BESIDES brains, the only tools an economist used to need were a pen and notebook. But massive improvements in computing power have turned the dismal science into an increasingly empirical one. Research by Daniel Hamermesh at Royal Holloway, University of London, finds the share of economics papers in leading journals focused on pure theory fell from 58% in 1983 to 19% in 2011.

Three types of empirical papers have taken their place. The first sort feeds on publicly available data, such as household surveys. The second relies on data from experiments, such as randomised controlled trials. Most leading empirical papers, however, now rely on other data, often administrative and acquired through extensive negotiation with government officials. Analysis by The Economist of work from the National Bureau of Economic Research finds that at least 28 papers it released last year featured the use of administrative data. Before 2000 hardly any did (see chart).

Collaboration between professors and governments can yield impressive (if often predictable) results. In March Raj Chetty, an economist at Stanford University, published a paper which linked tax records with survey data and found that white men in America earned significantly more than black men, even after controlling for their parents’ incomes.

Sharing tax records with researchers can also yield useful policy advice. Arun Advani of Warwick University analysed self-assessment tax records provided by the British government, and showed that tax audits made money for the government by scaring tax cheats into improving tax compliance in future.

But not everyone is happy about relying on government sources. In America economists grumble that prominent researchers are more likely to have the connections needed to obtain such data. In other words, this line of research might widen inequality within the profession, because academic superstars have more opportunities to do celebrated research. So academic economics, like the society it studies, would become a game of who, rather than what, you know.

Moreover, governments, wary of privacy concerns, themselves tend to be cautious about sharing citizens’ data. Britain’s is still haunted by the memory of the revelation in 2007 that 25m child-benefit records had been lost in the post. So when officials share data with academics, they do so selectively and with plenty of bureaucratic hurdles, often making any resulting research impossible for third parties to replicate.