Detecting electoral fraud in Japan

Picture: Flickr

Volume 2, no. 5, October 2014

Author: Yusaku Horiuchi, Darthmouth College

Moving house ‘on paper’?

It may seem fairly obvious, but only those people who fulfil particular requirements are given voting rights in an election. In Japan, one of these is that voters must be Japanese citizens aged 20 or over. Another is that voters have a registered address in a municipality within a relevant electoral district for more than three months. (In municipal elections, each municipality constitutes an at-large electoral district.) According to the Public Offices Elections Law, exploiting this requirement by moving one’s residential registration to another municipality — just on paper — for the purpose of voting, while continuing to reside in an original municipality, is illegal.

Does this kind of electoral fraud actually exist? Searching newspapers for the keywords, ‘false moving in’ (kakū tennyū) and ‘false registration’ (sagi tōroku), shows that the problem is often reported before or after municipal elections and takes place irrespective of a candidate’s party, the region or time period. It is clear that this is a prevalent and deeply rooted problem in the Japanese electoral process.

So how is this sort of electoral fraud actually committed? Cases where a candidate or supporters ask relatives and friends living in other municipalities to move their registration to the candidate’s own address or that of their electoral offices are the most common. There are other obviously shady cases, such as one instance where 202 people were registered as ‘living’ at an address of only 240 square metres, or another instance where a large number of staff members ‘lived’ in a teppanyaki restaurant. There is a case where the massive influx of people into a municipality some months before election increased the municipality’s population by more than 10 per cent.

It is possible that voters who have changed their ‘address’ cast an absentee vote or vote prior to the election. It is, however, more likely that someone living in a municipality voted by impersonating those who ‘moved’ their residential registration, and there have been reports of such instances.

Abuse of weak residential registration requirements

Why is this sort of electoral fraud possible? First, it is because not all municipalities hold their elections on the same day. Therefore, it is easy enough for someone to move their address from a municipality that doesn’t hold an election to another municipality that does. More importantly, however, it is because the paperwork required to change one’s address is extremely simple in Japan. Although a required form must be submitted within 14 days of moving, no proof of residence is required. It is even possible to ask someone else to submit the form on someone’s behalf. In a reported case, an agent abused this procedure and submitted the change-of-address form for up to 128 people in a day!

Importantly, three months after submitting the paperwork, individuals are automatically given voting rights in the municipality where they moved in. It is not necessary, as it is in the United States, to enrol to vote. Furthermore, candidates are able to get a list of automatically registered voters (which is based on their residential registration) prior to the election. As a result, the candidates can monitor whether their supporters actually change their addresses and sanction them if they do not.

Photo by Shinichi Higashi: Flickr

A small number of votes could change an election outcome

Why would candidates go to such petty lengths to increase their number of votes? In Japanese municipal elections, there is a high probability that a small number of votes could change the result of an election. Japanese municipal assembly elections use the single non-transferable voting system, which was used in the elections for the national House of Representatives until 1993. Under this electoral system, particularly when the number of seats is large, a large number of candidates compete with one another within the same district. More importantly, as the entire municipality constitutes an at-large electoral district in a municipal assembly election, voters must often choose a single candidate from 10 to 100 candidates (depending on the size of municipality).

Under this system, the difference in total votes between candidates tends to be extremely small. It is not rare for the winning margin between an elected candidate (with the smallest number of votes) and a losing candidate (with the largest number of votes) to be less than 10 votes. Generally speaking, the smaller the municipality size, the smaller this difference becomes — meaning that the incentive increases for candidates to try whatever means they can, including any slightly ‘shady’ methods, to gain additional votes.

The ‘shady’ cases reported in the media may give readers the impression that they are just one-offs, but we believe that this is in fact the tip of the iceberg. It is for this reason that it is important to uncover the extent of the problem through statistical analysis.

Photo by Yasunari(康就) Nakamura(中村): Flickr

The timing of municipal elections in Japan as a ‘natural experiment’

Our statistical analysis took into account the variation in election timing. In 1947, when the simultaneous municipal elections took place for the first time, all prefectural and municipal elections were held in April. But after this, due to various reasons, the number of municipalities that held their municipal elections off the four-year cycle increased.

