Improving Voter Participation through Mobile Phones and Newspapers in Mozambique

Paul Collier
Pedro C. Vicente
Ana Vaz
Cabo Delgado, Zambezia, Gaza, and Maputo-Province, Mozambique
1,154 households in 161 polling locations
2009 - 2009
Target group:
Outcome of interest:
Electoral participation
Intervention type:

Many countries have struggled to build democratic institutions that respond to the will of the general population. Researchers followed the introduction in Mozambique of an information campaign using SMS, a hotline for electoral misconduct, and a free newspaper to evaluate whether improved information and participation could affect voter turnout and other elections-related outcomes. All three programs had clear positive effects on turnout and some benefitted the incumbents’ vote share. However, the effects of the programs on political participation were weaker for more socially-connected individuals; these individuals appear to downplay the value of their own political participation as they realize that more of their peers will be voting because of the programs. 

Policy issue

Voter participation is an important way for citizens to guide the political process. In theory, governance improves when democratic institutions reflect the will of the majority. However, political participation is not guaranteed. Voters may not vote for many reasons, such as a lack of information, barriers to notifying authorities of campaign irregularities, an inability to share information with other citizens, or a belief that their vote won’t make a difference given the voting patterns of others. These challenges to participation may also limit voters’ willingness or ability to demand political accountability. Even with high participation rates, elections may not discipline politicians if incumbents are able to reduce political participation or competition before the election. Once in office, politicians may develop licit or illicit mechanisms to secure future victories well before the votes come in. Providing voters with accessible electoral information and encouragement to actively engage in the political process may increase participation and improve the electoral process overall.

Context of the evaluation

Mozambique suffered a devastating civil war from 1977 to 1992, which ended with an agreement to hold multi-party elections. Since then, the Frente de Libertação de Mozambique (FRELIMO) has won all national elections, earning an increased share of the vote in each one. Meanwhile, voter turnout has fallen rapidly, from 88 percent in 1994 to just 36 percent in 2004 (a 59 percent decline). All national elections have suffered from allegations of electoral irregularities, primarily ballot fraud, carried out by FRELIMO. In addition, numerous sources point to low levels of information and political accountability in Mozambique. Afrobarometer data find relatively low levels of support for democracy and only one quarter of Mozambicans believe that their elected representatives are concerned with looking after their interests or listening to their opinions.1 Citizens demonstrate resistance to proffer opinions about politics and do not seem to see democracy as important for improving economic outcomes.

This study took place during the presidential, primary, and provincial assembly elections of October 2009. The run-up to the elections was relatively calm, with FRELIMO and its candidate expected to win. The elections themselves were conducted in line with international standards, according to both international and national observers. FRELIMO won decisively, earning 75 percent of the vote in both presidential and parliamentary elections.

Red car with campaign posters supporting FRELIMO party
Car with FRELIMO campaign posters

Details of the intervention

Researchers conducted a randomized evaluation to test the impact of three voter education interventions on electoral participation in Mozambique’s 2009 elections. From a total sample of 161 polling areas, researchers randomly assigned 120 polling areas to one of three interventions, with the remaining 41 polling areas serving as a comparison group. Within each treated polling area, field staff randomly selected an average of eleven individuals to participate in the study, nine of whom received the intervention assigned to that polling area. Therefore, each treated polling area included a group of untargeted individuals who did not directly receive any of the interventions, but who may have indirectly received access to the programs through social interactions.

  • Civic Education: Participants received leaflets designed by the electoral commission that provided a detailed explanation of the steps needed to vote on election day. In addition, in the two weeks prior to the election, participants received five text messages per day focusing on the importance of voter participation.
  • Electoral Hotline: Participants were invited to send text messages reporting electoral problems to an electoral hotline. Each report was shared with all other program participants in the area via text message.
  • Civic Education plus Electoral Hotline (Newspaper): The newspaper program combined elements of other two interventions. Participants received weekly copies of a free newspaper @Verdade that featured the contents of the civic education leaflet and text messages. @Verdade also sponsored a national hotline to report electoral problems.

In addition to these three programs, after the interventions and the elections, all survey respondents received a leaflet inviting them to send SMS messages suggesting policy priorities to the winner of the election. Because each message had a small monetary cost, researchers interpreted the act of sending a message as a measure of demand for political accountability.

Within each area, researchers conducted surveys on electoral behavior and perceptions before and after the elections and matched this survey data to administrative data on turnout in each polling area. In addition, researchers analyzed information on electoral problems (campaign misconduct, election-day misconduct, or violence and intimidation) collected by the national hotline of the newspaper @Verdade and election observers.

Finally, researchers gathered information about targeted and untargeted individuals’ social relationships and connectedness. This allowed researchers to determine whether individuals responded differently to the interventions if they were more socially and geographically connected to other participants; in other words, the “peer effects” of the programs.

