There’s been a big leap in the number of Covid vaccines known to have phase 3 results since my last catch-up post…
Thanks to some virtuoso high-profile performances from a few women early on, this claim took off fast: having a woman at the head of your government was a major pandemic advantage. Let’s get real, though. There were just 21 female national leaders in May 2020 out of 195 UN-recognized countries – and some of those were heads of state with little or no involvement in political decision making. But even if all of them were individually fully responsible, that would still be too few to draw reliable conclusions about the influence of gender alone among all the complexity of a country’s response to a pandemic.
People of one gender don’t all behave the same way in leadership, of course. Other personal characteristics can lead to differences far more vast than the similarity of their gender. It’s not as simple, either, as female leaders being a good proxy for a country’s social and public health progress. More than 70 countries have had female heads of government, and that doesn’t mean they’re all more likely to be Norways or New Zealands: more than half aren’t from European or rich and progressive economies. And of course, if the next heads of government in Norway and New Zealand are men, that doesn’t on its own mean the countries will suddenly fundamentally change – or that countries previously led by women did when a man was subsequently leader.
For several reasons – and I’ll unpack that below – this “women versus men leaders” question on the pandemic doesn’t seem answerable to me. Yet I still find people’s attempts to do it fascinating. That’s mostly because I’m so interested in comparing countries generally. But I guess it’s also because I’d kind of like the claim to be true. And that personal bias puts me on red alert for jumping to a conclusion too quickly.
Exactly 2 years ago I was writing about my frustration at this claim going viral, and how uncritically people embraced it. And a couple of studies had recently seemed to confirm that women did better. They appeared in June 2020: a preprint by Supriya Garikipati and Uma Kambhampati (since published in a journal), and another by Soumik Purkayastha and colleagues. I discussed specific concerns about each of these studies, but beyond that, it was far too early in the pandemic to jump to conclusions: Covid had dug in for the long haul.
Within a few weeks of my op-ed, there were 5 more of these early studies, either released or submitted to journals. In the first of them, Luca Coscieme and colleagues agreed with the June authors that women leaders had done better for a couple of reasons – their policy responses to the pandemic, and the attributes of countries that had women leaders. Leah Windsor and co, though, concluded it was only that second factor, not the gender of the leaders.
The other authors also disagreed that gender was decisive. Andrea Aldrich and Nicholas Lotito analyzed data on specific preventive responses like school stay-at-home orders and public information campaigns in the early phase of the pandemic. They concluded countries with women leaders were not more likely to institute them. Mette Harder and Christoffer Harder studied pandemic responses, too, and they also concluded women leaders weren’t more likely to institute shutdowns – but if you narrowed it down to the OECD countries, the women-led countries shut down faster. And Jennifer Piscopo analyzed country wealth and capacity data, concluding the gender correlation others had found was “spurious…[because] many women-led countries score high on state capacity and that high-capacity states have low coronavirus mortality regardless of whether they are led by women or by men.”
Around that time – July 2020 – another study was about to start. Sebawit Bishu and colleagues ran surveys in 8 US states, having chosen states matched for gender and political party of governors. The results? Complicated. The governor’s gender didn’t have a clear impact on most of their measures of public perception and people’s intended behavior. Their political ideology did, though. On the other hand, Shrabanti Maity and Ummey Rummana Barlaskar, the authors of a later study, concluded that female leadership at the regional level in India was better at combatting Covid. They analyzed various data sources in states and territories across the whole of 2020.
Those regional studies point to some of the complexities I discussed back in 2020. In larger countries, a lot of power over pandemic response can reside at the provincial level. And for example in Australia, even some key national decisions were in response to pressure from state leaders (male and female) in a national cabinet.
What about studies that analyze more than just the first few months of the pandemic? I found a couple. A study that was published recently was the trigger for this post, although it only goes to the end of 2020. It’s by Dianna Chang and colleagues. There’s another that goes a bit longer – up to May 28, 2021 (by Cullen Hendrix). I haven’t found any yet with recent data, though. And that matters a lot: the last year has been very different from the pandemic’s first year-and-a-half. Female-led New Zealand’s population mortality rate has overtaken male-led Japan’s, for example, and is no longer so dramatically different to male-led Australia’s.
The authors of both of these studies looking at a longer time conclude that female leadership was an important factor in reducing Covid mortality. We’ll start with the Hendrix one, that looks at Covid outcomes in May 2021. Because he’s only looking at OECD countries, those led by women are a higher proportion than the 10% international rate – more like 20%. It’s still small, though: he classifies 8 countries as women-led. And therein lie a couple of major problems.
The first is immediately obvious when you look at which countries he names: half are 4 out of the 5 Nordic countries. The 5 Nordic countries are Denmark, Finland, Iceland, Norway, and Sweden. They’re important here, because they’re a geographic group of wealthy, equality-concerned countries with strong public health cultures, borders that are relatively easy to secure, large proportions of people living alone, and relatively small populations – these 4 have less than 6 million people each and Sweden has a little over 10 million.
Those 4 countries all having female leaders at the same time was new: 2019 was the first time it happened – and Sweden made it 5/5 late in 2021. (I’ve tallied up the years between 2011 and 2021 that each of them had a female head of government for at least part of the year here.) Had the pandemic hit in the few years before it did, there would have been 1 or 2 instead of 4: a couple of years later, and Sweden would have counted in the female-led column. Was the gender of these countries’ leaders at that point in time decisive?
