As individuals, we acknowledge the importance of economic forecasting and how it affects the decision-making process of all relevant entities. From individuals to businesses to governments, you could see that all are paying really close attention whenever any economic forecast is released. Despite its significance, more often than not, economic forecasts are not successful in yielding indicative results of the future economy, as has been consistently proven in the last decades. Although various augments in model technicality have been made in order to improve the accuracy of forecasts, the predictive power of these models continues to be in question. In fact, econometric models have not yet yielded more accurate results in the past 40 years. If forecasts fail to deliver, should we continue to trust them, or are they even relevant to our daily lives?
How do economists construct a forecast model?
If we are about to attempt to answer this question, let’s take a look at how scientists and economists build up such a model. Usually, we would want to explore the relevant causal factors that affect the outcome that we want to investigate. These potential variables can be found in literature, or even in your own imagination. In econometrics, the process is formally known as creating a “base specification”, which includes all of the most basic necessary factors of the model. Then, an “alternative specification” will be made, in which you would add other possible explanatory factors into your model and then use statistical methods to evaluate whether they are relevant. After making appropriate adjustments, we check whether these combined factors fit well to what we intend to explain. Now, these steps seem relatively easy on paper, but it is much more difficult to create a model in real life.
So, suppose that you want to determine the market price of milk based on an economic model. After a thorough investigation, your “base specification” might include seasonality factors, the number of cows in the country, or the price of cheese, a (partial) substitute for milk. However, the equilibrium price of milk might be based on other factors that constitute the “alternative specification”, such as the futures market, or even the exchange rate!
Why is it so difficult to construct a good model?
The economy is a highly complicated system with various kinds of interactions between factors, which is why it is so difficult to construct a concrete econometric model. For example, the price of cheese is also dependent on the number of cows and vice versa, so these factors interplay with each other. Because of such similar interconnections, a subtle change in an economic phenomenon might result in substantial changes in other variables, which directly affects the accuracy of the model. Thankfully, the application of probabilities in models records these random changes, but one weakness is that these models cannot anticipate sudden large shocks.
Econometric models are undoubtedly helpful to professionals attempting to make predictions about the economy, but they still have some flaws. For example, we often use statistical tests in econometrics in order to evaluate the validity of the models. However, statistical inference is only a means to test the degree of correlation between the outcome and the input variables, and it does not need to conclude such causal relationship between them. After all, the judgment of whether to include a variable still depends on the knowledge and expertise of the economists.
Another explanation is that the market is largely driven by biases and because of that, confidence plays a big role in its formation and development. In most classical models, we presuppose that humans are rational decision-makers and that they are not likely to be influenced by external motives. Recent models do take into account the deviation from rationality to some extent, but the effects of behavioral patterns are not easily quantified. Take Bitcoin as an example. The increasing attention to digital currency has only been around for a couple of years, and only in 2017 has Bitcoin grown by 600 percent (at the time of writing this article). When some businesses and institutions started to accept Bitcoin as a means of transactions along with fiat currencies, investors also started to believe in the actuality of the cryptocurrency. When such consistent confidence is accumulated, demand is quickly fueled into investing the currency and its buying price keeps surging.
Economic models usually do not take into account exogenous variables. Usually, they suppose that some exogenous adjustments should be made when the model is run. These exogenous factors can include, for example, fiscal or monetary policy changes. These decisions on economic policies are determined based on the judgment of an independent board that is not involved in the market mechanism (although their decisions are the reaction to the happenings of the economy), so it is improbable to incorporate these changes into the model appropriately. For example, forecasts generally are forward-looking in terms of years. During that period, if the president decides to expand public expenditures or the central banker adjusts the interest rate, while the model does not take into account these changes, then it is unlikely that the model remains correct. In addition, the economy is continuously evolving as well with even more complex economic and financial products. The economy is so complicated that any model would be impossible to capture the entirety of an economy at a certain time. However complex the constructed model seems to be, it can be only partly representative of reality.
Should economists be able to see if a crisis is coming?
When there is a crisis, many economic forecasts tend to misjudge the development of the economy. For example, when the Fed was asked in 2007 to estimate how the US economy would perform during 2008-09, its estimates showed that there was no anticipation of a breakdown. There were some indications that the economy would moderately slow down due to the initial impact of the bankruptcy of financial firms, but the forecast showed that it would be merely temporary. Contrary to the general belief, the US economy began to shrink. Even with their further downward adjustments in 2008, the Fed governors did not believe that the economy was heading into a deep recession. As we have seen in reality though, the US economy collapsed: domestic banks bailed out and the stock market plunged. The US suffered a 2.8 percent contraction in growth in 2009 and is still enduring the consequences of the crisis.
In this scenario, the failure of economic models could be attributed to the fact that many economists erroneously undervalued the importance and interconnections between financial institutions, both public and private. Similar results can be found when we investigate the extent of the Eurozone crisis, as OECD was too optimistic regarding how European countries would be capable of reacting properly to the crisis. Similar to the Fed, they had to make constant downward adjustments on the performance of European economies.
How should we react to one forecast?
Well, although we could see how unreliable economic forecasts may be, it is better to have forecasts rather than nothing at all. However, there are some ways in which we can utilize the available forecasts to our advantage. One approach is to compile all of the available forecasts at hand and then take the average values from the obtained figures. The approach is sensible if you do not have any other information, and statistically speaking, the more forecasts you have, the lesser the extent of the error. However, when there is additional information that one believes to be true, one should be more critical of what forecasts to choose based on the included inputs.
The accuracy of economic models, for some reason, is rarely evaluated in a timely manner. In the United States, however, forecasts made by the Congressional Budget Office and their associates are occasionally evaluated by comparing its projections to the actual performance of the domestic economy. Their annual assessments can give some indication to what forecasts can be trusted if you are interested in the US economy. Similar evaluations are performed by the European Central Bank for European countries as well, but they are much less common.