VoxEU Column Development

Measuring identity in the tropics: Evidence from the implicit association test

Ethnic identity can fracture nations – it is a leading explanation of the Boko Haram insurgency in Nigeria. This column presents evidence from a psychological test that reveals how members of different ethnic groups from the Democratic Republic of the Congo form their respective ethnic identities. This Implicit Association Test reveals that people have a small, statistically significant bias towards own-group. The magnitude suggests that biases are consciously held, rather than being a conditioned response.

For the past five years an armed group called Boko Haram has been terrorising northeastern Nigeria. As we write they are besieging Maiduguri, the capital of Bornu, one of Nigeria’s 36 states.

What motivates them? There are many theories, but one with wide currency is that the rebellion is fueled by ‘ethnic tensions’ between northern groups – particularly the Kanuri – and the national government.1 According to this view, the Kanuri – having a different language, culture, and history – do not identify as being Nigerian, and therefore do not recognise the authority of the central state. Instead they demand political independence and their own state – something they are willing to fight and kill for.

This is not the only case in which ethnicity has been blamed for adverse outcomes. A large empirical literature has emerged linking ethnic differences to a wide range of outcomes, including conflict (Fearon and Laitin 2003, Montalvo and Reynal-Querol 2005), the under-provision of public goods (Gugerty and Miguel 2005), voting patterns (Casey, forthcoming), labour productivity (Hjort 2015), and economic underdevelopment generally (Easterly and Levine 1997).

It is clear that identity – what Akerlof and Kranton (2000) call a ‘sense of self’ – is an important determinant of human behavior. Our identities influence the way we behave, what social norms we adopt, who we associate and cooperate with, and how we treat others. Our attitudes towards those we perceive to have a different identity can lead to discrimination and, in the limit, violence. This naturally leads to important questions. To what extent do individuals in Africa actually define themselves in terms of their ethnicity? And if they do, does this lead to the sort of own-ethnic bias and behaviors thought to underpin many of the existing empirical findings? If such a bias exists, at what level of consciousness does it reside? Is it part of what Kahneman (2011) calls our ‘System 1’ – our fast, subconscious, intuitive self – or our ‘System 2’ – our explicitly rational calculating self?

To date, there is limited research directly testing for ethnic bias and favouritism. An exception is Habyarimana et al. (2007, 2009), who infer it from the decisions made in behavioral games played in Uganda. In Lowes, Nunn, Robinson and Weigel (2015a), we argue that recent work in psychology allows us to develop a direct measure of ethnic bias – namely, one’s unspoken and perhaps subconscious preferences towards different ethnic groups, including one’s own.

Our measure is derived using from a variant of the Implicit Association Test (IAT), which is a computer-based sorting task developed to measure an individual’s implicit association between pairs of objects (Banaji and Greenwald 2013). We use a variant of the standard IAT, called the single-target IAT (ST-IAT) (Bluemke and Friese 2008), to measure participants’ implicit attitudes towards four major ethnic groups in the region: Luluwa, Luba, Lele, and Kuba.

Our ethnicity IAT was implemented among 536 participants living in Kananga, a city in the Democratic Republic of Congo, using ten-inch touch-screen tablets connected to headphones. The IAT is broken into four parts, one for each ethnic group. Each part comprises two blocks. Within each block, participants hear 24 randomly ordered words that are of three types: good words (e.g., nice, happy), bad words (e.g., terrible, wicked), and words associated with one of the four ethnic groups (e.g., Luba tribe, Luba culture).2

Participants are asked to sort the words they hear to either the left (by pushing a button on the bottom left of the screen) or the right (by pushing a button on the bottom right of the screen). The screen as seen by participants is shown in Figure 1. In all blocks of the IAT, good words are always sorted to the left, and bad words are always sorted to the right. Participants are reminded about where to sort the words by the image of a happy person on the top left of the screen and the image of a sad person on the top right of the screen.

Within each part of the IAT – recalling that there are four parts, one for each ethnic group – one of the two blocks instructs participants to sort ethnicity-related words to the same side as good words (left); the other block instructs them to sort ethnicity words to the same side as bad words (right). A small written label appears in the screen’s top left or right to remind the participant of the sorting pattern for that block. The screenshot shown in Figure 1 is from the block in which words related to the Luba ethnic group are sorted to the same side as good words (left).