Results and policy lessons

All three programs were effective in increasing voter turnout and the level of trust in the electoral process, although each worked in a slightly different way. Only the newspaper intervention resulted in increased voter demand for accountability and reduced the prevalence of electoral problems. However, the effects of the interventions on participation were weaker for more socially-connected individuals; these individuals may have realized that more of their peers will be voting due to the different programs, thereby making their own political participation less meaningful. Altogether, these results suggest that increasing citizens’ information and participation opportunities can affect their behavior as voters–though these effects may be different based on how socially connected an individual is-and that the tool used to deliver this information is important. Furthermore, presenting information about the elections in a neutral or positive light can boost support for the incumbent when the distinction between state and party is not clear. Based on these findings, researchers suggest that incumbents may have learned ways to bend the electoral system in their favor beginning well before election day. Thus, while there is value in making elections more transparent, those efforts alone may not be sufficient to encourage genuine electoral competition.

Voter Turnout: According to official polling data, all three interventions increased turnout by close to five percentage points (an 11 percent increase relative to the comparison group) on average in treated areas. These effects were not statistically significantly different across the different interventions.

Voting Patterns: On average, the civic education program had a positive effect on the incumbent’s vote share and a negative effect on the challengers’ vote share. The newspaper intervention had the same effects to a lesser degree, and the hotline intervention did not have any effect on vote shares. One possible explanation for these effects may be that the civic education program and the newspaper, to a lesser extent, focused on neutral or positive messages about the election. Because some of the information was sponsored by the electoral commission, these positive messages may have been associated with the incumbent. In comparison, the hotline included negative messages reporting electoral problems.

Demand for Political Accountability: Only the newspaper intervention significantly increased demand for political accountability, as measured by an individual’s likelihood to send a message to the president-elect. On average, respondents who received the newspaper were 9.9 percentage points (65.6 percent) more likely to send a message than those in the comparison group.

Electoral Problems: The newspaper program reduced both the incidence and intensity of electoral problems. Polling areas where @Verdade was distributed had 0.58 fewer electoral problems than comparison areas, relative to a baseline of 0.95 problems (a 61 percent decrease). These findings suggest that the newspaper was particularly effective at improving local politician behavior, perhaps due to increased perceived monitoring.

Peer Effects: Researchers found that the interventions did not affect targeted and untargeted individuals differently; on average, neither type of individual was significantly more likely than the other to turn out to vote after receiving any of the interventions, and both had similar increases in demands for political accountability,. However, the programs did impact individuals differently based on their level of social connectivity. Socially-connected individuals in treatment areas were less likely to participate in politics than those with fewer social connections. A socially-connected individual may realize that more of his/her peers will be voting due to the different interventions, thereby making her own political participation less meaningful.

More specifically, despite the positive average effects of the interventions on voter turnout in treatment areas, researchers found that their effects were smaller for more socially-connected individuals. For individuals directly targeted by any intervention, the impact on voter turnout was between 2.7 and 4.9 percentage points smaller (40 and 68 percent smaller) for those with an average level of social connectedness, relative to an individual with no social connections. A similar pattern occurred among untargeted individuals: effects on turnout were between 3.7 and 8.0 percentage points (36 and 79 percent) smaller for an untargeted individual with an average level of connectedness compared to an individual with no social connection. More socially-connected individuals were also less likely to send a message to the president-elect than their unconnected peers in treatment areas, though these differences are only occasionally statistically significant for the civic education treatment (depending on the measure of social connectivity). Conversely, however, individuals’ interest in elections increased with higher levels of social connectedness within treatment areas, particularly so for areas who received the free @Verdade newspaper.

For researchers, this relationship between social connectedness and weak intervention effects is likely due to how more socially-connected individuals (targeted or not) are in a better position to realize that their peers are more likely to vote, or that the incumbent will extend their lead, as a result of the intervention. Consequently, more-connected individuals conclude that their vote will make less of a difference in influencing the elections, making them less likely to vote than less-connected individuals in areas that have received an intervention. The same logic may explain central individuals’ relatively lower demands for political accountability in treatment areas, as they see more of their peers already voicing their political demands and therefore feel that their participation is less essential.

In the follow-up of this study, newspaper @Verdade implemented hotlines for a number of national events, using the same platform that was introduced for the research (ushahidi). This included Maputo riots in 2010, and elections in both 2013 and 2014. The websites for these events were among the most prominent sources of news about Mozambique during those times (both nationally and internationally). @Verdade continued its focus on civic education with particular efforts during electoral periods. The use of text messages and internet platforms became central to the newspaper’s activity. The electoral commission in Mozambique, in both its political (CNE) and technical (STAE) branches, is aware of the results of this research and has taken it into account in the design of its civic education activities.

Aker, Jenny, Paul Collier, and Pedro C. Vicente. 2017. “Is Information Power? Using Cell Phones and Free Newspapers during an Election in Mozambique.” Review of Economics and Statistics 99(2): 185-200. doi:10.1162/REST_a_00611

Fafchamps, Marcel, Ana Vaz, and Pedro C. Vicente. 2017. “Voting and Peer Effects: Experimental Evidence from Mozambique.” Economic Development and Cultural Change, forthcoming. Doi:10.1086/700634

Pereira, Joao, Ines Raimundo, Annie Chikwanha, Alda Saute, and Robert Mattes. “EightYears of Multiparty Democracy in Mozambique: The Public's View.” Afrobarometer WorkingPaper 30, 2003.