Predictions on Economy

Predictions on Economy

In economy, the problem of decision-making is always raised, that is, the choice of an option between different alternatives. When making a decision it is very important to have a vision of what is going to happen in the future: taking a decision requires considering all those alterations that may occur during the relevant time horizon for the subject in question. No decision should be made without considering the future evolution of all those events that condition it. In recent years a great emphasis has been placed on improving the decision-making process and this is where the idea of prediction comes in.

Origin

Predicting is a complex task, since it involves calculating some future event, in general, as a result of a rational analysis or a study of existing data. The aim of this document is to summarize the most important aspects that denote the most used prediction techniques in the area of economy and business.

In a broad sense we can say that the object of economic science is the study of the way in which economic agents make their decisions and the analysis of the consequences that result from the adoption of such decisions. Both in the economy of the company and in the macroeconomic field, the problem of decision-making is posed, that is, the choice of an option between different alternatives. Each option will result in a different result that can be measured in terms of utility, cost, benefit, or any other magnitude, depending on the problem being considered. However, the concrete result obtained will depend on situations that may occur outside the sphere of influence of the decision maker.

When making a decision it is very important to have a vision of what is going to happen in the future: taking a decision requires considering all those alterations that may occur during the relevant time horizon for the subject in question. In good logic, a decision should not be made without considering the future evolution of all those events that condition it. In recent years a great emphasis has been placed on improving the decision-making process and this is where the idea of prediction comes in.

When decisions are made, the decision maker is, in general, in an environment of uncertainty regarding the events that may occur in the future. The problem faced by the decision maker is to choose between alternative decisions, taking into account the usefulness of their decisions before each of the possible events. These events are facts, usually located in the future, or that the decision maker does not know. In any case, the decision maker will be able to achieve better results if to some extent he manages to reduce the uncertainty about future events, or that the decision maker does not know.

Prediction techniques are aimed, precisely, at reducing uncertainty about the future and, therefore, reducing the risk when making decisions. When predicting, it is a matter of calculating some future event, in general, as a result of a rational analysis or a study of existing data. For prediction to be useful in the planning process.

The field of application of the prediction is very broad in the area of the economy, and the following examples can be considered:

1. Predictions on macroeconomic variables such as GDP, inflation, current and investment spending, and others matters.

2. Predictions on the expected profitability in the international securities and merchandise markets.

3. Predictions on the behavior and economic results of companies.

There are different types of predictions depending on what is predicted and what approaches will require and different techniques will require, for example:

1. Predicting the effects of an event: In this case, we know that an event will occur in the future with certainty, and we want to determine what its effects will be. For example, to know who will win the next elections, or what effects will a law that will be promulgated soon or what will be the future sales of a new brand that comes to market, etc. The problem that arises is that the event can be unique, but with multiple effects or consequences, so the best possible approach is to search or generate the largest amount of relevant data.

2. Predicting the time an event occurs: This kind of predictions is questioned when, and if, a certain event is going to occur, that is, when the next elections will be or when the economy will recover, or when the competitors of a company will produce a new product at the time of the recovery of the economy market. In some of these examples, there was a sequence of similar events in the past, for example, the dates of the elections. In this case, observing the pattern of times between events, one could predict when the next one will occur. However, the usual way of working is to look for leading indicators, which are events that can happen before we are trying to predict. This approach is widely used to predict points of change in the evolution of the economy. For example, before launching a product on the market it can be seen that a company has reserved a lot of television advertising time.

3. Prediction of time series: A time series is a set of observations collected at regular intervals of time. For example, the hourly temperatures, the daily prices of shares at the close of the stock exchange, the monthly unemployment rate or the annual National Income, that is, if we are at time T and we want to predict what will happen to a variable at time T + 1. By definition this is the most complex type of prediction to perform, since it must involve the construction of a mathematical model that explains the behavior of the variable and is capable of predicting its quantification in the future.

Limitations on predictions

Most of the criticisms made to predictions (unexpected developments, predicted events that never happen, large prediction errors, errors in the moment, intensity of predicted changes, etc.) are well founded. However, it is necessary to understand or know how to interpret the predictions.

