How to create a panel dataset

Panel data refers to the pooling of observations on a cross-section of entities. These same entities are observed at each time interval, which means that panel data essentially consists of “cross-sectional time-series” data. When dealing with economic or financial data, entities may be households, countries, firms, etc over several time periods. Time periods, depending on the level of detail you wish to observe, may be quarterly, semi-annually, or yearly. In some occasions, the time span may be daily; especially when dealing with financial data. For a strongly balanced panel dataset, the same entities should be observed across the time interval in a consistent manner.

Panel data is useful, as it allows you to control for variables that you cannot observe or measure. For instance, you can control for cultural factors across countries, or difference in business practices across companies. Besides, panel data allows you to control for factors that change over time but not across entities. For instance, corporate governance requirements, or business regulations may subject to change over time, however, the same set of requirements and regulations would effectively apply across all companies (entities). Hence, panel datasets are useful in capturing individual heterogeneity.

Panel data can be presented in many different ways, however, for econometric analysis it should always look as in the figure below. In this figure, we have a panel dataset of countries (entities) and years (time). It depicts three countries with four variables (y, x1, x2 and x3) over a period of five years (2014 to 2018). As you can notice, the same representation applies to all of the three countries, that is, we observe the same variables (y, x1, x2 and x3) over the same timespan (2014 to 2018). Depending on the complexity of your task, the easiest way to approach a panel dataset is to create it in Excel; and then transfer the data (as shown in the form of the figure below) into an econometric software like Stata.