Why Economic Predictions Always Fail Us?
Submitted by Nic Lenoir of ICAP
There is a sense when one reads sufficiently educated publications that a lot of people feel betrayed by financial and economic forecasts. One can argue whether some analysts foresaw what was about to unfold as early as 2006, but fact is most people had it completely wrong.
Financial forecasts are always tilted to the upside because overall we have a recent history of solid growth globally so our expectations suffer from positive recency bias, and also because even when the outlook is a little most negative it is in the financial services industry's best interests to spur optimism as sleeping money does generate any fees.
When it comes to economic forecasts obviously official forecasts are always tilted to the upside because it increases chances of reelection and it also allows higher spending with future high tax revenue being the offsetting entry on the accounting books. However one should expect independent economists to be much more accurate in their forecasts assuming they do not work for an employer with direct vested interest in overestimation of growth. Maybe that does not transpire because man in general is by nature optimistic. What I think is a much more interesting subject is the lack of granularity in economic analysis and statistical releases backing the analysis, whether the resulting conclusions are bullish or bearish for the economy.
Not an economist by training, I always grow frustrated when developments in forecasts don't match at all my rational understanding of what is happening in the economy. I have tried to pay more attention to it because being bearish, I was surprised by the extent of the rebound in manufacturing activity since March, and despite viewing this rebound as artificially generated by public demand or publicly sponsored private demand, it remains that statistically the releases showed dramatic improvement. At it turns out the answer I always received when inquiring was: "look at new orders", "have you seen inventory levels". A few things became obvious. First is the fact that a lot of econometric models use cycle analysis corresponding to the average duration and unfolding of the "traditional" business cycle, and the second is that a lot of projections are based on leading versus lagging indicators, but without all that much in depth digging into the reason why fundamentally the leading indicators are leading the lagging and where. As an example, many people talked about the second derivative turning in March, and in some cases ahead of time, called the turn in equities. Yet despite these timely calls, no one really convincingly at the time showed in numbers how public spending was starting to outpace the slowdown in private spending. There were people arguing that bail-out efforts were unprecedented and would lead to a turn, but they were absolutely not quantifying their arguments. So the average spectator was left with unquantified fundamental arguments and econometric forecasts with no solid fundamental arguments to back them up, and both faced a number of equally unquantified bearish arguments or even some econometric bearish projections. At the end of the day if an analysis is not robust enough then agreeing with its conclusions requires being biased towards agreeing in the first place, which renders the analysis useless.
That's why I try to stick to technical analysis most of the time, because I believe that enough understanding of the technical underpinning of the market and investor sentiment can increase odds towards a greater move one way or the other, without guarantying it will happen however. When it comes to economic analysis I have no claim of being able to provide a more quantified framework than anyone else, nor do I really have the time to. All one can do is use logic and try to analyze the research coming across one's desk, and that's my predicament.
The most shocking yet consistent conclusion I always reach, is the lack of depth of the data being presented. Take real estate for example. There is very little work done trying to provide public statistical data beyond price median and number of sales. For example how is the change in price in housing for each house not put in perspective with the price at which that house (if it's an existing house as opposed to a new construction) last sold, and when. If one month only multi-million homes traded then it would be the new median and the change in price would reflect this absurdity... Relying on only one or two numbers to describe a complex market lowers the relevance of any findings. We could also mention the number of foreclosures, the curb placed on that number, or the amount of cash spent in renovation before resale, but the point is made here. In the case of real estate it is easy to understand: there is little public resource thrown at the analysis, and private research is either costly or published by realtors, who are just as biased as financial analysts making buy or sell recommendations.
However beyond real estate, look at a statistic such has the average household income: all we get is the percentage change. That hardly tells anything at all. How is that divided across sectors or on the level of income is anyone's guess. Why is there so much focus on averages, a statistical concept that is at junior-level in terms of mathematical understanding? Don't we know how to look at distribution, multi-factor analysis, statistical relevance tests? There are billions of dollars literally gambled daily based on averages. If you look at the percentage change of the median income versus the 95th percentile income, you realize high incomes have grown since 1960 at 3 times the speed of median incomes, and I would be willing to bet that 99th percentile numbers would be even more telling. My basic underlying understanding of the economy is that the middle and lower class have had very little income appreciation over the last 40 years, if not negative growth once adjusted for inflation. That gap has been filled by a huge increase in credit which has allowed to maintain consumption, while the savings rate tanked until it became negative at one point (and that is an average, just pause and think... yes it is scary). In the meantime the amount of free money being saved at the fringe has been a huge factor in asset inflation, helped by abundant liquidity provided by governments in the Western economies. It is easy to say, and looking at income growth across percentiles one gets a first clue to quantify it, but until our statistical analysis becomes a lot more evolved, results will be very irrelevant in terms of prediction value.
Good luck trading,