Strengths and weaknesses of the forecast

From using the forecasting method above it should have been possible to predict that the economy would stagnate, and with high likelihood enter a recession, about 1,5 years before the dated recession. Also I argued that forecasters using this method should have been able to predict that the future recession would with high likelihood be of great magnitude at least 6 months before December 2007.

But with this said, this forecasting approach carries a number of more or less vital strengths and weaknesses. With this in mind the following discussion of quality is not exhaustive, but carries some of the most important points. The discussion will be much compared to a more econometric approach to forecasting which arguably carries the biggest differences from the approach described in this paper.

The quality of the forecast

While most econometric approaches has a specified time horizon and give more exact predictions, the flexibility of the approach from this paper makes the analysis less specified. There is a rule for all forecasting that longer time-horizons are more difficult to predict, but with the discussed approach you create qualified expectations for the trends of the next 6-18 months, while the magnitude of the expected trends are often revised as more information is made available.

From the earlier analysis it was possible to predict the 2007 recession in 2006, but the expectations about magnitude was only obvious 6 months ahead of the peak. But with all the uncertainty that comes with forecasting you cannot depend on any models to forecast with a hundred percent accuracy. This means that it is the forecasted future trends which are of importance, and not necessarily the exact day the business cycle is going to turn.

This is probably the most argued drawback of forecasting in general. If the forecaster is not able to predict anything with certainty, then why should we spend time and resources on trying to predict the future? While it is for obvious reasons true that the forecast performed on the 2007 recession would not have eliminated all uncertainty in real time, the expectations created from the analysis would have been a great preparation, and of such quality that it should indeed have been taken seriously.

Economic forecasting have been researched for many years, and while this paper only holds evidence on the possibilities of forecasting the 2007 recession, other research have proved the possibilities of forecasting earlier recessions.

The basis for scenario building

While it is impossible to remove all uncertainty, the information gathered from the forecast gave clear indications that there was a recession coming, and that the housing market was likely to experience a significant downturn. This was known at a relatively early stage and could then have been used as the basis of realistic internal and external scenario building.

As the expectations on the depth of the recession did not come around until 6 months before the peak, forecasters could have used the information available to create scenarios on how a recession triggered by increased interest rates, high levels of debt and unwarranted growth in house prices, would affect company performance.

From this managers could get expectations on whether the markets they are competing in would be leading or lagging the business cycle, and approximately how much this would pressure their individual cash flows, and the demand for the goods or services they were providing.

This could help creating potential strategies to take the best possible advantage of both the upside and downside risks. In turn, this means that even though the future still holds uncertainty, also after performing economic forecasts, forecasting can help limit the surprises of the internal effects from changes in the business cycle.

Technical strengths and weaknesses

The most evident strength of the forecast above is its flexibility. While both the society and the economy are constantly evolving, flexibility is arguably a vital and much needed attribute. The freedom in the choice of indicators means that you can easily include different markets and sectors which are of importance to your particular analysis. And as the markets are continuously changing, the forecaster can easily change the indicators towards new information which are of more importance to the modern economy.

This can sometimes be a problem for more specified econometric approaches where you often are more dependent on a few time series with strong empirical data, and a long and stable statistical relationship. An example of a time series which suddenly became relevant is the number of initial claims for unemployment insurance. While this time series suffered from such high volatility that forecasters did not find it useful, the increased monitoring by the Labor Department have made it a helpful forecasting tool in the modern economy, and it certainly gave important signs of trouble ahead of the 2001 recession.

While historical relationships are important in terms of confirming the theories behind the indicators used and to make it possible to learn from the past, it plays a more direct role in econometric modeling. In those approaches you often depend on the movements from the past to be the same in the future.

While this indeed can be helpful, you must at the same time remember that these approaches move away from the possibilities of new and different correlations in the future. A good example of this is the famous Fed-model, from which Alan Greenspan predicted a negative relationship between short-term interest rates and the stock market. This proved empirically correct for many years and also seems intuitively correct.

But nevertheless, it failed fundamentally during the dot.com bubble where the stock markets crashed while the interest rates were falling (Shiller 2005). The approach explained in this paper on the other hand would have seen the unusually high prices, and would have found that the prices were not fundamentally justified ¹.

