The markets had a great finish to the year. From November 8, 2016, most stock markets worldwide went up – for instance, the S&P/TSX Canadian market was up 4.8%. And if you have a Canadian pension plan subject to solvency rules, your liabilities are down 4%, so your solvency position has improved by 6% since November 8. A big jump in just seven weeks.
Of course, November 8th was presidential election day in the U.S.
To recognize the vast improvement in investment funds (and pension funds in particular) immediately following the election, I suggested to my former colleagues (“former,” because I’ve retired) at George & Bell Consulting that I write an article titled In Praise of Donald Trump. The title of course was a bit tongue-in-cheek, but I had hardly hit the “Enter” button when I got replies: “There goes our client base!”, “If the article is cheeky/playful …”, “… but I still like the title as it will generate some talk given how ‘provocative’ it is.” So here is the article watered down, both in title and content. And an apology when I wander off topic a bit – there is lots to write about.
As actuaries, we have considerable education and experience creating and using “statistics.” More precisely, we actuaries use “statistical inference” in our work. Wikipedia gives a good definition:
Statistical inference is the process of deducing properties of an underlying distribution by analysis of data.
Statistical inference in turn relies on statistical sampling, defined as:
Statistical sampling is concerned with the selection of a subset of individuals from within statistical population to estimate characteristics of the whole population.
Simply put, to identify the characteristics of a large group of things (often people), which statisticians refer to as the “population”, without gathering the relevant information from each and every person or thing in that population, actuaries and statisticians will take a “sample” of the population from which they infer the characteristics of the group as a whole.
The mathematics underlying statistical sampling and statistical inference is very robust, and the results are often very useful if the process is done properly. A required condition of sampling “properly” is using a sample that resembles the population in every way, except size. In an election poll, the pollsters will survey individuals across the broad electorate, with the intention of creating a sample having similar regional, age, gender, education characteristics (amongst others) as the entire population. But because the process of polling is expensive and the pollsters feel compelled to conduct a survey each and every day in the run-up to the election, pollsters will minimize the size of the sample to the minimum size acceptable by mathematical theory. This leads to the statement often attached to the announced results of the survey, that the results are valid “19 times out of 20”.
But the sample must resemble the population, and conducting a poll of eligible voters isn’t the same as conducting a poll of individuals who are likely to vote. Much was made in the U.S. of Clinton supporters not voting; they may have told the pollsters that they would vote for Clinton if they chose to vote, but they didn’t vote. So the polls might have shown Clinton as the expected winner of the election because of her many supporters, but when they didn’t vote, she lost.
A second flaw is that a polled voter’s stated intention isn’t a verifiable objective fact. A survey might ask an individual their height (they might admit to the correct value) or their weight (although they might give the pollster their weight from 20 years ago) or their age (which they tell the cashier at the liquor store), but those measures are all objectively determinable. An individual on the other hand might be asked who they intend to vote for, and rather than admitting they’d be voting for a racist, misogynist bigot, who is attracting their vote because that candidate would “make their country great again”, they would tell the pollster instead that they would vote for the other candidate because who would ever admit to voting for a racist, misogynist bigot? No doubt some of the people polled told the pollsters fiction, but why not?
Pollsters have really gotten things wrong the last few years. Before the Trump election, there was Brexit. And in Canada polls have been absolutely wrong in forecasting the results in recent federal and provincial elections. Flawed polls may appear to be harmless, but they could be consequential if:
- individuals who are persuaded that their candidate is a “shoo-in” don’t vote, and their preferred candidate doesn’t win, or
- candidates themselves change their actions based on poll results (for example, Clinton not spending much time at all in states she thought she was solid in, but ended up losing).
Statistics are mighty useful in actuarial work, but they must be used with care because they can be dangerous if used the wrong way.
The Game of Bridge
Since retiring, I have returned to the game of bridge; it is something I played with enthusiasm when I was younger before family and career got in the way. It is a fine game that combines sharp analysis, crafty psychology applied against shifty opponents and even craftier psychology in trying to communicate (legally) with a partner who you hope is paying attention. And as a returnee to bridge, I have once again found myself to be the “youngster”, as the game is popular almost exclusively with the older crowd.
On November 8th, the day of the U.S. Presidential Election, I was playing in a bridge tournament in Whistler. It was a “Regional” event, which means there were people from beyond the local area, including the U.S. The fact that the word “trump” is both an important concept in bridge (describing a playing card of the suit chosen to rank above the others), and the name of the Republican presidential candidate was not lost on the savvy competitors. Some wore T-shirts and hats bearing slogans linking the two meanings of the word, for instance “It doesn’t matter if your hand is weak or strong, BID NO TRUMP.”
There were three sessions, one in each of the morning, afternoon and evening, on November 8th. At the dinner break, it had become clear that Trump might win, and between rounds during the evening session as people checked their phones for results and as whispers started spreading through the crowd (and these games are so quiet the whispers can be quite audible) it had almost become a certainty. One of our group had chosen to stay in our rented condo during the evening game to watch the results come in poll by poll, and as we called him between rounds it was clear he was becoming increasingly despondent, evidenced by the markedly lower level in the bottle of the Scottish elixir when we had returned to the room.
So the pollsters got it wrong. And as it became clear that Trump was likely going to win, there were also reports that the financial markets were tanking – although the North American markets were closed overnight, it was possible to “bet” on the American market in the futures market. And the betting all had the markets down. Really down.
But when the sun rose on the U.S. Eastern seaboard on November 9th, it shone on the American markets and most markets globally. And the markets have continued to rise in the seven weeks following the election. So the overnight financial punters joined the pollsters in the league of idiots.
The year 2016 was a remarkable one in the financial markets. Stock markets generally started poorly in January, then gradually reversed back into positive territory, increasing sharply after November 8th.
The S&P/TSX in Canada and the S&P 500 in the United States had strong returns both pre- and post-election. The MSCI EAFE index, which represents non-North American markets, was down 2% through to November 8th, but ended the year in positive territory.
|Up to November 8th
|After November 8th
|S&P 500 (local $)
|MSCI EAFE (local $)
The results of the fixed income market were equally startling. Yields on 30-year Government of Canada bonds dropped from 2.2% to 1.9% from the beginning of the year to November 8th, then bounced back to 2.3% at the end of the year. The changes in interest rates caused the FTSE TMX Canada Universe Bond Index to increase by 3.8% up to the election and decrease by 2.1% in the last seven weeks, for a total roundtrip return of 1.7%.
For a balanced fund with a typical mix of stocks and bonds, the fund was up 5.7% to November 8th, up an additional 1.8% to year-end, for a total of 7.7% for the year.
More importantly for pension funds subject to solvency rules, the increase in bond yields in the last seven weeks pushed liabilities down by 4%-5% , so the solvency position has improved by 6%-7% since the U.S. election (assuming a balanced portfolio). Even those pension and benefit plans not subject to solvency rules have improved because actuaries will be able to use a higher discount rate when conducting the next actuarial valuation.
But why have markets responded this way?
The most common explanation of the market behaviour is that president-elect Trump was elected on a pro-growth agenda. His government will borrow trillions of dollars to invest in the country’s infrastructure. According to the theory, and as anticipated by market participants, that will increase the country’s economic activity, employment and the profitability of companies participating in the activity. So the stock markets have responded positively.
In addition, the vast volume of borrowing will crowd the market of borrowers who will have to offer increasingly higher interest rates on their securities. So interest rates have gone up.
And that’s the theory. But it can’t help but feel like riding in the last car of a rollercoaster up a long hill and watching the front car disappear over the crest.