A necessary prerequisite to be able to predict, by whatever method, is that there is a pattern of behavior in the phenomenon we are studying. If a pattern of behavior does not exist, it is not possible to predict with a certain degree of precision, although, sometimes, subjective opinions can be given based on similar past situations.

In economics the predictability of a phenomenon varies from being almost nil (daily price of an action) to being excellent (Seasonal patterns based mainly on climatological reasons). The problem is that in economics, patterns and relationships are mixed with random components and can change unpredictably over time. Generally, changes in patterns or relationships are due to: i) Randomness of human behavior; and ii) Ability of people to influence the future with their own actions.

Factors of influence on predictability

When extracting the pattern of behavior followed in the past and being able to make conjectures about the future, a very important point is the amount of information that we have. For example, focusing on the time series, let us first consider a time series consisting of daily observations about the time of sunrise for fifty years. The problem would be to predict what time the sun will rise tomorrow. With a series of data of this type, it is very easy to make this prediction. It is true that, instead of using the observations of the past, prediction can also be made based on knowledge of the laws about the movement of the stars.

In any case and this is what we want to emphasize here, in a phenomenon of this type from the observation of the past, a good forecast can be made of what the future will be. Why does this happen? Simply because the series contains a lot of information being the past values of great utility to predict the future.

The choice of the prediction method to be used in a given situation entails finding a technique that satisfactorily answers the questions posed. The "best" prediction method is not always the most "accurate". The prediction method that should be used is one that covers our needs with the least cost and inconvenience. The professional should try to build simple models, easy to understand and, therefore, to explain. An elaborate model can generate more accurate predictions, but it can be very expensive and difficult to implement. The principle of parsimony tells us that when choosing between prediction models, if everything else is the same, we have to choose the simplest one.

Sometimes we just need very crude predictions, in other cases, accuracy is essential. In some applications the accuracy can be very expensive, for example; an imprecise prediction of an economic indicator can lead to the US Federal Reserve, to raise interest rates wrongly with all the consequences that this would have. On the other hand, increasing accuracy tends to increase the costs of both data acquisition, as well as of personnel or the use of data processing technology. If a small loss of precision is not very important, and the cost goes down substantially, we may prefer the simpler, less precise model, than the complexes that have been held over the last fifty years.

On the other hand, the greater the prediction horizon the greater the possibility of change in patterns or relationships because the behavior or attitudes of people can change. For example, fundamental changes in the environment may occur, for example, technological changes.

The short-term predictions are predictions for a period of less than three months, so there are two aspects to consider. On the one hand, changes in economic patterns and relationships can occur and actually occur. But due to the great inertia that most economic phenomena present, when a relationship changes, the result of this change is not immediate. This concept of inertia is very important in the economic field. For example, has taken almost a year since the so-called "oil crisis" for Western economies to enter recession. Due to this inertia and delays in the response, the current state of many variables is a good predictor of its value in the near future. That is, short-term established patterns can be extrapolated with a certain degree of precision.

Mid-term forecasts are those that cover the period from three months to two years and, in general, are derived from long-term predictions, or are constructed by accumulating short-term predictions. These predictions are usually not very accurate and, in general, it is often difficult to predict the points of change in business cycles, nor are recessions or periods of expansion. But these predictions are also necessary to make decisions about budgets or allocation of resources, therefore, planners must accept their limitations to predict recessions and booms in the economy and develop flexible plans that are capable of adjusting to cyclical changes.

The long-term predictions are those that cover a period of two years or more. The conclusions that are collected in the literature on the accuracy of these predictions are generally pessimistic; since in the long term there are many behaviors of variables directly or indirectly related to the fact that it tries to predict or simply generate new variables whose effect could not be estimated at the initial moment. In summary, long-term predictions tend to be imprecise, but they are necessary for strategic and budgetary planning. Therefore, all problems created by the uncertainty of these predictions should be studied and not ignored.

Guillermo Souto