Growth in the euro area, Sweden’s most vital export market, slowed considerably over the course of 2018. When presenting its new macroeconomic projection in December, however, the European Central Bank, or ECB, predicted that growth would recover. We have analysed a number of different forecast models and scenarios and can state that the ECB’s assessment, combined with several other forecasters, appears to be rather optimistic. In contrast, our results indicate that it is more likely that the euro area is in a recession that will persist over the next few years, leading to negative effects on Swedish export growth, among other areas.
Draghi has faith in recovery
Following the presentation of the ECB’s monetary policy decision in mid-December, its president, Mario Draghi, stated that growth in the euro area had proved to be weaker than expected. His conclusion was that this slowdown was in part temporary and in part an effect of the economy approaching its full growth potential. Based on strong domestic demand, improvements on the labour market and robust construction, the ECB projects that growth will pick up pace again, though at a slower rate than before – from 1.9% annually in 2018 to 1.7% in both 2019 and 2020.
The euro area represents the largest market for Swedish exports by far, and developments in that area are therefore not inconsequential for Sweden. Some 40% of all Swedish goods exports went to the euro area in 2017, as did around 30% of services, according to Statistics Sweden. The figure below shows that growth in the euro area and Swedish export growth are closely linked, although it can also be noted that Swedish exports to the euro area increased significantly in 2018 even as growth fell in the euro area overall – potentially signalling difficult times ahead for the Swedish export industry.
Labour market, interest rates and inflation
This article is an attempt to forecast growth in the euro area over the coming years in order to examine the validity of the ECB’s assessment. Assisting us in this endeavour is our new web-based tool, Indicio. We let Indicio run through time series of the past 20 years for some 50 different variables that could be deemed to say something about growth in the euro area up until the end of 2020. The result of this was that factors which best covary with, and contain information about, future growth in the euro area can principally be found in three areas: the labour market, interest rates and inflation.
Out of the various possible combinations of variables, we selected the set that gained the most points overall during the selection process and in which the highest number of models passed the thresholds to be deemed statistically defined. These variables include German and US 10-year government bond yields, unemployment and the euro area’s Consumer Price Index (CPI). Indicio’s selection process is called Var-Select and the model form is called VAR Lasso. The variables are given points according to their contribution to the accuracy of the forecast (Influence in Indicio, using a method known as Shapley Values). Indicio contains a library of some 30 different models that are evaluated by the system given the selected variables. Indicio also transforms the variables into differential and logarithmic forms and different lag structures while testing the various forecasting models. The aim is to find models with a strong ability to forecast (“out-of-sample accuracy”). Indicio identified 12 models that met these criteria. The GDP forecasts contained in these models are displayed in the image below. The spread between the forecasts is large from a short-term perspective, ranging from 0% to 3% growth, but they converge to between 1 and 2% by the end of 2020, with a few exceptions.
The multivariate models generate internally consistent scenarios for all input variables, so it is not necessary to produce separate forecasts in these cases. The internal scenarios for the 12 models are displayed in the figure below. Using own scenarios for the input variables can be justified at times, however – for example when there is a confident view of how one of them will develop or if you want to see how sensitive the forecast is in different scenarios.
ECB’s forecast the most optimistic
One option when attempting to pinpoint the “correct” forecast among all these is to combine them, depending on how accurate they have been historically in “out-of-sample” tests. There are several ways to do this in Indicio, but in this case we chose a variant that combines the models stepwise for each quarter, depending on their accuracy in terms of different forecast horizons (”stepwise MAPE” (mean-absolute-percentage-error)). The results of this combination are displayed in the image below, together with a similar combination of 12 univariate models and the ECB’s own forecast.
The ECB’s GDP forecast from December predicting growth between 1.5% and 2% over the coming years is clearly more optimistic than the combination of our model forecasts. The univariate models, which are based on different principles and interpretations of historical GDP data, are overall largely pessimistic and indicate growth between 0.5% and 1% at the end of 2020. The combination of the 12 multivariate models indicates somewhat stronger growth, but similarly points towards an imminent recession with GDP growth of around 1% at the end of 2020.
Is the ECB’s forecast too optimistic?
We use the ECB’s forecasts for inflation, unemployment and bond yields to produce scenario analyses for GDP growth in Indicio. The ECB predicts that inflation will be just below 2% (1.6% in 2019 and 1.7% in 2020), that unemployment will fall (from over 8% in 2018 to 7.8% in 2019 and 7.5% in 2020) and that bond yields will increase on average by 30 basis points over these years (we use German 10-year bond yields, while the ECB combines the euro area’s various 10-year yields).
The forecast results, combined with input from the ECB’s scenarios, are largely similar to the results produced by those generated by the original 12 models above. There is potentially a smaller spread, which is reasonable since they are based upon the same scenario for three out of four variables. If, however, we combine the GDP forecasts generated with input from the ECB against each other in the same way as the original 12 models, the result is roughly the same for GDP growth in the future. This is shown by both the light-green and light-blue lines in the figure below. The figure also displays the most optimistic and pessimistic outcomes of the original model forecasts (the broken lines), along with the ECB’s GDP forecast (red line).
In light of these exercises using different models and scenarios, the conclusion we can draw is that the ECB’s GDP forecast is higher than the various other growth trajectories. Interestingly, the Financial Times published a survey on new year’s eve conducted among 24 economists regarding the outlook for the euro area; their median forecast for GDP in 2019 is 1.6%, i.e. around the same as the ECB’s forecast. Only time will tell whether they are overly optimistic, but in light of the different models and scenarios we have used here, there appears to be an imminent risk that we will not see growth recover in 2019 and that the euro area will be stuck in a recession over the next few years.