Making Smarter Policies, Faster

Policies matter. Just look at the The Commonwealth Fund 's latest state scorecards on women’s health . (Spoiler alert: things are getting worse) If we are to make our communities healthier, and narrow disparities, we need to know how to learn from our policy successes and mistakes faster, and keep improving over time.

How can we learn to make smarter policies faster?

That’s the main question addressed by a breakthrough article published recently in Science magazine, “Climate policies that achieved major emissions reductions: Global evidence from two decades.” It is a remarkable piece of work with important implications across all policy-making, including in both planetary and population health.

Annika Stechemesser and colleagues from the PIK - Potsdam Institute for Climate Impact Research looked at 1500 climate policies implemented between 1998 and 2022 across 41 countries on 6 continents. The policies included command-and-control measures like emissions standards and technology mandates and also market-based policies like subsidies. (The authors noted that in their study, very few carbon pricing approaches like carbon taxes had been deployed.)

Instead of the usual approach of trying to isolate the effect of individual policies in a particular country, these authors used a clever machine learning method. They focused on modeling national emissions, especially large reductions, using machine learning-enhanced difference-in-difference methods, as a means of understanding which policy interventions mattered.

What did the team find??

The researchers identified 63 policies that were associated with lower emissions, and they shared certain features. For one, pricing mattered most. Taxation and reduced subsidies (for fossil fuels) led to some of the biggest successes, especially when combined with policies like energy efficiency mandates. Results varied by sector and country, but overall, this model was effective at predicting significant drops in carbon emissions in countries like the US, UK, Norway, Spain, Hungary, Germany, and China.

Now think about the implications of this work on health policy.

In health care, we have no shortage of policy instruments. Across communities, states, nationally, and globally, we could using similar machine learning models to ask, which policies directed at which players, alone or in combination, were most effective? The machine learning model also welcomes inclusion of other social policies that impact health, like food, housing, education, transportation, and environmental protection.?

Globally, 2024 is the year of elections and with new leaders come fresh opportunities to revisit policies in health and climate. It couldn’t be a better time for health services researchers to apply machine learning-enhanced capabilities to guide policy making. It's time we get smarter, faster.