Exchange rate effects on cross-border commuting: evidence from the Swiss–Italian border

Abstract

1. Introduction

While considerable attention has been paid to different kinds of long-term migration, few studies have analysed the determinants of short-term travel. This study intends to contribute to the relevant literature by looking at the effects of exchange rate movements on a new form of mobility: cross-border commuting.

Cross-border commuting is a form of migration that is not accompanied by a change of home or home-country (Tassinopoulos and Werner, 1999), it is periodic (Kondoh, 1999), and it occurs within smaller areas.¹ The decision to commute compared with long-term migration is thus associated with a much lower cost, while still offering the benefits of migration.

Many studies in the migration literature have analysed factors affecting mobility patterns, such as wage differentials, unemployment rates, and social security benefits (Todaro, 1969; Dustmann, 2003; Heitmueller, 2005; De Giorgi and Pellizzari, 2009; Fiva, 2009). However, there are very few studies looking specifically at the effects of exchange rate volatility (Nekoei, 2013; Nguyen and Duncan, 2017) and none of them focuses on cross-border commuting.

In this study, I exploit variation in traffic flows across municipalities in Ticino, the southernmost canton of Switzerland, together with three addition datasets, the Cross-border Commuter Statistics, the Swiss Earnings Structure Survey (SESS), and Google trends data, in order to analyse the responsiveness of cross-border commuters to changes in the exchange rate that increase the purchasing power of host country wages in their home country. Specifically, I examine the 2005–2015 period, which was characterised by a significant appreciation in the Swiss franc. During the euro crisis, the CHF soared by almost 20% in value against the EUR, with the rise being halted in November 2011 with the introduction of a minimum exchange rate of CHF 1.20 per EUR by the Swiss National Bank (SNB). In January 2015, the CHF appreciated even more sharply when the SNB decided to decouple the franc. From 2005 to 2015, the CHF increased in value by roughly 30%.

The main findings of this paper are as follows. First, I show that the appreciation in the Swiss franc produces an increase in traffic flows, in Ticino, near the Swiss–Italian border. In particular, a 10% appreciation in the CHF is associated with an increase of 1.6–2.7% in the number of cars along the border. Interestingly, I observe that the vehicle flow across the border is affected by the exchange rate only during specific hours, which differ according to the direction of the flow: the flow towards Ticino is affected only in early or late morning, while the flow from Ticino to Italy is affected during late morning and the afternoon. These temporal and traffic direction patterns point to the role of cross-border shopping and commuting in explaining this effect. Second, I show that there is a positive relationship between the intensity of Google searches from Lombardy² for a job in Ticino and the value of the CHF against the EUR. Third, I find that the number of cross-border workers in Swiss municipalities close to the Swiss–Italian border is positively correlated with the EUR/CHF exchange rate. In particular, a 10% appreciation in the Swiss franc increases the employment of cross-border workers by 2.5% in municipalities within a driving distance of up to 10 km from the border. Lastly, I provide evidence suggesting that cross-border workers also react to the monetary shock through the intensive margin. Specifically, I show that the number of hours worked increases more for cross-border workers than for natives when the Swiss franc appreciates. The estimated elasticity for hourly employees is equal to 1.1. All results are robust to using the Swiss/Italian wage ratio as the main explanatory variable instead of the EUR/CHF exchange rate.

This paper is related to some recent studies that look at cross-border commuting as another form of spatial labour mobility, as distinct from long-term migration (Dustmann et al., 2017; Beerli et al., 2018). In particular, Beerli et al. (2018) focus on Switzerland and analyse the effect of opening the border to cross-border workers on labour market outcomes and firm performance. They exploit the timing of the implementation of the Agreement on the Free Movement of Persons (AFMP), between the EU and Switzerland, which occurred in discrete steps between 1999 and 2007. They find that, between 1999 and 2010, opening the border to cross-border workers produced an increase of foreign workers equal to 10 percentage points of the 1998 total employment in municipalities within 15 min travel time to the border. This increase is mostly attributable to the inflow of cross-border workers. Unlike Beerli et al. (2018), this paper looks at what happened between 2010 and 2015, after the monetary shock, when the effects of the agreement are expected to be vanished. While the effects of pull factors are not taken into account in the above-mentioned study, this paper focuses on the role of appreciation in the Swiss franc in explaining the rise of cross-border workers in Switzerland. Certainly, the liberalization of the Swiss labour market allowed cross-border workers to react fully to EUR/CHF exchange rate movements. The findings of this study are thus interesting and informative when compared with those in Beerli et al. (2018).

