Is it Easier for Investors to Forecast the Long-Term or Short-Term?

I am in a forecasting competition and asked to predict global iPhone sales. I can choose to make an estimate over one of two time periods – either the next three months or the next ten years. Which would I prefer? This is a simple decision – the next three months. There will be far less noise and uncertainty over the quarter then there will be over the decade. This aligns with Philip Tetlock’s work on super-forecasting, which suggests that predictions for the near-term are likely to be more accurate than those we make for the distant future. Does it therefore follow that investors are best placed to focus their attentions on the short-term? Almost certainly not. In fact, where we do have to make forecasts, it typically is in our interest to take a long-term view. Why is investing an exception?

Before exploring this idea further, it is worth starting with a caveat. Forecasting most things is difficult and we are notoriously bad at it. Predicting the behaviour of a complex adaptive system such as financial markets is particularly challenging. Our default setting should be to avoid it, where at all possible.

But as investors we must make certain types of projections, even if we don’t realise it. When we build a portfolio the decisions we make (how much to allocate to equities? How much to bonds?) are underpinned by expectations about the future. Not all forecasts are created equal however, even when they are about the same asset class.

Let’s take equities. In very simple terms if we want to take a view on the prospects for global equities then there are two things that matter: sentiment (how other investors ‘value’ them) and fundamentals (the earnings stream we receive through time). The importance of each element alters dramatically depending on the time horizon applied.

The shorter the time period the more sentiment dominates outcomes. This creates a prediction problem. It is close to impossible to hold confident views about what events will occur and how market participants (in aggregate) will react to them. The amount of noise and randomness is pronounced, and the range of potential outcomes vast.

As the horizon extends, however, the fundamental factors start to matter more. The compound impact of earnings begins to overwhelm the influence of fluctuating sentiment. If we are making a ten year forecast for an asset class, we are thinking about the accumulation of cash flows / earnings over that entire period, not just what they are by the end of it.

The influence of multiple years of earnings should mean that the range of outcomes narrows as our time horizon increases. In the graph below we can see the breadth of annualised returns for global equities since the late 1980s:

A one year horizon is the point of peak sentiment driven uncertainty, which then tapers materially as the period extends.

If we are investing in global equity markets today, we know the starting yield, the valuation and can set a prudent expectation for growth (long-run global GDP). It is reasonable to expect our returns to gravitate towards this over the long-run. There are clearly no guarantees here and there remain profound uncertainties. We should, however, have more confidence in these types of assumptions than in what might happen in equity markets over the next year.

Although it is more likely that we will enjoy success in making long-term asset class return forecasts, it is important to define success. We are not going to predict ten year returns accurately to 2 decimal places (or indeed 1); one of the few things we can be certain about is that such point forecasts will be wrong. This does not mean, however, that long run views are without value; there are many situations where they get the odds on our side.

Take the below three examples. These are all scenarios where I would prefer to make a ten year to a one year prediction:

  • Setting a sensible range of expectations of where asset class returns will reside: I would expect the dispersion of returns to narrow as the time horizon moves past twelve months.
  • Ranking asset classes by relative preference: I have no idea about how equities will fare compared to government bonds over the next year, give me ten years and I will be far more confident.
  • Judging the likelihood of positive absolute performance: The odds of positive returns from most traditional asset classes improve as our time horizon extends.

The benefits of adopting a long-term approach when setting return expectations is reliant on being diversified. If we are heavily concentrated (for example owning a single stock) we should not expect the spread of potential outcomes to contract materially as our time horizons extend because of the ever-present spectre of idiosyncratic risk.

Whenever we make a forecast we should seek to understand what the most influential variables are and how unpredictable they are likely to be. For investors making decisions based on short-run asset class performance it is critical to be aware that outcomes will be incredibly noisy, driven by erratic sentiment and have a wide range of potential paths. Although imperfect, extending our time horizons can give us a little more confidence.

My first book has just been published! The Intelligent Fund Investor explores the beliefs and behaviours that lead investors astray, and shows how we can make better decisions. You can get a copy here (UK) or here (US).