What percentage of the Greek police voted for the far-right Golden Dawn ("Chrysi Avgi") party in the general elections of May and June?
We will never know exactly, because ballots are cast in secret. But when we studied the results from individual polling stations, we noticed a strange pattern.
Golden Dawn's share of the vote, which in central Athens (the electoral constituency known as "Athens A") was 7.8 per cent, fluctuated wildly between polling stations, some of which are very close to each other.
Several months ago the Greek newspaper To Vima also noticed this pattern and matched the polling stations where Golden Dawn did very well in Athens with those where on-duty police officers are given special permission to vote on election day.
I was intrigued by this, but was reluctant to draw any conclusions until I could be absolutely sure that I knew where the police had voted. In theory, this is public information that should be provided by the Greek ministry of the interior.
In practice, it proved extremely hard to obtain. Eventually, we did receive this information, and were able to see a consistent match between polling stations where Golden Dawn did well, and the 88 polling stations in Athens where uniformed officers (not just the police, but also the military, the fire brigade and first-aid paramedics) were instructed to vote, alongside local residents.
This applied across a wide variety of neighbourhoods, of varying social and economic characteristics.
The most intriguing fluctuations came in polling stations around the Attica General Police Directorate (or GADA), where we know hundreds of police officers voted.
In 13 nearby polling stations, Golden Dawn averaged more than 20 per cent of the vote, whereas in the neighbouring "civilian only" polling stations it received 6 per cent of the vote, or below the Athenian average.
In other words, it seems that it is the police presence that is the variable, which is pushing the Golden Dawn vote upwards.
If we make this assumption (and yes, it is an assumption), we can conclude that the percentage of police who are voting for Golden Dawn is consistently higher than the actual result from the polling station.
To get a more accurate idea of the percentage of Greek policemen who voted for Golden Dawn, you'd have to know what proportion of the voters at any polling station were police, and also how many actually cast their ballots.
I did not have this information. Several Greek publications have come to their own startling conclusions in this regard.
I went to the Greek police, and asked them what they made of this evidence. A police spokesman said that any Greek citizen has the right to vote for any legally registered party.
Voting preferences, he said, are a private affair, provided an on-duty officer carries out his (or her) work with integrity, professionalism, and without discrimination towards any specific person or group of people.
So why might so many policemen in Athens vote for an extremist party like Golden Dawn? Here are my own three theories, which are not mutually exclusive.
1) There is a tradition of authoritarianism in parts of the Greek police, which dates back to the time of the military Junta and beyond.
2) There is a feeling among some Greek policemen, working on the rundown streets of central Athens, that "my enemy's enemy is my friend". The police feel overwhelmed by the influx of immigrants, and they also feel hostility from anarchist and leftist groups. In some neighbourhoods, they will see Golden Dawn as a more sympathetic alternative.
3) The average Greek policeman, like many other Greek civil servants, is disillusioned and frightened. Morale is low. Pay has fallen by approximately 30 per cent in the past two years, and there are more cuts to come. The policeman voted for Golden Dawn as a protest, like hundreds of thousands of other Greeks.