“Things are rarely as bad, or as good as they seem”
Variants of this quote have been attributed to different people over the years. It’s a pithy formulation that helps people to unpack some of the confusing and unexpected outcomes that life has a tendency to produce. For the first few days of this week many people in Moldova were shocked to find that the election and referendum results were not as good as expected. It would not be hyperbolic to say that Monday and Tuesday had a distinctly funereal atmosphere among pro-EU voters in Moldova.
But this quote goes both ways, things are often not as bad as they appear either - they’re just confusing. In today’s perspective article I’m going to try and unpack what happened last weekend and provide my analysis and opinion about why the vote went the way it did - and what it means for November 3rd’s runoff.
Understanding the Results
One of the main narratives this week is around the failure of pre-election polls to accurately forecast the results. In past articles we’ve mostly paid attention to polls conducted by Watchdog MD in partnership with CBS-Research. This is because they were the most prolific pollster tracking public opinion month to month and because they have a strong track record from past elections. Other polls, such as those produced by iData and Intelect Group showed broadly similar trends both in the presidential election and referendum.
In a nutshell these polls predicted “Yes” vote of between 44% and 55% and a “No” vote of between 23% and 35% inside Moldova. At the same time a huge percentage of voters, at least 10% and as high as 30% in some polls were undecided or unsure if they would vote at all. Given past electoral trends which would assume many of these voters would stay home, plus the fact that the very pro-EU diaspora is not included in these polls, there was an assumption that the referendum would pass by a healthy margin. So what went wrong?
iData did a post election analysis and identified 3 problems that they believe account for the polling miss - in addition to vote buying (which we will come to later). Firstly, they stated that the turnout skewed older than their sampling predicted. Secondly, there was a relatively high correlation between a candidate’s preference on the EU and voters choice in the referendum. Since Maia Sandu was the only significant pro-EU candidate this had an effect. And thirdly, they identified a protest vote by correlating questions asking whether the country is on the right track / wrong track with referendum votes. Those who believed the country was on the wrong track voted more heavily against the referendum, which iData asserts was a protest vote against the current government.
Even with these corrections their poll doesn’t match the reality of the vote and iData estimates a maximum of 60,450 voters being directly bribed - or 4%.
Watchdog’s preliminary post election analysis, as given by Vasile Cantarji of CBS-Research, focused much more directly on vote buying. He stated:
"It seems that we all know and accept that there is a phenomenon of vote buying by Shor. Let's see how serious we understand this. The police claimed 130,000 votes were bought, while Maia Sandu spoke of 300,000. I estimate the number of voters involved in this scheme at around 250,000,"
He noted that even removing the lower number of 130,000 votes against the referendum the Watchdog polls predicting around 55% for “Yes” come out as accurate.
While the Watchdog and iData analyses seem to diverge a bit that isn’t necessarily the case. Changes in turnout - especially the mass mobilization of the older generation is as likely a result of interference as it is a sampling error. Importantly vote buying was not the only mechanism of Russian interference - we’ll get into that more below.