Specifically, our analysis focused on municipal elections in 2003 and compared municipalities that held elections on 27 April 2003 and those that did not. We consider that this variation approximates to an ‘as-if’ random assignment, because the reasons for ‘dropping out’ of the four-year simultaneous election cycle (for example, the governor’s death, the Showa mergers of 1950) could not plausibly have affected the politics, economics and social environments in the 2000s.

In medical testing, samples (such as mice) are divided randomly into two groups and only one of these groups is given a ‘treatment’. Before the experiment begins, the characteristics of both groups (for example, their size) should be basically the same because of random assignment. As a result, any significant difference between the control group and the treatment group, after the experiment, must therefore be attributable to the treatment itself.

In natural experiments, the subjects (in our case, municipalities) are not divided randomly by the researchers but instead by naturally occurring random events. This approach is currently going through a massive boom in political science and economics.

We obtained municipality-level data of the number of people who moved into each municipality from January 2001 to December 2004, and looked at the increases and decreases in the reported number of ‘new’ residents compared to the same month from the previous year. We then compared these statistics between those municipalities with an election on 27 April 2003 and those without. The results showed that the number of people moving into municipalities with an election (vis-à-vis municipalities without an election) spiked at around January 2003, particularly in small municipalities. In order to vote on 27 April 2003, people would have to have moved their residential registration three months earlier, by around mid-January. So, it can be argued that the abnormally sharp increase in residential registrations in January 2003 was a result of people moving their registration in time to vote in the upcoming election.

Based on the estimated parameters of our statistical analysis, we found that the number of fraudulent residential registrations was on average around 10 people in municipalities with an assembly election (but without a mayoral election). In around 20 per cent of the municipalities we investigated, the difference between the elected candidate with the least votes and the losing candidate with the most votes was within just 10 votes. So, the small number of fraudulent residential registrations we discovered was large enough to influence election results in many Japanese municipal elections.

Why detecting electoral fraud is important

This kind of fraudulent residential registration is not particularly rare and not limited to Japan. In fact, it has been reported across the world. However, much of the previous research on this issue has only looked at a few individual cases or undertaken relatively simple statistical analyses. Approaching this through the framework of a ‘natural experiment’, our paper is the first to estimate the scope of this kind of behaviour through detailed data analysis.

The results of our research may imply a dilemma regarding the design of the democratic system. One of the requirements of a healthy democracy is to ensure the political participation of as many people as possible. One of the ways to encourage this is to make the processes and procedures to gain voting rights as simple as possible. Yet, if these procedures are too simple, the risk is that they encourage dishonest political participation through fraudulent residential registrations. This is indeed the case in Japan.

What kind of system would increase the participation of honest voters and at the same time prevent exploitation of the system? This problem is important to all democracies but has yet to be debated satisfactorily in political science. The use of cutting edge methods to uncover fraudulent voters will be indispensable to furthering future discussion on the issue.


Fukumoto, Kentaro, and Yusaku Horiuchi. 2011. “Making Outsiders’ Votes Count: Detecting Electoral Fraud through a Natural Experiment.” American Political Science Review 105(03): 586–603.

This article was translated by Jill Mowbray-Tsutsumi. A version of this article was first published in Japanese in Nikkei Business Online on 19 December 2011.

About the author

Yusaku Horiuchi is Associate Professor in the Department of Government at Dartmouth College and holds the Mitsui Chair in the Study of Japan. He is also an AJRC Research Associate. He received his M.A. in international and development economics from Yale University in 1995 and Ph.D. in political science from the Massachusetts Institute of Technology in 2001. His research and teaching interests include comparative politics (electoral politics, political economy, public opinion, Japan) and political methodology (statistical methods, research design). Horiuchi is the author of Institutions, Incentives and Electoral Participation in Japan: Cross-Level and Cross-National Perspectives (Routledge 2005) and published numerous articles from leading political science journals.

Updated:  23 June 2017/Responsible Officer:  Crawford Engagement/Page Contact:  CAP Web Services Team