This paper is structured as follows. Section 2 presents the background and motivation. Section 3 describes the data. Section 4 presents the identification strategy and the results regarding traffic flows. Section 5 provides supporting evidence by using Google Trends data and official statistics on cross-border workers. Section 6 analyses the intensive labour supply response of cross-border workers. Finally, Sections 7 and 8, respectively, discuss the results and conclude.

2. Background and motivation

2.1. Exchange rate movements, migration and labour supply

An individual migrates if the expected real wage or income differential between the host and home country exceeds the cost of migration³ (Todaro, 1969; Harris and Todaro, 1970; Dustmann, 2003).⁴ Moreover, for immigrants who want to transfer money home, or plan to return to their home country at some point in the future, the exchange rate, by affecting the purchasing power of wages in their home country, represents another important determinant of their migration and labour supply decisions. Hanson and Spilimbergo (1999) focusing on the USA, in fact, show that the number of illegal Mexican immigrants (as estimated by the number of border crossing apprehensions) increases as the purchasing power of US wages in Mexican pesos increases.

While the effects of foreign currency appreciation on the extensive labour supply response of immigrants is more predictable, the sign of the intensive labour supply response is a priori ambiguous and depends on whether the substitution effect predominates the income effect or vice versa. For instance, Nekoei (2013), using data on Mexican immigrants to the USA, shows that US dollar appreciation, by decreasing the price of remittances, pushes immigrants to cut their annual hours worked. However, Nguyen and Duncan (2017) show that in Australia, a negative relationship between the exchange rate and the labour supply exists only for male immigrants, while female immigrants do not adjust their labour activity in response to exchange rate movements.

Cross-border workers differ from long-term immigrants as they work in a different country from the one where they live, crossing the border every day or at least once per week. Theoretically, they can spend most or all their income in their home country. Kondoh (1999), specifically, defines cross-border commuting as a case of migration with complete repatriation of income. In this extreme case, the utility function of cross-border workers is defined depending only on their host country wage, the nominal exchange rate between the host country and home country, the price level in the home country, and the cost of commuting. Therefore, commuting and labour supply decisions of cross-border workers are expected to be strongly affected by exchange rate movements. Nevertheless, to my knowledge, there are no studies that explicitly analyse the effects of exchange-rate fluctuations on the labour supply of cross-border workers.

2.2. The free movement policy and the labour market in Ticino

The extent to which the exchange rate can affect mobility across the Swiss–Italian border depends on the level of integration between the labour markets of the two countries.

Since 1999, when the Swiss electorate rejected the opportunity to join the European Economic Area, the economic and political relations between Switzerland and Italy have been pursued on a bilateral basis. The first set of agreements, the Bilateral Agreement I, was signed in 1999 and included, among others, the agreement of the Free Movement of Persons. This agreement made Swiss and Euro-area citizens free to choose their place of work and place of domicile inside the boundaries of the contracting parties. The agreement was gradually implemented, with three distinct phases identified: partial liberalization (1999–2004), full liberalization for cross-border workers (2005–2007), and full liberalization for cross-border workers and EU immigrants (post-2007). Before 1999, Swiss employers had to adhere to work permit quotas or give preference to Swiss nationals during the recruitment process. Moreover, workers with a cross-border work permit were allowed to work only in the Border Region (BR) of Switzerland, had to have resided in the adjacent BR of the neighbouring countries for the previous 6 months, and were obliged to commute on a daily basis.⁵ From 1999 to 2004, these constraints were gradually eased; foreign workers were not required to have resided in the neighbouring BR for the previous 6 months and could decide to commute weekly instead of daily. Moreover, in 2004, a first step towards labour market liberalization took place when cross-border workers gained full access to the labour market in the BR and firms were allowed to hire cross-border workers under the same conditions as resident workers. However, they still had no access to other regions of Switzerland until 2007, when the labour market was fully opened to both cross-border workers and resident immigrants.

