We apply a simple linear regression model using the price of the euro in dollars to predict travel to Europe from America lagged by two, three, and four months. While negative coefficients for EUR/USD indicate that a cheaper euro may encourage travel to Europe, poor model-fit statistics mean that we can be no more than 15% confident that any of these models expresses a statistically significant relationship.

Key Points

• We analyze average monthly FX rates and lagged monthly travel rates to predict travel to Europe from the US. This lag reflects the exchange rate at the time tickets to Europe were likely purchased.

• We present simple linear regression models predicting travel to Europe two, three, and four months after the observed FX rate.

• None of the models are statistically significant, with R-squared values of <.001 and the lowest F-stat of the three being the four-month lagged model, at .853.

• The coefficients for all three models were negative, implying that, if there were to be a relationship, it would be directionally in line with our hypothesis - that a stronger dollar encourages travel to Europe.

As the euro continues its slow-motion-car-crash decline against the dollar, stateside Europhiles may be tempted by a discount vacation across the Atlantic. With the euro hovering around parity, roughly 25% off its historical 20-year average, the promise of lunchtime rosé and Mediterranean sunburns has rarely been so appealing. Here we answer the question, does a strong dollar relative to the euro result in more Americans travelling to Europe?

We start with data from the US Department of Commerce’s National Travel and Tourism Office. This gives us figures on outbound travel to eight regions: Europe, South and Central America, the Caribbean, Asia, Oceania, the Middle East, and Africa. We aggregate this data (published annually) for the past twenty years, ending with a dataset showing outbound travel from the US to each region since 2002, by month.

Next, we need currency data for comparison, and this we get from the Federal Reserve’s H.10 Foreign Exchange Rates release. To match time periods, we download a monthly average exchange rate for our currencies of interest, for each month since 2002. These are then converted into the direct quotation format – i.e., the price of the foreign currency in dollars. As the direct quote rises, the dollar is depreciating and vice versa.

Comparing monthly figures wouldn’t be the right approach, however, since people book their international travel several months in advance. So, we take the additional step of lagging our travel figures by one, two, three, and four months. This means that, if there is a relationship between currency strength and outbound travel, it would be reflected in trips made one, two, three, or four months after would-be travelers noticed their enhanced purchasing power. Since international ticket prices rise the most between four months and one month pre-flight, this is likely when traveler demand is strongest, and most bookings occur. Our hypothesis then, is that the strongest relationship would be between a month of observed dollar strength vs the euro, and a subsequent increase in travel by Americans to Europe three or four months later.

To test this hypothesis, we subject our data to a simple linear regression analysis. We would expect our model to show a negative relationship between the direct quote vs the euro (recall that a direct quote shows the price of the foreign currency in USD) and travel to Europe.

Below are the regression outputs for the two, three, and four months lagged data:

--------------------------------------------------------------------------------------------------------------------------

EuropeLag2 | Coefficient Std. err. t P>|t| [95% conf. interval]

-------------------+-----------------------------------------------------------------------------------------------------

EUR | -5272.753 211798.4 -0.02 0.980 -422493.9 411948.3

cons | 1001762 262854.3 3.81 0.000 483965.5 1519558

--------------------------------------------------------------------------------------------------------------------------

EuropeLag3 | Coefficient Std. err. t P>|t| [95% conf. interval]

-------------------+-----------------------------------------------------------------------------------------------------

EUR | -17039.31 212460.5 -0.08 0.936 -435573.6 401495

cons | 1016636 263765.6 3.85 0.000 497033.5 1536238

--------------------------------------------------------------------------------------------------------------------------

EuropeLag4 | Coefficient Std. err. t P>|t| [95% conf. interval]

-------------------+-----------------------------------------------------------------------------------------------------

EUR | -39469.34 213034.8 -0.19 0.853 -459144 380205.3

cons | 1045278 264571.7 3.95 0.000 524076.4 1566479

--------------------------------------------------------------------------------------------------------------------------

There are two notable elements of these outputs: first, the sigln of the coefficients does agree with our hypothesis: a negative coefficient for EUR here means that as EUR increases, i.e., euros get more expensive in dollar terms, travelers from the US to Europe decrease.

But as you can see from our confidence intervals and p-values, that relationship is weak at best. In fact, the p-values for each one of our regression outputs suggests that for all three of our models there is at least an 85% chance that there is no statistically significant relationship at all between the euro’s price and Americans’ travel to Europe.

Based on our models’ parameters, we can test how effectively they would have predicted outbound travel for the most recent month for which we have data, May 2022. During that month the average direct quote for euros was $1.0579. Based on that quote these are the predicted travel statistics for our three models:

Predicted Observed Miss %Miss

EuropeLag2 Model 996,184 1,663,664 667,480 40.1%

EuropeLag3 Model 998,610 1,663,664 665,054 39.9%

EuropeLag4 Model 1,003,523 1,663,664 660,141 39.7%

All three models missed by about 40%, with our model for a four-month lag performing slightly better – not surprising as this had stronger model fit statistics than the others. The size of these residual values – 40% of the observed figure - highlights the weakness of the euro/dollar FX rate in predicting travel patterns. While the negative coefficients confirm our hypothesis directionally, that a stronger dollar may encourage travel to Europe, we can be no more than 15% confident (based on the Europe lag 4 model’s p-value of .85, which was the strongest of the three) that any of these models expresses a statistically significant relationship.