Hendrix thinks it was. He writes, “…the COVID mortality rate in Sweden – the only Nordic country with a male executive – is three times higher than that of the next-highest Nordic country (Denmark…)”. Hmmm. I think this is a great example of the potholes people are falling into with simplistic takes on this issue. I don’t think Sweden is the Nordic anomaly because its head of government was male. His influence was limited, because the Swedish government was legislatively obliged to follow the advice of its public health agency. And the head of that agency was a major public health outlier in Covid advice – who has expressed regret for his contribution to the high mortality rate. Sweden arguably squandered a Nordic-wide advantage, not because there wasn’t a woman at the country’s helm, but because of who had their hands on a more powerful steering wheel in a public health emergency than the country’s leader.
The second major problem in the Hendrix analysis of 8 countries versus 29 is which countries he’s classified as women-led. Getting “us versus them” right is obviously critical in this kind of analysis – and he doesn’t get that right. Other than the 4 Nordic countries, he counts Germany, Lithuania, New Zealand, and Switzerland as women-led.
Lithuania is the only country in that group that had a similar mortality rate to Sweden at the time of his study. Until late November, though, it was traveling at around Norway’s very low rate. That matters, because it’s head of government was a man at that point: Lithuania only counts as woman-led from December, less than half the study period. Just a few weeks later, Estonia became woman-led, but he counts that country as male-led. He also counts Belgium as male-led, even though it was led by a woman for most of the time its mortality rate soared to the highest for a large-ish country in Europe. And finally, he counts Switzerland as women-led, even though the country has a collective form of government – a Federal Council, with a president that’s essentially a chairperson role and equal to the other members, rotating each year. In 2020 that chair was held by a woman, but from January it was a man. Given there are a couple of dozen other factors in the Hendrix analysis, it’s impossible to know how much impact this classification problem would have. But I don’t think it’s a valid gender comparison.
Which brings us to the Chang study. Those authors analyze 99 countries to the end of 2020, and they code 14 countries as female-led. However, they don’t report which ones they were. That’s an important thing to be transparent about. The authors say they chose “political leaders”, and that could mean they included only heads of government: but if they did, their sample happens to include pretty much every country with a female one, while not counting about 100 with male heads of government. According to my quick rough count on Wikipedia’s list, there were 15 female heads of government at some point in 2020 after the pandemic started, and 11 female heads of state (monarchs not included). Some had only been in office a few days, or only fairly early in the year. (Wikipedia also has a list of openly LGBT leaders – there were no transgender national leaders.)
The half of male-led countries included in the Chang study is not evenly distributed across the globe. From a quick look, for example, it’s less than half of Africa’s countries, but way over half of Europe’s, and all of the Americas. And on the face of it, it seems impossible that the countries with male and female political leaders are represented in the same way. It looks as though if you’re a woman leading a very small country, for example, you’re more likely to be included than if you are a man.
The authors don’t report at what time point they classified gender of the countries’ leadership for 2020. Some countries with very high Covid mortality rates had leadership changes between October and December. Some of those new leaders were of different gender: for example, Belgium and Bolivia went from women to men leaders; Lithuania and Moldova from male leaders to female. When a correlation is as fragile as the one reported in this paper, it doesn’t take much to tip the scales.
Gender of the national leader is one of 21 factors about countries that they include – none of them consider pandemic response. When you have a lot of elements in an analysis like this, you run a lot of risks: on the one hand, trawling for “hits” with so many datasets drives up your chances of getting a coincidental correlation. (For more on that, see my post on the potholes of statistical significance testing.)
Yet at the same time, amongst your big set of data points, you could be misled by not taking critical issues into account. For example, the factors include a country’s weekly average temperature, but not how easy it is for a national leader to ensure borders are secured. It includes age, but not what proportion of people live in multi-generational households or congregate homes.
That last one is pretty critical for an infectious airborne disease. This study includes population density and urbanization, but not household size. So that means that highly urbanized countries with large tracts of unpopulated or sparsely populated land are seen as fundamentally different from those without them. But the critical difference of how much easier it is to reduce exposure to the virus if you live in a single-person household isn’t taken into account. Even within the EU, those differences are major: less than 20% of households are single adult in Portugal and Slovakia for example, but it’s close to 50% in the EU’s Nordic countries. Or consider the OECD: only 13% of all seniors live alone in Mexico, compared to 46% in Denmark.
Beyond the usual problems of data on key points not being equally reliable from country to country – including for Covid mortality – there are data points in this particular study that are particularly fuzzy, like a country’s rated happiness. They also aggregate data on investment in emergent technology, 4G mobile coverage, and social media use into a single data point called technology. They argue those were critical for contact tracing and largely accurate dissemination of information, and that was a key pandemic advantage. While that constellation may be a positive one in some countries, I’m dubious about that assumption globally. Consider the incredibly impressive contact tracing in African countries that would score lower on “technology” here, coupled with an awful level of misinformation on social media.
When you’re picking 21 factors for something as complex as international comparisons, there are just so many ways to end up with spurious findings. Using simple datasets with often wildly different quality across countries to answer extremely complex socio-political questions isn’t a solid path to get to reliable conclusions. The sight of all that data should set off alarm bells. Too often, though, it seems to push people into an excess of conviction about the findings.
Disclosures: I identify as female. I grew up in Australia, and have lived in Germany and the US. I was back in Australia from before the pandemic. It had a male leader, and so did the state I live in (Victoria). Australia has had 1 female head of government, and 1 female head of state. More about me.