According to SECO (2017), in Switzerland, a net immigration level of 65.500 people was registered during the years following the implementation of the agreement. Beerli et al. (2018) analyse, specifically, the response of the employment of immigrants to the labour market liberalization and shows that their labour supply increased by about 10 percentage points in municipalities within 0–15 min from the border and that two-thirds of this increase is attributed to the inflow of cross-border workers.

SECO (2017) points out how, in addition to the agreement, Switzerland’s generally favourable economic condition in those years explains the strong and positive migration flows. However, the relative proportion of each country of origin to total immigration has changed according to the economic evolution within the European Union (EU) zone. In fact, while during first years following the free movement of persons, the majority of immigrants came from north and north-western Europe (mainly from Germany) and to a lesser extent from southern Europe, during the financial and economic crisis the net migration of Germany began to decline gradually while immigration flows from Southern Europe increased sharply.

Moreover, SECO (2017) underlines how the appreciation of the Swiss currency resulting from the euro crisis, led a specific type of worker, the cross-border worker, to benefit from an increase in the domestic value of their wages by 5% on annual average (or 30% in cumulative terms) between 2009 and 2015. This exchange advantage was only partly offset by a higher annual inflation rate in the neighbouring countries, specifically, 1.4% in Italy, 1.2% in France and 1.1% in Germany.

In Ticino, in total, there are more than 60,000 cross-border workers coming from Italy. Table 1 shows the composition of commuters by country and the Italian province of provenience. More than 80% of cross-border workers consist of people coming from the two closest Italian provinces, Varese and Como.

3. Data

The main dataset consists of data on road traffic flows. The data source is the Sezione Mobilità del Canton Ticino, which provides data on traffic flows in Ticino using information obtained from 60 automatic traffic-counting stations. These traffic-counting stations located along the main roads of the canton record all traffic movement in terms of direction and time every day. I collected data from these stations for the years 2005–2015. The data consist of vehicle counts crossing a specific traffic-counting station at a certain time of the day and on a certain day of the year, for each direction of travel.

Moreover, I rely on three additional datasets. The first is the SESS, which is conducted every 2 years in enterprises in Switzerland.⁶ The survey includes information on the nationality, place of work (macro-region), education level, wage, hours worked and other individual characteristics of a random sample of workers. The other two consist, respectively, of search volume data derived from Google Trends and Cross-Border Commuter Statistics.⁷ The latter contains quarterly information on the number of foreign cross-border commuters working in Switzerland and their main characteristics.

Finally, I obtained daily data on the EUR-CHF exchange rate from the SNB and collected data on the annual average wage at current prices in Switzerland and in Italy from the Organisation for Economic Cooperation and Development (OECD) website.

4. Identification and the results on traffic flows

4.1. Identification

My identification strategy relies on geographical variation in the location of traffic-counting stations. To analyse the impact of the CHF appreciation on cross-border mobility, I identify traffic-counting stations in municipalities located within a driving distance up to 10 km from the border (CLOSE), and compare them to other stations, located further away from the border (FAR). Figure 1 shows the location of the traffic-counting stations. To ensure that long-distance goods-transport traffic flows are not included in the analysis, I disregard all traffic-counting stations located on highways. Stations not active over the entire period of analysis are similarly excluded. Thus, I obtain an unbalanced panel of 25 traffic-counting stations, with 11 located within 10 km from the border.

Figure 1

Location of traffic-counting stations.

Notes: The map shows the location of traffic-counting stations throughout the canton of Ticino. Traffic-counting stations located within 10 km of the border are in green on the map, while the remaining traffic-counting stations used in the analysis are in yellow. Traffic-counting stations on highways and those not active over the entire period of analysis are not included in the analysis.

Open in new tabDownload slide

The distribution of traffic counting stations based on their distance from the border is presented in Figure 2.

Figure 2

Distribution of traffic-counting stations.

Notes: Distribution of traffic-counting stations based on their distance from the border. For readability purposes, data for traffic counting stations located further than 30 km from the border are excluded from the graph. In total, there are four traffic-counting stations located further than 30 km. In four cases a single point represents two traffic-counting stations.