The same was also the case for the housing market in the forecast just performed for the 2007 recession. From history and theory one would expect house prices to stabilize as the interest rates kept falling, but as the record prices were not driven by fundamentals, the prices kept falling towards a new equilibrium even after the interest rates were set to record low levels.

In other words, while econometric approaches are expecting the same to happen in the future as in the past, the approach described in this paper learns from the past but is still flexible enough to evolve together with the economy.

The point that more statistical approaches follow a certain methodology can arguably be a strength compared to more flexible approaches. This often makes them easier to work with and return more straight forward results. While econometric models have carefully designed steps which are done the same way for each forecast, the approach in this paper on the other hand leaves more freedom to the forecaster on how to interpret the indicators and draw conclusions.

This leaves more weight to the analytical skills and experience of the forecaster which arguably could bias the forecast. As most forecasting approaches learn how to make qualified predictions about the future through understanding history, experience can make an important difference especially in flexible approaches.

The biased forecast

While I have argued that the flexibility in the forecasting approach from this paper makes the forecast stronger in the ever evolving world of economics, this flexibility could certainly bias the forecast as well. The statements from Van Der Stede (2009) on overly positive and negative markets arguably support Linda M. H. Lai (1994) in the view that some psychological phenomenons also play an important role in business cycle economics.

As the forecasting approach analyzed in this paper is flexible and gives much freedom to the interpretation and choice of indicators, this approach could be especially prone to suffer from forecaster biases.

Overconfidence and the confirmation trap

Linda M. H. Lai (1994) argues that managers overestimate their own abilities and enter a stage of overconfidence in upturns, while they get overly pessimistic when the economy enters recessions. Such behavior is in accordance with Van Der Stede’s (2009) suggestions that the level of scrutiny moves inversely to performance. A god example is how strong beliefs in the “new economy” resulted in stock prices far beyond rational and fundamental pricing during the dot.com bubble.

Equally the housing bubble and credit crunch ahead of the 2007 recession was fueled by high risk taking in a goal to add to future yields in the growing market. The confidence gained from long periods of high performance seems to remove the respect of fundamentals and business cycle risks, and make managers overly risk-loving.

The bubble in the housing market is a good example which we have seen before, and which is often explained through the same reasoning; that because the population is growing and land is scarce, house prices always have to grow at a high rate. As already discussed, Akerlof and Shiller (2009) and Shiller (2005) show that this is fundamentally wrong², and that real house prices actually have not grown at such an impressive rate at a long term perspective, except during short term bubbles.

Linda M. H. Lai (1994) suggests that managers, especially those of crisis-prone organizations, often give little attention to potential signs of a crisis and ignores indications that a bubble is under development. Instead they seek information which confirms that their actions towards further investments and risk taking are correct while they look the other way to conflicting information.

This phenomenon is named “the confirmation trap”, and could indeed be part of an understanding to why some bubbles explained by the same rational are recurrent. A forecaster biased through such phenomenons is also likely to produce overly positive forecasts, and hence put even more fuel to the potential bubble.

In other words, managers seem to neglect the business cycle after long periods of growth.

This tendency is confirmed by the vast amount of articles and literature produced during booms suggesting that the business cycle might be dead. During the Great Moderation these stories were supported by the theories such as the existence of a Greenspan Put. This theory suggested that the US stock markets were safe from downturns of any significant magnitude because the Fed would be expected to increase liquidity at signs of weakness (Miller, Weller and Zhang, 2001).

This resulted in increased risk taking much based upon a theory which was based on the confirmation trap. Investors sought confirmation to justify increased risk taking, and to why markets would keep on growing. While the theory that the Fed would expand monetary policy in times of trouble is correct, both the dot.com bubble and the housing and credit crunch of 2007-2008 stands as proof that even the Fed cannot stop bubbles from bursting, and the business cycle from living.

The flexibility in the interpretation and choice of indicators means that these psychological phenomenons are of especially great importance to such forecasting analysis as the one presented in this paper. As managers forecast the future environment for prestige projects and investments during growth stages, they are likely to have short term incentives towards investing, and while also suffering from overconfidence, the forecast might be biased through the confirmation trap.

But with that said, the vast availability of indicators means that this forecasting approach has great opportunities to discover developments which are not supported by fundamentals. This should have made it harder for overconfident managers to justify their investments to the executive board.

