Concern that too little saving is done by a significant number of UK households has long been a motivation for government policy. This is particularly the case for retirement saving: the Pensions Commission estimated in 2004 that perhaps nine million people were under-saving for their retirement. The issue is potentially wider than retirement saving alone: recent IFS research found that the median family had little more than a thousand pounds in liquid financial wealth in 2005. However, any intervention by policymakers designed to increase household saving rates should ideally be based on high quality evidence on the efficacy of different policies.

Today, a new British Academy policy centre report, authored by IFS researchers, is being launched. The report examines the evidence base for four different types of policy to promote household savings:

• Using financial incentives such as tax-favouring and matching;
• Financial education, training and information provision;
• Choice architecture and 'nudges';
• Social marketing, using techniques from commercial marketing to promote social goals.

The report critically assesses the literature on whether policies in these areas, both in the UK and abroad, have been successful in encouraging households to save more. Many of these policies have been actively pursued in the UK. Financial incentives include generous tax treatment for pension saving, and matching policies such as the previously piloted Saving Gateway. The government has recently consulted on a review of Personal, Social, Health and Economic education in schools which could recommend making financial education a statutory component. Under choice architecture, 'auto-enrolment' into employer provided pensions is set to be rolled out from October. This compels employers to default most workers into a pension scheme from which employees can subsequently choose to opt-out, rather than the current system into which people typically need to opt-in.

The report concludes that, despite the obvious importance attached to the issue in the UK and internationally, significant gaps in the evidence base for whether policymakers can raise overall household saving remain. There are of course some high quality exceptions, but relatively few studies are able to combine credible counterfactuals (what would happen to saving and wealth in the absence of the policy?) with good measures of total savings. For example, there is convincing evidence that financial incentives can raise the amount saved in the favoured savings vehicle. There is much less compelling evidence on whether these incentives increase total savings or just represent assets being shifted from one form of saving to another.

Similar issues arise in assessing 'nudge'-inspired policies like changing default options for pensions saving. There is plenty of evidence that auto-enrolment raises participation in pension schemes (often substantially) but much less sense of the extent to which it raises total saving, if at all. Auto-enrolment can lead some people to save less than they otherwise would have done: most people who auto-enrol stick to the default contribution levels and investment funds which are often quite conservative.

Many studies assessing financial education or training look at the impact of policies on financial knowledge or on intended changes to savings behaviour, but there is not much good evidence that these outcomes translate into genuinely higher household saving rates.

Some other key conclusions from the study are:

• Not much evidence can be found on the long-term impacts of these policies, and whether they help entrench saving habits;
• Interventions are sometimes packaged together (such as combining matching with financial education). It is hard to disentangle which aspects of the package are effective in isolation, and whether the impact of the policy as a whole exceeds the sum of its parts;
• Relatively little evidence, except for that around financial incentives, is specific to the UK;
• The possible use of 'social marketing' to promote savings has been little-researched.

While the evidence base is in general limited, this is not a reason for inaction if there is a genuine need for intervention. Indeed, it presents something of an opportunity for the government. The move to auto-enrolment in the UK, for example, ought to be robustly evaluated, to provide evidence not just on how the UK scheme should be improved but also for other countries thinking of similar reforms. New pro-savings policies should also be introduced in ways which allow their effects, short- and long-term, to be assessed properly, ideally through randomised trials. This requires proper measurement of total saving and wealth, considering possible spillovers to those not directly affected (for example, some studies of workplace financial education show an effect on the saving behaviour of non-treated workers or family members outside work), and perhaps allowing for random variation in the way policies are 'framed' which also appears to affect behaviour.

Of course while trials are important they may not always be appropriate or easy to implement. Thus evidence in this area needs to feed into the continued development of theoretical models of saving behaviour. Together with good household-level data on saving and assets (which has been much improved in the UK by the introduction of the Wealth and Assets Survey in 2006), this will allow researchers to assess the likely impact of policy compared to a modelled counterfactual even when trials are not feasible. Using multiple, complementary approaches in this way will hopefully mean a significant improvement in the evidence base in the future.