Open in new tabDownload slide

The assumption is that the traffic flows along the border are those mostly affected by cross-border travel, while changes in trade traffic flows due to appreciation in the Swiss franc pertain to both groups of traffic-counting stations. Thus, by exploiting the geographical variation in traffic flows, I am able to disentangle the impact of monetary shocks on mobility due to the increase in both cross-border commuting and cross-border shopping from other kinds of exchange rate effects on traffic. In addition, by looking at the time of day and traffic-flow direction, I can distinguish flows related to cross-border shopping from those related to cross-border labour supply.

As in Kaiser and Siegenthaler (2016), I claim that CHF exchange-rate fluctuations act as exogenous shocks on the Swiss economy, given the special status of this currency.⁸

4.2. Descriptive statistics

Figure 3 presents the evolution of the EUR-CHF exchange rate, from 2005 to 2015. During the financial turmoil of 2008, the CHF gained appreciably, since investors started depositing their money in CHF. In the midst of the Eurozone crisis, the value of the CHF soared—so much so that the SNB decided to provide support for the EUR by printing new CHF notes and using them to buy EUR. On 6 September 2011, the SNB introduced a minimum EUR-CHF exchange rate of CHF 1.20; the hope in doing so was to bring down the value of the CHF and help the country maintain its price competitiveness in the export market. Then, in January 2015, the SNB announced that it would no longer hold the CHF at a fixed exchange rate with the EUR and thus removed the floor. The value of EUR 1 fell to just CHF 0.85. This happened ahead of the move by the European Central Bank (ECB), among other things, to buy up huge amounts of state bonds. In so doing, the SNB managed to protect the Swiss economy from further volatility and upwards pressure on the CHF. Moreover, the figure shows that this appreciation translated to an increase in the value of Swiss wages in euro terms. The ratio between the value of Swiss wages (in euros) and Italian wages increased by the same size of the appreciation.⁹

Figure 3

The EUR-CHF exchange rate and the Swiss–Italian wage ratio.

Notes: The graph shows the annual EUR/CHF exchange rate (right axis) and the relative average annual wage of Switzerland to Italy in euros and current prices (left axis). On 6 September 2011, the SNB introduced the minimum exchange rate of CHF 1.20 per EUR. On January 2015, this floor was removed.

Source: SNB and OECD website (2005–2015).

Open in new tabDownload slide

Before turning to direct estimation of the effects, it would be interesting to look at the raw traffic data to appreciate the relevant differences in the traffic flow patterns between the two groups of stations, and to establish that, in terms of traffic flow, appreciation of the Swiss franc matters. Figures 4 and 5 display the hourly distribution of the average number of vehicles in the traffic counting stations close to the border (CLOSE) and in those further from the border (FAR). Figure 4(a, b) shows the two distributions during working days (Monday–Friday), while Figure 5(a, b) analyses traffic flows during nonworking days (Saturday and Sunday). Two driving directions are analysed (i.e. north-to-south, and south-to-north). The traffic flows in CLOSE municipalities on working days are clearly affected by cross-border workers who commute on a daily basis, and who enter Switzerland in the morning and return to Italy in the afternoon. For the other group, the two distributions are instead closely aligned, showing that traffic flows in areas further from the border are not related to cross-border workers. Differences between the two groups of traffic-counting stations are found also on nonworking days: while the two distributions overlap for the second group of stations, they remain staggered for those located close to the border.

Figure 4

Distribution of traffic flows by time and direction (Monday–Friday).

Open in new tabDownload slide

Figure 5

Distribution of traffic flows by time and direction (Saturday–Sunday).

Notes: The figures show the average number of cars per hour and per traffic-counting station by time of the day and direction across the two groups of traffic-counting stations. The symbol ‘CH->IT’ indicates north-to-south flows, and ‘IT->CH’ indicates south-to-north flows. The first group (CLOSE) consists of traffic-counting stations located within a driving distance of up to 10 km from the border, and the second group (FAR) includes the remaining traffic-counting stations.

Source: Sezione Mobilità del Canton Ticino (2005–2015).

Open in new tabDownload slide

Evidence of similar patterns is provided in Table 2, which reports, for each group of traffic-counting stations, the average number of vehicles per hour in five time intervals, which are 5–9 a.m., 10 a.m.–1 p.m., 2–8 p.m., 9–11 p.m. and 12–4 a.m.

Finally, Figures 6 and 7 show the correlation between the average number of vehicles per hour and the exchange rate, by showing the time patterns in trips for the two directions. A clear surge is seen in traffic flows after the currency began to appreciate in 2009, for the first group of stations (CLOSE) in both graphs; meanwhile, the flows seem to remain constant over time among the remaining traffic counting stations (FAR).

