Since 1994, voter turnouts in Mozambique have dropped. In 1994, 80% of the voter population voted on election-day (ballots cast/voters registered). In 2004, this figure had dropped to 33%, increasing only to 45% and 49% in 2009 and 2014 respectively.

While low voter turnouts are not uncommon, they are problematic. When people do not not vote, it undermines the legitimacy of the winner and can lead to suboptimal allocation of resources, as the preferences of the entire population are not fed into the political system.

Since the early 2000s, the Afrobarometer has included questions about voter behaviour. By pairing the answers with other information about the respondents, we can explore the correlates of voter participation in Mozambique over the last three electoral cycles (2004, 2009, and 2014).

It is important to note that self-reported voter participation is consistently higher than actual voter turnout. This is the case also in Mozambique. In 2015, 80% of the Afrobarometer respondents said they voted in the last national election (excluding those that were too young to vote). Yet, we know that only 49% actually did. In other words, around one third of the respondents did not tell the truth. While this is a problem in terms of estimating overall voter turnouts, it is not a problem when it comes to estimating correlates and marginal effects, assuming that the respondents who did not tell the truth are randomly distributed.

Election year Self-reported voter participation Voter turnout as recorded by CNE/STAE Discrepancy
1994 NA 87.9 NA
1999 NA 70.3 NA
2004 82.8 32.5 50.3
2009 83.4 44.6 38.8
2014 80.2 48.5 31.7

We derive three main findings from our analysis. The first is that older people are more likely to vote than younger people. A 60-year-old is around 5% more likely to vote than a 40-year-old; and more than 15% more likely to vote than a 20-year-old, all else being equal. We see this pattern also in other countries but it represents a greater problem in Mozambique, where younger people make up a growing share of the population. The figure below plots the results. The black line indicates the predicted probabilities above and below the mean. The grey area reflects the 95% confidence intervals.

The second finding is that people who are satisfied with the president are more likely to vote than people who are not. A person who ‘strongly approves’ of his performance is 13% more likely to vote than a person who ‘strongly disapproves’ of his performance, again, all else being equal. What this suggests is that people who are dissatisfied do not necessarily respond by voting for an opposition candidate. Many respond simply by staying at home on election-day.

The final finding is that people who trust the national electoral commission (CNE) are more likely to vote than people who do not. Again, the difference is around 10-15%, all else being equal. This is the case also with people who trust the police and the courts. In other words, people who generally trust public institutions are more likely to particpate in elections.

The present analysis is preliminary. That being said, it does suggest that a considerable share of the population–younger people and people who are dissatisfied with the government and generally do not trust public institutions–consciously choose to stay at home on election-day. This interpretation of the results is supported by the fact that 58% of the respondents who admitted to not have voted in the last national election (excluding those that were too young to vote) said they “decided not to vote” or “did not have time”.

For further information, including acces to data and regression results, please contact Halfdan Lynge-Mangueira at