Temporal dependence is a characteristic of many time series data that indicates that the past affects the future.
In other words, the value of a variable collected this year will depend largely on the value of that variable’s data last year. Temporary dependence is a characteristic to be taken into account in econometric studies . In some cases you can improve the estimates. Now, if analyzed incorrectly, it can lead to large estimation errors.
Of course, time dependence does not affect all time series data equally. There are four major types of time series . These four great types are summarized in the following table:
The type they normally affect is type 2 and 4. That is, when the average is increasing or decreasing (not constant). In other words, when the variable is in an uptrend or a downtrend. However, temporary dependence may also affect type 3. Very rarely, it will affect type 4.
The reasoning is simple. Type 4 is not altered by temporal dependence, because regardless of the time that passes, the series will only revolve around limits. What happened in the past does not serve to predict the future.
Examples of temporary dependence
Temporal dependence defines the fact that the data we collect currently depends on past evolution. For example, the current value of the Gross Domestic Product (GDP) depends, among other things, on the past value. If last year it had a value of 100 million, this year at most will range between 90 and 110. And we would be talking about a very large range of variation. In any case, the value will not exceed 100 to 10 million. The following graph shows the evolution of the GDP of a country for 40 years.
Another example also very simple, could be the shares of a publicly traded company. If yesterday each share had a value of 5 euros, the usual thing is that today its price is between 4.5 euros and 5.5 euros. And, we repeat, we are talking about a very wide range of variation. What would be very rare is that it will be worth 5 to 10 euros in one day or 5 euros to 1 euro. That is, it is very unlikely. The following graph shows the evolution of an action over 5 years.
As a conclusion, we can say that many time series data of economic variables have temporal dependence. Presenting temporary dependence is neither bad nor good, just different. So that difference must be treated with the relevant techniques.