Figure 6

Annual and monthly variations in traffic flows. Direction: IT.

Open in new tabDownload slide

Figure 7

Annual and monthly variations in traffic flows. Direction: CH.

Notes: The figures show how the average number of cars per hour and per traffic-counting station changed over time across the two groups of traffic-counting stations. The symbol ‘CH->IT’ indicates north-to-south flows, and ‘IT->CH’ indicates south-to-north flows. The first group (CLOSE) includes traffic-counting stations located within a driving distance of up to 10 km from the border, and the second group (FAR) includes the remaining traffic-counting stations. The two straight lines represent the average number of cars in two periods: January 2005–September 2010 and October 2010–December 2015.

Source: Sezione Mobilità del Canton Ticino (2005–2015).

Open in new tabDownload slide

4.3. Empirical strategy

To analyse the effects of the exchange rate on cross-border mobility, the empirical strategy entails comparing the elasticity of crossings with regard to the exchange rate between the two groups: the group of stations located within 10 km from the border and the rest of them.¹⁰ I aggregate the daily traffic flow data on a monthly basis. I preserve the hourly structure of the dataset and distinguish working days from nonworking days. Moreover, I divide the daily hours into five time intervals and utilize the two driving directions separately.

Variable ln e_my represents the log of the EUR-CHF nominal exchange rate, whereas CLOSE_i is a dummy variable that has value of 1 only for those traffic-counting stations located within 10 km of the border and zero otherwise. ln e_my* CLOSE_i is the interaction term between the dummy and the exchange rate, and it measures the effect of interest. f_i are traffic-counting station-fixed effects, which account for differences across municipalities, and m_m are month-fixed effects, which control for strong seasonality in trips; X_my is a vector of control variables: the log of the Swiss gross domestic product (GDP), the log of the Italian GDP and the log of the unemployment rate in Lombardy.¹¹ Finally, Post2007_y is a dummy for the period after the full liberalization of the Swiss labour market, which was implemented in 2007.

I estimate this equation separately for working days and nonworking days, for each travel direction, and for the five time intervals, i.e. 5–9 a.m., 10 a.m.–1 p.m., 2–8 p.m., 9–11 p.m. and 12–4 a.m.

Moreover, in order to check the robustness of my results, I replicate my analysis using the log of the relative nominal wage expressed in euros (w_CH/w_IT) instead of the EUR-CHF nominal exchange rate. This time the observations are aggregated at the annual level.¹²

4.4. Results regarding traffic flows

The results are presented in Table 3. Traffic flow from Italy to Switzerland (namely, from south to north) is analysed in Panel 1, and that from Switzerland to Italy (from north to south) is analysed in Panel 2. For any temporal interval, there are two columns, one for working days and another for nonworking days.

The results in Columns 1 and 2 show that the traffic flows from Italy to Switzerland during the first time interval (5–9 a.m.) near the border increase more over time compared with those in the rest of Ticino. A 10% appreciation in the CHF seems to increase the number of cars along the Swiss–Italian border in early morning by 2.7% more than in the rest of the canton.

The results suggest that Italian workers responded to exchange-rate variations by increasing their labour supply in Switzerland.¹³ In line with this hypothesis, I find that the interaction term is negative and statistically significant also in Column 5 of Panel 2, which speaks to the traffic flows from Switzerland to Italy during 2–8 p.m., when, presumably, cross-border workers are returning to Italy. Conversely, the interaction terms in Columns 1 and 2 of Panel 2 and 5–6 of Panel 1 are never statistically significant; this suggests that, instead, the labour force commuting from Switzerland to Italy is not affected by the monetary shocks. Columns 3 and 4 of Panels 1 and 2 refer to the second time interval (10 a.m.–1 p.m.). The interaction terms remain negative and statistically significant in both panels. The estimated effect for this time interval is 0.16–0.18%. A cross-border shopping hypothesis can explain these results: the stronger the CHF, the larger the number of trips by Swiss consumers to the less-expensive Italian shops.¹⁴ Moreover, Table 4 presents the results for night-time traffic flows. It seems that border crossings made during the night do not adjust in response to exchange rate fluctuations. The coefficient of the interaction term is never statistically significant, with only one exception, in Column 2 of Panel 1, where the coefficient is positive.^15,16

Finally, the estimates obtained using the relative wage instead of the exchange rate are shown in Tables B.1 and B.2. The coefficients of the interaction terms appear even stronger. In particular, a 10% increase in the annual wage ratio between Switzerland and Italy leads to a 4% increase in the number of cars crossing the Swiss–Italian border in early morning (Column 1, Table B.1).

