Making predictions just got harder

Question marks over our ability to predict the future provided two of the highlights of our annual pensions conference. First, Jon Palin, a longevity actuary at Barnett Waddingham, showed how our assumptions around improving mortality seem to have been disrupted by a decline in healthcare spending, compounded by a virulent flu season. And then, unscripted, our keynote speaker, the broadcaster Andrew Neil, gave a bravura analysis on the failings of pollsters to guess recent election results.

Our analysis of recent mortality data shows that the steady 2.5 percent improvement in pensioner mortality between 2000 and 2011 turned into a 0.3 percent annual improvement in the five years that followed to 2016. The question everyone is asking is, was this just a blip or is part of a new trend?  

Well, it appears to be a bit of both. There was a spike in flu-related deaths in early 2015 due in part to the low effectiveness of the vaccine that year, although this isn’t the whole story because we have also seen more deaths than expected in successive summers. But Jon’s contention was that the most compelling evidence for a new trend so far is the discrepancy in health and social care spending by the previous two governments. In the Blair-Brown years, the percentage of GDP spent protecting the health and wellbeing of our population grew at the highest rate in the history of the NHS. Growth since then has been flat and waiting lists are rising. Does this explain the decline in mortality improvement? It’s a highly intriguing theory and one that should be factored into how trustees set the mortality improvement rate for their schemes.

We expect to see much more debate between companies and trustees on this issue, particularly as the CMI has suggested that the rate of improvement is less certain. It has added a new “period smoothing parameter” to its model to make it easier for users to take a view on initial rates of mortality improvement, as well as the traditional long-term improvement assumption. (It should be noted that the CMI’s new core view on mortality improvements is 1.4 percent).

Another area of uncertainty in our lives is whether we can ever again rely on the predictions of pollsters ahead of large elections. A member of our audience quizzed Andrew Neil for his wisdom on this. His theory is that voters are becoming more shy of giving their intentions at a time when politics are veering from the post-war consensus. He cited shy Tories for the predictions of a Milliband victory in the 2015 election. He cited the experience of a woman interviewed by pollsters in the south of Glasgow who had a Vote Yes poster outside her home just so that she “did not get intimidated”. He warned that low polling figures for Le Pen in the French elections might be wrong as a result of such behaviour. He added that pollsters were also at fault for massaging their figures so as not to appear too different to other researchers, in case their figures proved to be wrong. He cited how one pollster did predict a David Cameron victory in 2015 but buried the figures.

It was thrilling to hear such insights and I hope that, along with our own analysis, we can deliver the same level of interest to you at this event next year.