To obtain a measure of a participant’s view of an ethnic group, we compare the average response time when the words about the ethnic group in question are sorted (1) to the same side as good words, and (2) to the same side as bad words. The logic behind the test is that if an individual has an underlying positive implicit association of an ethnic group, then the sorting task will be easier and – and response times will be quicker – when the ethnic group’s words are sorted to the same side as good words.

Figure 1. ST-IAT screen shot. In this block, Luba ethnic words are sorted to the same side as good words.

The findings

We use these measures to test whether individuals have an implicit bias towards their own ethnicity. Our evidence suggests that they do. Participants in our sample tend to have a more positive implicit view of their own ethnic group than they have of other ethnic groups. What is surprising is that although the effect is precisely estimated, it is remarkably small. In fact, this implicit own-ethnicity bias is considerably smaller than the explicit own-ethnicity bias we found by asking participants survey questions about their attitudes towards the same four ethnic groups. In short, implicit bias is significantly weaker than explicit bias. Ethnic discrimination appears to be more a product of System 2 thinking than System 1 thinking.3 This suggests that people tend not to have strong innate preferences for coethnics over non-coethnics, though they may consciously choose to act as if they do.


Adida, C L, K E Ferree, D N Posner, and A L Robinson (2014), “Who’s Asking? Interviewer Coethnicity Effects in African Survey Data”, Mimeo.

Akerlof, G and R Kranton (2010), “Economics and Identity”, Quarterly Journal of Economics, 115 (3): 715-753.

Banaji, M R and A G Greenwald (2013), Blindspot: Hidden Biases of Good People, New York: Delacorte Press.

Bluemke, M and M Friese (2008), “Reliability and Validity of the Single-Target IAT (ST-IAT): Assessing Automatic Affect Toward Multiple Attitude Objects”, European Journal of Social Psychology, 38: 977-977.

Casey, K E (2014), “Crossing Party Lines: The Effects of Information on Redistributive Politics,” forthcoming in the American Economic Review.

Easterly, W and R Levine (1997), “Africa’s Growth Tragedy: Policies and Ethnic Divisions”, Quarterly Journal of Economics, CXII(4), 1203-1250.

Fearon, J D and D D Laitin (2003), “Ethnicity, Insurgeny and Civil War”, American Political Science Review, 97(1), 75-90.

Gugerty, M K and E Miguel (2005), “Ethnic Divisions, Social Sanctions, and Public Goods in Kenya”, Journal of Public Economics, 2005, 89(11-12), 2325-2368.

Habyarimana, J, M Humphreys, D N Posner, and J M Weinstein (2007), “Why Does Ethnic Diversity Undermine Public Goods Provision?”, American Political Science Review, 101(4): 709-725.

Habyarimana J, M Humphreys, D N Posner, and J M Weinstein (2009), Coethnicity: Diversity and the Dilemmas of Collective Action, New York: Russell Sage.

Hjort, J (2015), “Ethnic Divisions and Production in Firms”, Quarterly Journal of Economics, forthcoming.

Kahneman, D (2011), Thinking, Fast and Slow, New York: Farrar, Straus and Giroux.

Lowes, Sara, Nathan Nunn, James A. Robinson, and Jonathan Weigel. 2015a. “Understanding Ethnic Identity in Africa: Evidence from the Implicit Association Test (IAT),” NBER Working Paper #20885.

Montalvo, Jose G. and Marta Reynal-Querol. 2005. “Ethnic Polarization, Potential Conflict and Civil War,” American Economic Review, 95(3): 796-816.


1 See for example the remarks of Samuel Uche, Archbishop of the Methodist Church of Nigeria on January 15, 2015: http://www.premiumtimesng.com/news/headlines/174730-fulani-kanuri-behind...

2 While IATs are usually administered with written words that appear on the screen, due to low literacy levels, we used recordings of words instead.

3 This finding is consistent with earlier work by Habyarimana et al. (2009), who, using behavioral games, found that individuals’ implicit bias towards coethnics was weak (or actually non-existent), but that their (observed) explicit behavior showed a coethnic bias.

1,050 Reads