5. Robustness checks

5.1. Official statistics on cross-border workers

In the previous section, I showed that early morning traffic flows from Italy to Switzerland, and afternoon traffic flows from Switzerland to Italy, are affected by the exchange rate. These differences across hours and directions point to the role of cross-border labour supply in explaining the exchange rate effect on traffic flows. The appreciation of the CHF against the EUR induces more Italian workers to cross the border to work in Switzerland by increasing the value of Swiss wages relative to the EUR wages. To provide additional supporting evidence for this hypothesis, I use official statistics on commuters in Ticino, to show that the number of cross-border workers in municipalities close to the border is strongly affected by the value of the CHF. I use driving distance to determine the distance between each municipality and the closest border post and then distinguish those municipalities exposed to the shock according to alternative distance cut-offs (5, 10, 15 or 20 km). Table 5 shows the average number of cross-border workers across groups of municipalities.

Table 6 presents the results of the estimation. In line with the previous analysis, the nominal exchange rate seems to affect the number of cross-border workers in municipalities within a driving distance up to 10 km from the border. The supply of cross-border workers increases by 2.5% in response to a 10% appreciation in the CHF.

In Table B.3, I show the estimates obtained using the wage ratio instead of the exchange rate as the main explanatory variable. The magnitude of the coefficient of the variable of interest appears, again, even larger.¹⁷

5.2. Google Trends data

Does the increase in cross-border workers reflect changes in the labour supply, rather than in labour demand? To answer this question, I look for evidence in Google Trends search volume data. I test whether people in Lombardy increased their job hunting in Ticino when the CHF appreciated, compared with the previous period—in other words, in response to the economic incentives created by appreciation in the Swiss franc.¹⁸

Google has started to provide access to aggregated information on the volume of queries for different search terms, and on how these volumes change over time, via the publicly available service Google Trends. This facility shows how often a particular search-term is entered, relative to the total search volume for a given region and time.

The aim of this section is, therefore, to demonstrate that a link exists between the volume of job-related searches from Lombardy for positions in Ticino, and the EUR-CHF exchange rate. I collect data for three different queries: ‘Offerte lavoro in Ticino’ (‘Job Vacancies in Ticino’), ‘Offerte lavoro in Lombardia’ (‘Job Vacancies in Lombardy’) and ‘Cambio Euro Franco Svizzero’ (‘Euro Swiss Franc exchange rate’). I focus exclusively on data from Lombardy.

Before turning to the direct estimation, in Figure 8, I plot the volume of searches for the query ‘Offerte lavoro in Ticino’ (‘Job Vacancies in Ticino’) and for the query ‘Cambio Euro Franco Svizzero’ (‘Euro Swiss Franc exchange rate’) over time, along with the value of the EUR/CHF exchange rate. The figure clearly shows how the volumes of both queries increase in correspondence with the beginning of Swiss franc appreciation. Moreover, the volume of the query ‘Job vacancies in Ticino’ is quite high also in 2005, the year after the full liberalization of the Swiss labour market for cross-border workers. This represents an evidence of the explanatory power of the Google Trends data.

Table 7 reports the results. Columns 1 and 2 estimate the equation without the interaction term, so as to analyse solely the trend in the search volume of the two different queries, ‘Job Vacancies in Ticino’ and ‘Job Vacancies in Lombardy,’ respectively. As the coefficient on the term ln e_my in Column 1 is significant and negative, no statistically significant effect of the exchange rate emerges in Column 2. This result is robust to the use of the full model.

The coefficient on the interaction term in Column 3 is in fact negative and statistically significant. This suggests that the appreciation of the Swiss franc leads to an increase in the volume of searches made in Lombardy for a job in Ticino, while it does not affect the volume of searches for a job in Lombardy.

