2. Impact of anomalous weather conditions on economic sectors

2.1 How do we measure the economic impact of anomalous weather?

This part of the study investigated the relationship between climate and the economy through a time series analysis. We were particularly interested in whether anomalies in the weather, compared to normal weather conditions, affected the value added of certain economic sectors, as measured in the CBS national accounts. We used long-series macroeconomic data and weather measurements to estimate those effects. Using an econometric time series model, we reconstructed the structural connections between economic data and weather data. The model described the trends in the economic series, as well as seasonal patterns and possible special series interruptions. It was then enhanced further through the addition of weather variables. These were used to estimate the economic impact of weather anomalies; they also partly explained deviations in the economic trends. The chances of these weather anomalies occurring depend on the climate; as the climate changes, so do those chances. The number of frost days is an example of a weather variable. That number will decrease as a result of climate change, which may affect the output of certain sectors. We aimed to demonstrate that effect in this study by looking at weather and economic data from the past thirty years. By doing so, the study allowed us to show the influence of the (changing) climate in various economic sectors. 

An earlier study⁶⁾ demonstrated which of the tested weather variables has the most statistically significant effect on the following sectors: construction, manufacturing, food & accommodation services, mining & quarrying, and energy, as well as the size of the effect. This chapter’s results are therefore an updated account of prior research. We will discuss plans for refining and expanding said research in section 2.3. 

2.2 The effects of weather on the sectors construction, manufacturing, food & accommodation services, mining & quarrying, and energy

2.2.1 Weather variables

We used time series of the weather variables in Table 2.2.1.1.

2.2.2 Economic data

Our study used the macro-economic time series from Q1 1995 to Q2 2024. We applied the definition of GDP based on output, as shown in StatLine⁷⁾ , i.e. based on the value added for each economic sector. Specifically, we selected the following figures for each economic sector:

• value in millions of euros
• measured in constant prices (price level 2021=100)
• not corrected for structural, seasonal fluctuations. 

We investigated several sectors of the Standard Industrial Classification (SBI), namely B: Mining and quarrying; C: Manufacturing; D: Energy; F: Construction; and I: Food and accommodation services. We also looked at the sum total of these sectors.  

2.2.3 Updating the results

Tables 2.2.3.1 and 2.2.3.2 show the influence of particular weather anomalies in Q4 2023 and Q1 2024. We calculated the effects of the most significant variables for each sector (see Table 2.2.1.2). In the sectors construction and manufacturing, this effect is the result of several days of deep frost; in food and accommodation services, it is the effect of maximum temperatures; and in mining & quarrying, as well as energy, it is the effect of degree days.

2.3 Plans for refining and expanding the study

The figures shown in section 2.2.2 serve as an update to an earlier CBS study on this subject. We hope to refine these results in future research. We will also expand them in terms of the economic data that we wish to explain, the weather effects we wish to investigate, and the econometric time series models we wish to use for that purpose. We have listed the most important suggested improvements for each of these subjects below. 

Economic data:

• Our principal focus will be on the agriculture, transportation, and food & accommodation services sectors, as that is where we expect to find clear weather effects. However, we also hope to develop a more general approach that can be applied to other sectors. 
• If we look at the economy as a whole, we can also investigate which climate factor – such as drought or extreme precipitation – has the largest effect in which sector.
• In addition to value added per sector, it is also vital that we consider changes in volume and pricing separately. If volumes decrease due to extreme weather conditions (e.g. due to a failed harvest), this could lead to price rises even if turnover (price x volume) remains relatively stable. In other words, certain weather effects will show up in the data of one series, but not in the others.
• In the previous study, we looked at macro-economic quarterly figures. However, some weather effects can be difficult to detect at those frequencies or levels of aggregation because they cancel each other out in sub frequencies and sectors. This is why we plan to look at various levels of aggregation; smaller aggregates could be temporal (e.g. monthly series instead of quarterly) or intrasectional (e.g. several agricultural product sectors within the broader agriculture sector).

Weather data:

• We looked at a limited number of weather effects in the first study, mainly involving temperature (cold and heat). These could only partially explain the economic series, however. In the research that is currently ongoing, we will take a broader set of variables into account, based on multiple data sources, other defined weather variables, and combinations of weather effects. In doing so, we hope to make our research results more compatible with the pressure factors of the KNMI climate scenarios. 

Econometric time series models:

• We will look at a larger set of time series models. Finding the best model will be managed through the standard methods of model selection. 
• The time series models will take fluctuations into account, such as outliers, fractions, calendar effects, and the disturbances caused by the COVID-19 pandemic. This will ensure that those effects cannot be erroneously attributed to the weather. 
• Each sector may be affected by multiple weather variables, so we will look at models with multiple independent variables.  
• Because we hope look at a large number of weather effects, we will develop a method for selecting the most influential weather variable per sector. 
• The influence of weather effects can vary between quarters. For instance, a warmer winter does not affect the food and accommodation services sector in the same way as a warmer summer does. To account for this, the weather effects will be modelled quarterly.
• Time dependency: we will research whether the influence of a weather effect changes over the years. The construction sector used to be heavily affected by frost, for example, but technical developments may help them to cope with this better.
  
  

2.4 Technical explanation

Establishing weather variables

We used the daily data from the KNMI⁸⁾ to establish weather variables. Data was available for fifty weather stations across the Netherlands, but we only used measurements from the five homogenised stations: De Bilt, De Kooy (Den Helder), Eelde (Groningen), Vlissingen, and Beek (Maastricht). These measurements were then corrected slightly to make time series comparisons easier. Because the five stations are distributed evenly across the Netherlands, the average of their measurements can serve as a proxy for national weather conditions. The measurements were made up of 38 variables, which can be categorised as follows:

• temperature
• precipitation
• sun, clouds
• wind
• air pressure
• vision
• relative humidity 

1. Calculate the quarterly totals and averages of the variables for each station.
2. Calculate the long-term average over those same quarters for each station, and subtract this from step 1. Because we were looking for the influence of weather anomalies, we looked at anomalies in the long-term average. 
3. Average across all stations

Time series analysis

For our analysis, we used the methodology by Ouwehand and Van Ruth (2014)⁹⁾, including adaptations presented in Ouwehand (2020)¹⁰⁾. Please see those sources for a technical description of time series analyses.