Intro

I’ve done this brief analysis to answer a forecasting question (illustrated below) posed by the Good Judgement - Open platform.

What will be the closing value of the euro to US dollar exchange rate on 14 October 2022?

Amid the war in Ukraine and concerns about energy supplies in Europe, the euro to US dollar exchange rate hit parity for the first time in two decades on 13 July 2022, meaning that one euro was worth one dollar (CNBC, AP). The outcome will be determined using data as reported by Google Finance (Google Finance, set to “6M”).

The following possibilities are made available:

Method

My forecasting process follows two simple steps:

  1. Examine historical evidence to create a baseline estimation.
  2. Examine other sources of evidence, specifically related to the war in Ukraine and its impact on the EU’s economy.

1. Baseline estimation

I start this forecast by loading a set of standard libraries and data, followed by a brief description of the dataset.

I obtained the data using the Google Finance function in Google Sheets, with the following arguments: =GOOGLEFINANCE("CURRENCY:EURUSD","price",DATE(1982,8,27),DATE(2022,7,21),"DAILY"). The dataset has two variables:

This dataset contains all the historical daily data available via the Google Finance function. I then downloaded it into a csv for this analysis (see graphical representation below).

head(df)
## # A tibble: 6 x 2
##   Date       Close
##   <date>     <dbl>
## 1 1982-08-27 0.965
## 2 1982-08-30 0.945
## 3 1982-08-31 0.944
## 4 1982-09-01 0.945
## 5 1982-09-02 0.956
## 6 1982-09-03 0.959

Modelling

For modelling the baseline, I use Facebook’s prophet package. I then make use of their quick start tutorial to guide my baseline forecast. The model returns a point estimate of 1.14 with a 80% uncertainty interval of [1.07, 1.21] (see table below).

##               ds     yhat yhat_lower yhat_upper
## 12771 2022-10-09 1.140304   1.036212   1.244404
## 12772 2022-10-10 1.138989   1.033644   1.244205
## 12773 2022-10-11 1.139182   1.032111   1.246236
## 12774 2022-10-12 1.139084   1.037338   1.252803
## 12775 2022-10-13 1.139440   1.041644   1.243183
## 12776 2022-10-14 1.139150   1.026650   1.231585

The plot below illustrates the actual prices and trend, as well as the forecasted values on the right end of the plot. Since the timeline is large (1982 to 2022), it is difficult to see see the difference between actual and forecasted values on the chart.

Plotting the model’s components illustrates the trend a bit more clear, but it also illustrates the seasonality components. From July to October, the ratio seems to increase by about 0.01 points. But this effect is small, for whatever reason it arises.

Baseline conclusion:

The point estimate is at 1.14, and given the wide values of the confidence interval, the last 4 possibilities (ranging from a ratio of 1.0 to >1.5) are still in the running.

Another question to ask about the model is whether historical data is useful in telling us anything about the future. The graphs above don’t seem to pick up the seemingly downward change in trend.

The war in Ukraine puts unprecedented pressure on EU’s economy. And, with the winter looming and gas shortages on the horizon - things might be quite different this time. Of course, there have been crisis in EU before, including Brexit and the 2007/2008 financial crisis. But none of them seem to have hit the euro as badly. Still, the question is how much impact will this have.

Altogether, the baseline is deemed insufficient to make a forecast. Methods quantifying the impact the current war will have need to be employed.

For now, I will submit a first prediction with an equal split between the last 5 possibilities presented.