The evidence presented supports the hypothesis that Italian workers increase their labour supply in Ticino in response to a stronger CHF and confirms the role of cross-border commuters in explaining the surge in traffic flows at the border during CHF appreciation phases. I replicate the analysis using the wage ratio as the main explanatory variable, as in the previous sections. The results are presented in Table B.4 and appear quite similar.

6. Intensive labour supply response

So far, I have provided evidence for an extensive labour supply response of cross-border workers to changes in the EUR/CH exchange rate. In this section, I investigate whether they react also along the intensive margin, namely, by increasing their number of hours worked. To do this, I use data from the SESS. The dataset lacks a municipality identifier, so I use data on the entire Ticino canton for my analysis.²⁰ Descriptive statistics are shown in Table 8. I distinguish hourly employees from salary workers. For salaried workers, the survey provides the number of hours worked per week, while for hourly employees, the information refers to the total number of hours worked per month.²¹

In order to investigate the intensive labour supply response of cross-border workers, I compare the sensitivity of hours worked to changes in the exchange rate, between Swiss resident and cross-border workers, separately, for salaried workers and hourly employees.

The results, which are reported in Table 9, show that cross-border workers paid on an hourly basis respond strongly to the Swiss franc appreciation. Specifically, the average number of hours worked increases by 11% more for cross-border workers than for natives when the Swiss franc appreciates by 10%. The coefficient of the interaction term in the column regarding salaried workers is also significant but closer to 0. Table B.5 shows the results for the same analysis conducted using the wage ratio instead of the exchange rate. The results appear similar in sign and in magnitude.

7. Discussion

In the previous sections, I showed that cross-border workers increased their labour supply in response to exchange rate movements. However, since the introduction of the free movement of persons between Europe and Switzerland took place some years before appreciation of the Swiss franc, it may raise the concern that my results are driven by a gradual response to this policy change. In this section, I discuss why this is not likely to be the case.

The first step of the liberalisation for cross-border workers was enacted in 2002, when cross-border workers were no longer required to have resided in the BR for the previous 6 months and were allowed to commute weekly rather than daily. Then, in 2004, when the second step of the reform took place, employers in BR municipalities were allowed to hire cross-border workers under the same conditions as Swiss residents, though cross-border workers were still not allowed to work in municipalities not included in the BR (NBR) (Beerli et al., 2018). Finally, in 2007, the residence requirement and the distinction between the BR and NBR in Italy and Switzerland were completely removed, and cross-border workers as well as other migrants were allowed to work anywhere in Switzerland (Losa et al., 2012; SECO, 2017).

The first concern might be that my results are driven by the increase in the Italian population that is ‘eligible’ as cross-border workers following the removal of the residence requirement in 2007. I address this concern as follows.

First, in Table 1, I present the share of cross-border workers by Italian province of origin.²³ The Italian municipalities included in the BR (defined until 2007) were those located within 20 km of the border. The BR included almost all the municipalities in the province of Como, around 90% of the municipalities in the province of Varese, 80% of the municipalities of the province of Verbano-Cusio-Ossola and 30% of the municipalities of the province of Lecco.²⁴ The table shows that, even in 2015, the number of cross-border workers from provinces that were not included in the BR represented a small percentage of the total number of cross-border workers. One reason for such a low response might be the different tax treatment that has remained in place even after 2007 between the two groups of Italian provinces. In fact, while people from municipalities within 20 km from the border still benefit in Italy from a special tax treatment that allows them to pay taxes only in Switzerland, cross-border workers resident in other provinces are taxed in Italy.²⁵ Given the huge differences in tax rates between the two countries, it is likely that after 2007, people from the second group of provinces, who were willing to work in Switzerland, decided to apply for a residence permit rather than a cross-border commuter permit and to move their residence to the foreign country.

Second, in Table A.1, I report the growth rate of the number of cross-border workers coming from the province of Como (almost all of whose municipalities were included in the BR) and that regarding cross-border workers coming from ‘other Italian provinces’²⁶ (all of whose municipalities were not included in the BR). The table clearly shows how the number of workers coming from the province of Como grew at a higher rate after the beginning of the appreciation, that is, after 2010. The growth rate in the number of people coming from ‘other’ Italian provinces instead decreased after an initial jump.

