I recently proposed that any assessment of a country’s statistical capacity be structured around the functions of government, such as those offered by the UN statistical office here. When this list is fully expanded, it includes all of the data that advanced countries like the US or Japan use to manage government and inform citizens. Most developing countries will fall below such an ambitious standard. So how should investments in improved statistical capacity be prioritized?
Higher priority should be attached to filling statistical gaps that are low in cost and high in informational value. Consider the cost of gap filling. This concept is pretty straightforward, though the cost will vary enormously from one gap to the next, depending for example on whether filling it will require the design and deployment of a completely new data collection instrument or, at the other extreme, only adding a single question to an existing routine survey.
Now consider the value of information, a decidedly less straightforward concept. At a recent meeting of the Partnership for Statistics for the 21st Century (PARIS21), Chris Gingerich of the Bill and Melinda Gates Foundation (@ChrisJGingerich) raised the question of whether a dollar value could be attached to the benefits of filling any statistical gap. What is the value, he asked, of statistical information?
The study of the Value of Information (VOI) originally grew out of the field of decision theory as pioneered in the 1950s by Howard Raiffa (here and here). Today Raiffa’s intellectual descendants work in operations research, engineering, climatology, game theory and specifically in applied health economics. (See the work by Karl Claxton and Mark Sculpher here and here. Also see Macauley and Laxminarayan’s summary of an edited volume on the value of information in health and environment, to which I contributed.)
From the decision theory perspective, information has the most value when it can influence a decision, and when the consequences of the decision are, in some sense, “large”. For example, a recent presentation at the PEPFAR Scientific Advisory Board by Tim Hallett of Imperial College London showed that in Africa a given HIV prevention budget prevents 16% more infections when targeted at the neighborhoods where risk of HIV infection is largest than if the funds are spent across the entire adult population. In this case, the “value” of the information about the risk of infection in each neighborhood would be equal to the increase in the expected return on investment in targeting prevention efforts. One might define the return by valuing those averted infections in dollar terms (e.g. using the human capital or the willingness to pay approach) or in DALY terms (if one is content with cost-effectiveness rather than cost-benefit analysis). In either case, one arrives at a measure of the priority that should be placed on having HIV survey data that is sufficiently granular and frequent to guide targeted interventions.
But how would one value an improvement in the information on gross national product, consumer price inflation, housing prices, crime statistics, health facility utilization and the many other statistical domains? To my knowledge, decision theory and VOI have never been formally applied to estimate the dollar value of improvements in a national statistical system. Research on this topic could pay dividends for the data revolution movement.
To start, one might distinguish four sets of decision makers whose choices might be improved by better statistical information: citizens, government, foreign investors and donors. An important attribute for any given statistic would be the degree to which that statistic (if measured with sufficient frequency, spatial granularity and precision) could improve decisions by each of the above four groups. Here are some preliminary thoughts on this approach:
Citizens. When statistical information is released to the public through a vigorous open government mechanism it can help citizens directly. Citizens need data both to hold their government accountable and to improve their private decision-making. (On the CGD website, see discussions of the value of public disclosure for climate policy here and for AIDS foreign assistance here.)
Governments. Statistical information helps citizens indirectly when it enables elected representatives and government to improve decisions on the public’s behalf. Targetting HIV prevention to epidemic hotspots, as proposed by Tim Hallett, is an example.
Foreign investors. Statistics on the prices and availability of key factors of production, such as labor, energy, transportation and even weather, can reduce investor uncertainty and thus encourage foreign investment.
International donors. The call by international donors to improve statistics on population well-being is welcome, since it suggests they would use better data to verify that development objectives are achieved and to enable cash-on-delivery type contracts.
In order to prioritize investments in statistical capacity, a study like that being launched by PARIS21 could experiment with using a panel of experts to score on a scale of, say, 1 to 7, the potential improvement in the decisions of each of these four groups that could result from the availability of a frequent, granular precise value of any given statistic. This would be a first step towards assuring that national statistics are “of the people, by the people and for the people.”