Lesen Sie die deutsche Version hier.

ZEIT ONLINE polled users around the world in seven different languages, asking the question: "Who do you choose? Trump or Clinton?" So far, 273,000 votes from 220 countries have been submitted. 

Many readers shared the poll using social media and ZEIT ONLINE also conducted multilingual campaigns on Facebook. Meanwhile, other media, including the Italian daily La Repubblica, embedded the tool on their sites, as did German international broadcaster Deutsche Welle, which built the voting tool into a number of its multilingual channels, from Russian to Swahili. US national public radio station NPR likewise reported on our survey on its morning show.

As such, our lighthearted survey achieved its primary goal: that of getting people around the world to think about their attitudes on the US election, one that is of major importance to far more than just American citizens. So what did we learn from it?

Real-Time Manipulation

So far, we have received exactly 273,000 votes, but we have validated only 67 percent of them, having filtered out the others as suspicious. We figured some individuals might try to manipulate our poll results, but the speed with which this happened took us by surprise. Not even two hours after initial publication, a freshly dispatched voting bot had already registered the first 1,000 Trump votes for Austria.

Such bots generated a high number of invalid votes and shortly after launching the survey, we felt compelled to take steps to significantly increase our anti-bot measures. To be on the safe side, we purged all initial votes, even ones that had been cast by real people.

Once we refined our filtering mechanism, enabling it to weed out certain types of manipulation, the share of invalid votes shrank to 9 percent. Even our current voting tool isn't immune to more sophisticated attacks, but the numbers have been far more stable since resetting our database.

If bots could vote, though, Trump would be their man. Some 76 percent of the invalid votes that we have purged were cast for the Republican candidate.

What We Know about the Adjusted Figures

Overall, Hillary Clinton received 69 percent of the valid votes, with 31 percent going to Donald Trump. A total of 61 percent of all the votes cast originated from Germany. Over 1,000 votes each came from 18 different countries and more than 100 votes each from 61 countries. In total, people from 220 countries participated.

Of course, neither the high number of votes from Germany nor those collected from the individual countries offer a representative sample. The fact that we track the country where a user is located on the basis on their IP address also leads to data impurities. But users did have the option of changing that information when they submitted their votes.

So even if the data does suggest it, we cannot definitively say that the majority of people Poland or Russia, for example, would vote for Trump.

It is nevertheless interesting to compare our results with those of more representative polls. Germany's ARD-DeutschlandTrend survey, conducted on behalf of the country's leading public broadcaster, for example, shows that 75 percent of Germans would vote for Hillary Clinton if they were allowed to participate in the US election. Our voting tool showed 71 percent casting ballots in her favor. The results of WIN/Gallup International's Global Poll on the American Election shows similar figures.

But in some rare instances, the ZEIT ONLINE results vary significantly, particularly in France. There is also a divergence in the US, where Trump -- depending on the representative survey -- can count on roughly 44 percent support, a result that is much higher than the 29 percent found by the ZEIT ONLINE survey.

What, then, have we learned? Many, but certainly not all, citizens of the world favor Hillary Clinton. In some countries, including Russia and Poland, there may even be a preference for Donald Trump. The significant support for the Republican candidate here in Germany, at 29 percent, came as a surprise.