Large Language Models (LLMs) offer the potential of personalised learning on any topic and today I will try this out by documenting my effort to explore the Gross Domestic Product economic indicator.

Background

Large Language Models (LLMs) are very useful in learning. A user can explore a topic by asking questions and follow on questions, rather as a child would with a tutor. Perhaps not quite, “Socratic learning“ (see below), but similar.

Top Economic Indicators

Let’s try this by my asking ChatGPT (4.0) about country specific economic indicators.

What are the top 3 most common economic indicators used to compare countries?

Nice, got GDP first as I expected, with both variants, total and per capita explained.

Unemployment and Inflation rates are also good choices for the top 3.

Calculation of GDP

Provide a short summary of how GDP is calculated.

Assuming there can be differences in the GDP measure calculated by each of the 3 methods, in practice how large are these differences?

A good summary of sources of discrepancies and a 1% bound for differences for well developed economies. Different wording of this prompt gave similar, but not exactly the same response, in some cases noting an up to 5% error for developing countries.

Comparison Charts

Now I like to see data in charts to help with understanding, so let’s do that.

Create a chart of GDP per capita for the most recent 5 years for the US, UK, Germany, France and Japan

Hmm, not a chart, but still the table looks good.

Convert this table into a line graph and also add at the bottom a link to the source or sources used

The United States powering ahead, France trending down. (Note the figures here are in nominal terms, so not real – meaning adjusted for inflation).

Source of Data

I am not comfortable about the source given below the chart. Clicking on the link takes me to the talkmarkets.com website, (not one I was aware of) and an article posted by Visual Capitalist, which has a table for these countries plus Canada and Italy for the years 2019 to 2029 (forecast) and states that it is IMF data.

So that is ok, but I wonder why ChatGPT used this web page?

Presumably because the countries listed and 5 year period matched to this data, but it does beg the question on source of results, so how do I know that Visual Capitalist has not made a mistake in copying the data.

Can you verify that this data is indeed the same as published by IMF?

Oh dear, but we have a link, so let’s use that to verify, as not a bad habit to check our sources as any good researcher would do.

Unfortunately the link in the above does not work but a Google search for “IMF GDP Figures” as the first result, the IMF World Economic Outlook website and clicking on this, I can see.

What a nice website with loads of official IMF data to look at and download.

Checking this I can see that there are differences, the numbers are close but not exactly the same, odd, I assume GDP figures are subject to small revisions and the article from Visual Capitalist published in June 2024, used older figures.

Anyway, now I am on this IMF World Economic Outlook website, it is a treasure trove of information and not much point in asking ChatGPT for other data, as it is all here; GDP, GDP per capita (current prices/nominal terms), Real GDP Growth and more.

I could spend a lot of time on this site, which I will at some point, but not a task for today’s blog on LLMs.

GDP trends

So let’s try something different.

You may know and certainly most of us in the UK know this, that real GDP growth has not recovered after the Great Financial Crises to its long term trend.

Let’s see if ChatGPT can provide an explanation.

Real GDP growth has not recovered after the Great Financial Crises to its long term trend in many countries, using the US and UK as examples, verify the accuracy of this statement and provide a plausible explanation

Nice.

This time I am more cognisant of the attribution of sources after each paragraph to IMF or the Institute for Fiscal Studies.

In-fact clicking on the Sources image at the bottom (not shown) opens up a panel on the right with a list of Citations and easy navigation to each.

Great.

What Else?

I did try:

What else should I know about GDP metrics?

But while the response was somewhat interesting, it is too verbose to re-produce here, so you will need to try yourself.

I will stop there.

My Learnings

I was able to quickly get details on GDP.

Both explanations and a summary of the calculation.

And then real world data for countries.

Checking the sources used, led me to the IMF Website.

Which was new to me and a great resource of official data.

I could ask many questions and get good responses with citations.

In many ways similar to Google Search but it felt more productive.

All in all a positive experience.

I particularly like how ChatGPT has tool support.

For example, it goes out to search the Internet as needed.

What’s Next

That’s all for today.

Next week, I will try a different angle, possibly also on GDP.

Stay tuned for that.