Nevertheless, a forecaster biased by overconfidence and the confirmation trap is likely to produce biased predictions, and corporations should hence be careful with the choice of incentives put on forecasters to protect their objectivity. While macroeconomic forecasting does not come without uncertainty, it certainly creates qualified expectations about the future which can be of vital importance to the planning of corporate or investor decision making. Carried out with the right experience and with an unbiased and analytical mindset, it can indeed help managers prepare for changes in external macroeconomic factors which often gets the blame when things are going downward.

The possibility of a combined forecast

While both econometric and judgmental forecasts have different strengths and weaknesses, they are not mutually exclusive. As the econometric approaches are based on strong empirical relationships, but come with less flexibility, the judgmental approach introduced in this paper leaves much freedom and flexibility to the forecaster, but are sometimes an offer for human biases.

Many of the advanced computer programs such as SAS makes programming forecasts through econometric models such as the probit model a quick and easy job, and even though this model will not be examined in detail, Andrew J. Filardo (Dua 2004, pp 134-160) argues that this model have much empirical success in forecasting recessions, and could hence be valuable as a combined forecasting procedure together with the judgmental approach suggested in this paper.

Conclusion

This paper has argued that the ever recurrent business cycle is still very much alive and continues to possess a relevant risk to all market participants. But even after all these years with booms and busts, and with extensive research performed on business cycles, many market participants still seems unprepared as the economy reaches a period peak and enters a recession.

While the economy is moving in cycles, this paper has introduced a forecasting approach to help managers prepare for the future macroeconomic developments. It has produced relevant and updated research on a number of different economic indicators, and tested the flexible and dynamic forecasting approach on an ex post forecast of the business cycle peak from December 2007.

The forecasting approach suggested in this paper take both the importance of history and future flexibility into consideration. It learns from the past through an understanding of the physics of the business cycle and through empirical analysis of past developments in the economic indicators ahead of peaks and troughs.

From this the forecaster generates clear expectations on what to experience as the business cycle moves towards a change in growth. But while it learns from the past it still has the dynamics needed to notice the differences in the modern cycles from the cycles of the past. The new developments in the relevant economic indicators are also controlled against underlying fundamentals which gives a more focused picture on the validity of the current developments.

With this said, we must expect that future changes in the economic environment will affect the forecasting abilities of the different economic indicators. But the flexibility of this approach in regards of easily including or excluding indicators combined with the fundamental analysis where we get an understanding of the reasoning behind the current developments in the indicators, helps ensuring its relevancy also in the future.

False signals in single indicators will be detected through the fundamental analysis, and from the information gathered from other indicators. Also, if a single indicator for some reason should lose its relevance in the future, it can easily be switched with other indicators of bigger importance.

In the ex post forecast of the 2007 recession, the signs towards a future recession was unmistakable. Both business cycle theory and the developments in the respective indicators painted a clear picture of the potenţial dangers that lied ahead. As I have stressed throughout this paper; the fact that this forecast was made ex post, with the power of hindsight, arguably makes the interpretation of the developments easier, and somewhat biased. But still the signs from the different indicators were at such a magnitude and without fundamental support that a similar real-time analysis should have generated the same conclusion; that a recession was inevitable.

Another important acknowledgement from the ex post forecast is that the signs of the approaching recession were available at such an early point in time that it indeed could be of value to enterprise risk management. This means that the economic indicators, and the approach to economic forecasting analyzed in this paper was still of high relevance ahead of the business cycle peak of December 2007, which again increases the likelihood that this approach will carry relevancy also in the future.

As with all forecasting, this approach and the forecast of the 2007 recession, is far from perfect. It holds some important strengths and weaknesses, both technical and human, which the forecaster needs to be aware of. But this does not mean that economic forecasting is not a vital resource to macroeconomic risk management. We will probably never be able to predict the future with a hundred percent accuracy, but I have shown in this paper that we certainly can generate qualified expectations which can be used as the basis of internal awareness of future upside and downside risks resulting from changes in the growth of the business cycle.

1 There would probably be a need to use more indicators than the ones used in this basic approach, but it would not be a problem to gain this information.

2 They argue that house prices increased at a higher rate than can be explained by population growth, lack of land, increased costs of building or increased income.