Seasonal affective disorder is a real and frighteningly common disorder that includes symptoms of depression that are triggered by certain times of the year. Contrary to popular belief, it’s not just caused by winter-time, and it got me thinking – what is the worst month of the year? I didn’t like the answers that were out there already, so I dove into some data and today I can present to you a definitive, irrefutable, data-derived answer to what is the worst month of the year.
It’s November. The answer is November.
If you want to see how I came to this conclusion, or just to understand what the hell that picture is telling you, read on…
What the others say
T.S. Eliot said April is the cruelest month, but that might just be because he’s never experienced February in New England. Meanwhile, global climate change has made vast swathes of the country tortuous hellscapes during the summer months.
I recently – and randomly – came across this article that talked about what was the best and worst months of the year, and which may have prompted this article. Before they broke it down by political party because it’s the Washington Post and they like to stay on brand, they showed the results of a Gallup Poll of what people selected as their favorite month:
The problem is that I don’t know what’s making people pick the month they pick. My guess is that for many of them, their favorite is their birth month, or possibly their anniversary month, or maybe it contains their favorite holiday. I suspect this is an example of availability bias, where one outside example tends to define the category. So maybe you love March because it’s your birthday, but that’s just one day out of 30.
Others have tried to use data to answer this question, producing Blue Monday, or the algorithmically determined worst day of the year. And that algorithm was:
where W=weather, d=debt, D=monthly salary, T=time since Christmas, Q=time since failing our new year’s resolutions, M=low motivational levels, and Na=the feeling of a need to take action.
The problem is that they didn’t give any indication of how they knew when we failed our new year’s resolution or our motivation levels, or what the hell the feeling of a need to take action even means. The other problem is that the day they selected as the worst day of the year was Martin Luther King Jr. day, which is really, really icky.
So no one else was giving a good answer, but thankfully I can now provide a definitive conclusion.
How I Went About It
I have to admit, I liked the idea from Blue Monday of measuring the time since you’ve failed your New Year’s resolution as a way of determining how bad a day is. I also remembered a story that I heard about how some sweets company, I want to say Nabisco, would track the ratio of 🙂 to 🙁 on Twitter and release coupons on days where 🙁 was more common, as a way to cheer the world up. I couldn’t track down any actual evidence of this, so if anyone else has heard of it please let me know in the comments.
Anyway, this gave me the idea of using Google search queries as a quick and dirty way of gaining insight into the mental health of America in a particular snapshot in time. Google kindly runs a pretty fun service called Google Trends that lets you see search term popularity over time. For example, you can see how many people search for the word “diet”:
And you can see a definitely yearly pattern of a spike right around new years, then a quick drop followed by a steady gradual decline. Sometimes you can even see the effects of moments in history, like when you search for suicide:
And that big spike there is when Robin Williams killed himself.
There is a bit of evidence that shows that using Google Trends is not a bad way to measure the salience – or mental awareness – of certain topics among the public over time. So from here it’s just a matter of determining which search terms will best represent overall mood.
Gathering the Data
I chose a mix of positive and negative search terms. For positive I chose:
- Exercise – assuming that people searching for exercise are trying to do exercises.
- Diet – for the same reason.
- Fun things to do in… – as a way of determine general activity level
- How to… – as a way of determining motivation to get things done
- I love… – just with an assumption that people are more likely to talk about the things they love if they’re in a good mood.
On the negative side I chose:
- Depression – thinking that people searching for depression are thinking about depression.
- Anxiety – ditto.
- I hate… – for the same but opposite reason as I love.
- Ways to kill yourself – I don’t really have to explain this one do I?
I pulled together my results in one data set that you can see for yourself if you want to check my work. I made the negative search terms into negative values, but I decided not to weight the search terms since Google Trends doesn’t give actual numbers of searches, just an “index number”, and I don’t know how weighting the values might throw that off. Beyond that, I didn’t do any actual statistics beyond just averaging the numbers and throwing them on a plot, so behold:
I know that plot is pretty busy, so feel free to click around. You can click on the search term to remove it or add it to the plot. The thick black line in the middle is the overall average. I think the best way to look at this is to take all of the terms off, then add them back on one at a time.
As far as what we’re looking at here, this is year-by-year since 2004. This isn’t going to tell us much about which month is the worst, but we can start to look at some patterning of the data. Keep in mind, I made the negative terms negatively valued, so when you see anxiety exponentially going down in the plot there, that’s a bad thing, as that means that people are searching for anxiety much more over time.
The point is, we do see strong regular cycles for “Diet”, “Exercise”, “Fun things to do in…” and “I hate…”
So what I did next was just take the results for each year and average it by month, so BEHOLD:
Once again, the big black line in the middle is the overall average, and… Well it’s not exactly breathtaking.
“Fun things to do near…” spikes in the warmer months. “Exercise” and “diet” gradually decline over the year, which I was surprised to see that there’s no “beach body” effect. “Depression” and “How to commit suicide” are highest in the late winter and early spring. “How to…”, “I love…”, and “Anxiety” are pretty much flat, and so is the overall average.
We shouldn’t be too surprised at this, after all the month of the year has much less to do with how people are feeling than the actual circumstances of their lives. But I wanted to zoom in more to see what those minor differences are, and I wanted to look at in a way that doesn’t put December and January on opposite sides of a chart even though they happen one right after another, so BEHOLD:
The further in the mark is towards the middle, the worse the month is. I want to highlight again that these are minor, minor differences, but now at least we can see them. So what does this tell us?
The nice thing about the circular plot is now we can see the big jump from December to January, so this highlights the importance of those New Year’s Resolutions. Then there’s a steady fall until – credit to T.S. Eliot – April. There’s a big jump for the summer months, then a steep decline starting in the dog days of August and reaching it’s nadir in the worst month: November.
This makes sense when you look at the search terms separately again. Healthy eating and good exercise falls steadily over the course of the year. The increase in activity and decrease in depression gives a big bump in the summer, but come fall those blues are in full swing. December gets a little bump from a slight decrease in anxiety and depression, leaving November as the absolute bottom.