Potential GDP

Introduction

This exercise will utilise the IMF World Economic Outlook database. We will use this to get some data that can be used to assess potential GDP. The main work will be to smooth or filter timeseries data and to assess the performance of models.

Circular flow

1. IMF World Economic Outlook

Go to the IMF World Economic Outlook database. These are updated twice a year (April and October). Choose the latest data and pick ‘By Countries’ for ‘country-level data’ and then select ‘Advanced Countries’ and the ‘United States’ and ‘continue’.

You will have national account, monetary, trade, people, government finance and balance of payments data. We need to select:

• Gross domestic product, constant prices (NATIONAL CURRENCY)
• Inflation, average consumer prices (PERCENTAGE CHANGE)
• Unemployment rate (PERCENTAGE OF TOTAL LABOR FORCE)

Make sure you are happy about what each of these mean. Choose ‘continue’ and then select years 1980 to 2022 and ‘PREPARE REPORT’ and then ‘DOWNLOAD REPORT’.

The data will be delivered as Tab-separated variables. This should be fine, but if there are any issues, download the file and open with excel. If a wizard opens, remember that the variables are separated by tabs in this case.

Save the data as an excel file in an appropriate place.

2. Prepare the data

The IMF present the data as rows of variables with each column a separate year. I find this really annoying and unusual. If you want to work like this, do that. I would prefer to have columns of variables and rows of years. To transform the data you need to copy all the data and then paste special and transpose. This will flip the data from rows to columns.

Tidy up the data. Give the columns clear, simple titles. Remove the row that says ‘Estimates’ Reduce the risk of making a mistake.

3. Smooth or filter the data to get the underlying trend

There are a number of ways that we will filter or smooth the data to try to find the underling trend. We will use these three:

• A linear model. This will assume that the underlying growth of the economy is a straight line.
  
• A quadratic model. This will allow the underlying growth of the economy to be curved.
• A moving average model. This will adjust to recent performance of the economy.

Each of these have their own strengths and weaknesses. Consider what these are.

4. Assess the performance of the models. How accurate are they?

We could use the standard way that models can be assessed: Root, Mean, Square Error (RMSE). This will look at the difference between the model and the actual GDP, square that difference (so that over-estimating and under-estimating can be added together) and takes the average and the root (to return to original values).

What are the advantages of using this RMSE metric? What alternatives can you consider?

5. Structural breaks

One issue that we may have with time series models is that fundamental changes may create structural breaks. Think about major events that could have caused affected the underlying performance of the US economy. Does it look as if something changed there? Can you re-estimate the model for different regimes?