This article was published in Harvard Magazine. It highlights the work of SAI Steering Committee Member Asim Khwaja.
By Marina N. Bolotnikova
AROUND THE WORLD, tax receipts in low- and middle-income countries are much lower than they ought to be. Poor recordkeeping makes it easy for people to pay less than they owe; distrust that taxes will be returned as government services undermines people’s willingness to pay. Absent a strong culture of civic participation, policymakers need to find ways to improve tax compliance without further degrading public faith in institutions.
Asim Khwaja, Sumitomo-FASID professor of international finance and development, who directs the Evidence for Policy Design program at the Harvard Kennedy School, has spent the last few years working with the government of Punjab (the most populous province in Pakistan, home to Lahore and more than 100 million people) to study that problem. In Pakistan, tax collectors’ salaries are largely predetermined, based on experience and education. With no financial incentive to bring in more revenue, collectors frequently collude with taxpayers, accepting large bribes in exchange for tax write-offs. Both taxpayer and tax collector thus benefit—at the expense of the state. Khwaja and his colleagues, London School of Economics professor Adnan Khan and MIT professor Benjamin Olken, thought that improving the tax collectors’ performance would be straightforward: pay them based on how much revenue they collected. And indeed this is what they found: employees incentivized in this way collected much more than employees paid through the existing system.
Khwaja tested three different versions of this incentive system. One simply paid the collectors large bonuses based on how much money they raised; one paid based on amount of money raised, taxpayer satisfaction (did citizens become angry when forced to pay higher taxes?), and accuracy of assessed property values; the third, and most subjective, awarded bonuses based on all these factors, plus a rating by the employee’s manager. On average, the incentive system paid employees 30 percent of any tax revenue collected above a historical benchmark. The plan based on revenue alone did the most to increase tax receipts, which after two years grew 46 percent under that system, compared to 28 percent in a control group. The other two versions raised receipts by 41 percent and 36 percent, respectively. These more complex programs can confuse employees about how best to improve their performance, particularly if some criteria appear directly contradictory—such as collecting more tax money while increasing taxpayer satisfaction.
But a compensation system based on total revenue alone poses potentially enormous moral and political risks. Khwaja warns, “There’s a certain laziness, intellectually, in thinking that there’s such a thing as a straightforward best practice.” Because a simple revenue-based incentive system increases the reward for collecting more money, it also increases the bargaining power of tax collectors relative to taxpayers, allowing them to demand higher bribes. The researchers predicted that this would have a significant effect on taxpayers’ attitudes toward the government—and yet surveys of taxpayers didn’t show one. “The scheme should have had huge political costs,” Khwaja explains. “But the tradeoff between raising money and limiting political repercussions didn’t seem to be that severe.”
To understand why, he analyzed how tax payments changed under the new system. Total revenue shot up, but most of that money came from a small group of relatively well-off taxpayers. Everyone else, to avoid paying taxes in full, had to pay higher bribes: after the new program took effect, property owners reported that the “going rate” for bribe payments had risen. “We forgot that the tax collector is living in the community that he’s taxing,” Khwaja says. “He had already internalized the political cost.” To increase revenue while minimizing backlash from their fellow citizens, for example, tax collectors may have targeted for full payment those property owners with relatively little political power, though this is difficult to confirm through the experimental data.
Of course, an increase in bribe payments itself poses a risk to the legitimacy of the tax system. Khwaja suggests that policymakers ought to be cautious about implementing a crude pay-for-performance system that lacks a mechanism to discourage bribery. And increasing tax compliance alone can’t create a civil society with faith in its government. This has moved Khwaja to ask a much more ambitious question: his next project aims to link taxes to public benefits that directly meet the expressed needs of people in the community, to test whether this can improve government’s legitimacy. He jokes, “It’s only the entire basis of the social compact of the state.”