What responsibility do universities have in preparing public policy students for a world increasing informed by data and technology? David Eaves, a lecturer in Public Policy at the Harvard Kennedy School and Ben McGuire, a Kennedy School student delve into this issue.
Over the last several years, a fascinating trend has emerged among students applying and entering to Harvard’s Kennedy School of Government. A small but fast-growing number are taking classes in data science and machine learning. They’re asking professors for more direct coding experience – and increasingly are critical of the skills imparted by required courses in statistics and data. They’re filling up optional tutorials offered by more tech-savvy parts of the University on GIS, data visualization, and web scraping. There is a broad, large interest in learning more about technology – especially the interaction between technology and governance. This change is being driven by student demand, not supply – and there’s a risk that supply is moving too slowly.
In thinking about their future roles in government, policy students have recognized the need to learn three key skills that policy schools have been slow to offer. First, open data and the internet have created huge sources of messy data; students see that unlocking this data could have a big impact on the public good. Second, students who plan to work in positions where they build, buy, and manage technologies know that they need competence and confidence in understanding the relationship between technology and policy. Third, running and implementing advanced analytics in policy demands some fluency with more (and more powerful) programs than the traditional policy toolkit.
The internet and open data have created a massive opportunity for students with technology skills to solve public problems. Thirty years ago, policy programs viewed teaching students the granularities of data extraction and management as a waste of time. Our goal was helping students turn data into insights into recommendations into policies – and datasets at the time were largely limited in number and scope, expensive to produce, relatively high quality, and created for statistical analysis by agencies and economists. Today, nearly the opposite is true, and policymakers are awash in rich, messy datasets coming from myriad public and private sources. Learning how to write a script that can pull down, clean, and restructure data from weather stations, or run a language-processing algorithm on complaints to a financial regulator isn’t just an academic exercise. It’s a core skill set which can give policy students the edge, and give future managers the capacity to achieve their teams’ full potential.
Additionally, having a sophisticated fluency in technological tools isn’t merely another skill that students can tack on to their resumes. Rather, it’s the key that unlocks a wealth of management and leadership tools. When our graduates enter the public or non-profit sector, there’s a near certainty that they will need to make a decision about whether to build or buy a technology tool for their organization. Policymakers with no grasp of technology fundamentals will have a hard time evaluating a complex set of choices – many will simply be at the mercy of vendors who have every reason to oversell options and underrate complexity. Learning Python or R won’t make students tech procurement experts, but it is one way to provide them with the mindset to ask probing, skeptical questions. Even more importantly, students who end up managing analysts or others whose jobs include coding or data analysis will be exponentially better in their management roles if they comprehend the underlying problem and solution sets. They will be able to understand what is realistically achievable, make educated guesses about how long it will take, and manage the entire team more effectively because they are able to translate a deceptively simple question about data into an actionable answer.
Finally, the advanced analytics quickly becoming the norm in public policy conversations demand that students have fluency and depth with more complex programming languages. It’s no longer enough that public policy students are forced into a semester of SPSS or Stata – nor is it enough to be slightly more progressive and allow R or Python for regression analyses. Running and effectively explaining a permutation test with synthetic controls requires deep understanding of complex mathematics and precise, individualized code. And with machine learning algorithms, myriad blockchain applications, and early-stage artificial intelligence entering the mainstream, policy students who aren’t prepared to read and understand – if not write – the code that guides complex systems could be left out in the cold.
So why has the policy education world been slow to embrace technology? There are plenty of excuses – already-crowded core curricula, fast-changing and expensive technologies – but few satisfying answers. The real answer may be that policy programs tend to replicate one of our biggest criticisms of the public sector: we like things that work pretty well, and we’re very scared to innovate without overwhelming evidence. The problem is, policymaking is pivoting to a complex, data-rich world faster than we are helping students meet it. This stuff has been mature for ten to twenty years – it’s time for us to become fast followers of what works. If we aren’t able to move at least as fast as the world of policy, we could eventually be doing many of our students a disservice.
We don’t need to overhaul every aspect of our curriculum, nor should we throw out the fundamentals of our mission for students and for the world. Coding shouldn’t be the menu – but it needs to be on the menu. We can and we must be more willing to recognize the changing nature of the policy world, and its rapidly-growing orientation toward data and technology. We can and we must identify the fundamental coding and technology tools that can inform students ability to leverage new data for policy, manage technology in the public-sector workplace, and work with advanced analytics methods. The students are ready, they’re looking for our help, and they’re pushing to learn more. The only question that remains is – how long will it take the faculty and staff of public policy programs to catch up?