newly introduced pension system to build up in terms of coverage and credibility, the link to fertility rates would operate with a lag.” Of course, other factors also influence fertility choices, such as values, traditions and norms.

Using diagrams 1 to 3, Holmqvist illustrates the 2005 fertility rates of sub-Saharan African countries as plotted against three key indicators, namely: their GDP/capita, urban share of population and child mortality. The high coverage pension countries are labelled at their data points (even Swaziland and Lesotho are labelled although their systems were introduced as late as 2004 and 2006).

Diagrams 1-3

These diagrams tell us that:

• Lower fertility rates among sub-Saharan African countries are clearly correlated with higher GDP/capita, higher urban share and lower child mortality
• The pension countries are at the lower end in terms of fertility rates and child mortality, and at the higher end when it comes to GDP/capita and urbanisation
• The pension countries are grouped below the trend-lines which shows that these countries are at the lower end in terms of fertility rates also given their level of GDP/capita, urban share and child mortality. Their lower fertility levels can hence not solely be ascribed to the fact that they have reached a higher level of economic and social development as measured here. Measurements of the distance to the trend-lines indicate that their fertility rates are very roughly one unit lower (i.e. one child less per woman) than expected given their position in terms of these three development indicators.

Using diagrams 4 and 5, Holmqvist illustrates that – as postulated – fertility rates in pension countries tended to decline (with some lag) after the introduction of pensions, and more so than in comparable countries (the control group).

Diagrams 4-5

Holmqvist recognises that the correlations illustrated by theses diagrams are far from proving the existence of a causal link between pensions and fertility rates. The usual objections may be raised, such as omission of variable bias, underlying trends in time series, reversed causality etc. But, after applying various econometric tools to address these objections, he shows that a significant link between pensions and fertility remains. The findings indicate that having had a high-coverage pension system for 10 years is associated with a reduction in the fertility rate by something in the range of 0.5 to 1.5 children per woman, depending on model specification.

He also understands that reversed causality bias is also an issue – the argument being that “lower fertility means fewer children to take care of more elderly” resulting in pressure being applied to politicians “from the elderly as well as from their potential care-givers, to set-up public pension schemes. Fewer children would also free-up public resources that could be used for this purpose.” However, using the examples of South Africa and Mauritius, he points out that although reduced fertility rate is a factor that increases the likelihood of the introduction of a pension system, in these countries “pensions were introduced while fertility levels were high and the fertility decline came in the decades after the introduction.” He further demonstrates that in many of the political contexts in which pensions were introduced, “the scope for a majority of voters to directly influence pension policies was likely to be limited.”

In reviewing the results of his analysis, Holmqvist does acknowledge that “a weak spot is the fact that the results depend on such a limited number of countries with ’high-coverage’ pensions; just five sub-Saharan African countries have had high-coverage pensions in place for more than ten years. Secondly the pension variable used has been a rough proxy, as historical data on exact coverage and on pension expenditures is lacking. Thirdly, it has not been possible to operate with a wider range of fertility determinants as control variables without severely affecting the sample size, something that has been dealt with by controlling for fixed country- and time effects.” Nevertheless, the theoretic and empirical evidence does largely support the argument that having a high-coverage pension system in sub-Saharan Africa is associated with a reduction of the fertility rate.

In examining the policy implications of these findings, Holmqvist points out that while a “fertility impact of the indicated magnitude is clearly significant when assessing costs and benefits of introducing old age pension schemes,” he still “does not want to argue that fertility reduction should be the main motive for introducing old-age pensions, as considerations related to poverty, dignity, social cohesion and humanitarian principles are likely to weigh heavier, just as in the case of child-related cash transfers.” He adds that there are many other policy instruments in place that can influence fertility rates, which are “likely to be cheaper and more efficient than pensions in reaching that objective.”

Holmqvist concludes his paper by pointing out that similar old-age security mechanisms seem to be at play in Africa as has been shown to be the case on other continents. “Social pensions for the elderly and child-related cash transfers are both forms of cash transfers that are part of the ‘social protection floor’ advocated by the United Nations, but while the fertility impact of child-related cash transfer in developing countries is heatedly debated, scant attention is paid to the fertility impact of old-age cash transfers. The results presented here serve to remind us that a balanced social protection approach, caring for both elderly and children as proposed by the UN social protection floor, might contribute to a balanced fertility impact as well.”

To read the full paper by Göran Holmqvist, click here.