Rivalry's Poor Returns

New research suggests that, in some settings, ethnic loyalties trump productivity.
June 24, 2014 | Research Feature
Print this page

In most countries, ethnic diversity goes hand in hand with limited economic growth: diverse nations are, on average, poorer than their less diverse neighbors. Virtually all the evidence for this comes from studies of the public sector that look at local and national decision making, and they show that diverse groups have a hard time governing, from agreeing on policies to deciding how much to spend on public goods such as schools and roads.

Economists understand far less about the effect of diverse groups on productivity in the private sector, a key element of nations’ overall economic status. Professor Jonas Hjort wanted to know: is there a different, more direct effect of ethnic diversity on individuals’ income? Are diverse teams more or less productive than homogeneous teams? His new research, which studies production and human resources data from a large Kenyan flower farm, shows how productivity in the private sector can vary depending on relationships between ethnic groups.

On the farm, workers rotate through different teams. For each three-person team in a packing hall, one upstream worker distributes flowers downstream to a two-person team, who arrange flowers into bunches. Hjort has daily production data from the farm for over two years, including team members and which of two ethnic groups they belonged to. Hjort could see when each worker was on a mixed team or a homogeneous team; by comparing the average productivity of the homogeneous and mixed teams at different points in time he was able to document how ethnic divisions affected productivity.

Horizontally mixed teams — in which the supplier was of a different ethnicity than one of the assemblers — were 5 percent less productive than homogenous teams, while vertically mixed groups — teams in which the supplier was a different ethnicity than both assemblers — were 8 percent less productive than homogeneous teams. (So, teams in which all members belonged to the same ethnic group were the most productive.)

In short, the suppliers were undersupplying members of the rival ethnic group on the production teams by distributing more flowers to teams made up of their same tribe. This lowered their own productivity as well as pay, but the upstream workers were willing to accept lower pay themselves in order to lower the pay of team members of rival ethnic groups.

In late 2007, ethnic tensions flared up in Kenya in response to controversial national election results that pitted candidates from the two ethnic groups against each other. On the farm, the output gap between homogeneous and diverse teams almost doubled during the initial post-election conflict. Hjort attributes this to an increase in discrimination fueled by the political controversy, which split along ethnic lines. “This demonstrates that the economic costs of ethnic diversity vary with the political environment, and conflict periods have real costs,” he says.

When the farm owner shifted from individual to team pay for assemblers six weeks into the conflict period, productivity increased for horizontally mixed teams, so much so that while production fell modestly in homogeneous teams and teams where the supplier was of a different ethnicity than both assemblers, overall productivity still edged up. “It’s not clear if the plant owners were aware of the effect of ethnicity on teams before the election conflict,” Hjort says. “But they clearly reacted to decreasing productivity during the conflict by changing the incentives of upstream workers: when the firm started basing pay on team output, there was no longer any reason for the supplier to misallocate flowers in mixed assembly teams, since both would get paid the same.”  But although the firm responded to the productivity problem spurred by the election crisis, it could not eliminate it, and Hjort found evidence that discrimination was affecting productivity long after the initial election conflict died down.

Diverse groups need not doom an industry or a nation  — the flower industry in Kenya has been successful for several decades in spite of ethnic strife in its ranks — but this research shows how they can dampen productivity. Interacting economically with individuals of other ethnic backgrounds is hard to avoid when urbanization and economic modernization bring larger groups of workers together. Firms would do well to anticipate ways in which they can help diverse workforces reduce or eliminate conflict.

Hjort emphasizes that these results are not definitive: ethnic diversity  may not always inhibit productivity. Other research has shown that diversity is often quite good for productivity: Professor Katherine Phillips has shown that diverse teams are better at solving problems than homogeneous teams, at least in the United States.

Hjort speculates that one reason for the findings in the current study is that he looks at a simple production process where there’s little place for the potentially positive effects of diversity to surface. “If you’re trying to solve a complicated problem, it may be better to have people of different backgrounds,” he says. “As a starting point, we should expect the effects of ethnic diversity to depend on many circumstances.”

Jonas Hjort is assistant professor of finance and economics at Columbia Business School.

This paper won the 2013 CESifo Prize for Best Paper in Applied Microeconomics.

Jonas Hjort

Professor Hjort joined Columbia Business School in 2012. Before coming to Columbia, Professor Hjort received a BSc Economics from the London School of Economics, an MA International & Development Economics from Yale and a Ph.D. Economics from UC Berkeley. His research spans issues in development, labor and behavioral economics, with a particular focus on topics in their intersection. In recent work Professor Hjort has analyzed, for...

View full profilePersonal Website

Read the Research

Jonas Hjort

"Ethnic Divisions and Production in Firms"


Download PDF View abstract/citation