As the Swiss parliament debates the Confederation's budget, we are regularly reminded that spending has risen faster than GDP since 1990. But this observation is based on the pattern of the early 1990s, and the situation has remained stable ever since. In this kind of analysis, one should be careful about the years taken as a starting point. It is more rigorous to estimate trends taking all data into account, which shows a situation under control.
Comparing two years can be tricky
Between 1990 and 2023, federal spending increased by 153%, while GDP and population grew by 115% and 31% respectively. These figures suggest that the trajectory of public spending is indeed unsustainable. The fact that they are rising faster than the population is not a problem in an economy where GDP per capita is increasing.
But the situation is highly sensitive to the starting point considered. The figure below shows the average annual growth in GDP (blue bars), federal expenditure (red bars) and population (green bars), depending on whether we start the analysis in 1990, 1995 or 2000. Starting in 1990 shows a problematic trend, but this is not the case if we take data since 1995, and even less so since 2000.
In other words, expenditure slippage is a problem that was solved during the 1990s, as shown in the figure below that displays of expenditure trends as a percentage of GDP. We can clearly see that, after rising during the first decade, the line has since fluctuated around a stable level.
Growth figures between two years should always be treated with caution. Indeed, the calculation of the bars above is based entirely on the values of the initial year and those of the last year, and ignores entirely the data for the other years. In other words, the calculation over the 1990-2024 horizon puts 32 out of 34 data points in the garbage can!
Take all the information into account
There are two possible approaches for taking all information into account. The first is simply to analyze the graph showing the evolution of the expenditure-to-GDP ratio as a whole. The good old critical eyeball Mk I remains a tried-and-tested method.
The second, more formal approach is to estimate a trend using statistical software. This allows us to see whether any trends are simply due to chance, or whether they are firmly rooted in the figures. The table below shows this analysis, depending on whether we start in 1990, 1995 or 2000. We can clearly see upward trends in GDP and spending, unsurprising in a growing economy. What's more, we can see that spending has a stronger growth trend than GDP, especially if we start in 1990 (growth gap of 0.26% per year between GDP and spending), with the trend being much weaker thereafter.
But most importantly, we can estimate whether the growth gap reflects the randomness of fluctuations in spending and GDP, or whether, on the contrary, it shows a real trend. The last two columns of the table correspond to this assessment.
As in any econometric analysis, the estimated value of the growth trend is accompanied by a “Student t” indicator, which is associated with the likelihood that the estimated trend is simply the result of chance, and not a true trend. We usually consider that the trend is really there if the likelihood is 5% or less, which corresponds to a Student t of just over 2. The table clearly shows that if we start in 1990, the trend is clearly there. On the other hand, shifting the starting point by 5 or 10 years shows that the estimated weak trend could very well in fact be zero.
A ghost to hunt
What can we conclude from this exercise? First of all, we should never simply calculate the changes between a starting year and an ending year, but rather take the entire evolution into account.
Second, while Switzerland had a problem with excessive spending growth in the 1990s, this has not been the case for a quarter of a century by now. The ghost of that difficult period should not haunt current discussions.