Third, I claim that this policy change might even have contributed to reducing my estimation of the exchange rate effect on the labour supply of cross-border workers. In my empirical strategy, I compare municipalities within 10 km from the border to municipalities located farther away from the border. My results might be downward biased, as this additional level of liberalization allows cross-border workers to work everywhere in Switzerland.²⁷

The second concern might be that, since the 6 months requirement was no longer valid after 2002, people from other parts of Italy moved to Italian provinces at the border to work as cross-border workers. To deal with this point, I collected data on the number of inhabitants at the province level in Lombardy and Piedmont, the two Italian provinces bordering Ticino, and using a year-by-year analysis, I check that, since 2002, the number of residents in Italian provinces included in the BR has not increased in a manner different from that in other provinces. The results shown in Table A.2 confirm that the two groups experienced similar dynamics.

The last concern might be that I am capturing the effect of the full liberalisation of the Swiss labour market to cross-border workers implemented in 2004. Since the appreciation of the Swiss franc started in 2008–2010, it seems unlikely that I am confounding the two effects. However, to remove any concern related to the introduction of the bilateral agreement, by using the traffic flows data, I run a regression similar to my main specification in which I replace the exchange rate with year dummies. Figure A.1 shows the plot of the coefficients of year dummies interacting with the CLOSE dummy.²⁸ The figure shows that the coefficients regarding the period before 2009 are almost not statistically significant, with the exception of 2008. The coefficient becomes statistically significant and increases in magnitude in correspondence with appreciation of the Swiss franc. This graph appears similar to Figure 8, which shows how the interest of Lombardy people towards the EUR/Swiss franc exchange rate increases after 2009 and is correlated with the volume of job searches in Ticino.

Figure 8

Google Search Volume data.

Notes: The figure shows how the Google search volume for the queries ‘Job Vacancies in Ticino’ and ‘Euro Swiss franc Exchange rate’ in Lombardy changed over time along with the value of the EUR/CHF exchange rate.

Source: Google Trends (2005–2015).

Open in new tabDownload slide

It is reasonable to conclude that the implementation of the free movement of people allowed cross-border workers to react fully to the changes in the EUR/CHF exchange rate that were making Swiss wages more and more attractive to Italian workers.

8. Concluding remarks

According to Fries-Tersch et al. (2018), the number of cross-border workers in EFTA and EU countries was approximately 1.8 million in 2016. Given the magnitude of this new form of mobility, it is essential to understand its determinants in order to design optimal taxation, migration and labour policies.

In this study, I analysed the effects of exchange-rate fluctuations on cross-border mobility and provided the first estimate of cross-border commuter sensitivity to a currency appreciation. I focused on Switzerland, which is the main country of work for EU-28 cross-border workers (Fries-Tersch et al., 2018). Switzerland is a small country with relatively easy-to-cross borders, given the Schengen agreement, and has strong integration with neighbouring countries in terms of language and culture. Indeed, it represents the ideal setting for analysing the response of border crossings to exchange rate fluctuations. Moreover, since most of the other European countries have adopted the euro, and therefore do not see any kind of relative currency change, it is one of the few European countries in which a cross-border analysis of this sort is possible.

By exploiting changes in traffic flow along the Swiss–Italian border, I provide evidence for an extensive labour supply response of cross-border workers to Swiss franc appreciation. The results are confirmed by using search volume data from Google Trends and official statistics on cross-border commuters in Switzerland. When the CHF is stronger, more people resident in Lombardy look for jobs in Ticino, and the number of cross-border workers increases in municipalities within a driving distance of up to 10 km from the border. Moreover, I analysed the intensive labour supply response of cross-border workers to appreciation of the Swiss franc. My results suggest that the average number of hours worked by cross-border workers paid on an hourly basis increases when the purchasing power of their wages in their home country increases.

Figure A.1

Plot of the year-by-year regression coefficients.

Notes: Plot of the coefficients of interaction variables of the following equation:

$nihmy=α+∑n=20062015βnYearny+∑n=20062015γnYearnyCLOSEi+η1Xmy+fi+mm+εihmy,$

where $nihmy$ is the log of the average number of cars crossing the traffic-counting station i at hour h, month m and year y,$∑n=20062015Yearny$ represents year dummies for the years 2006–2015 and $CLOSEi$ is a dummy for traffic-counting stations located within 10 km of the border. $∑n=20062015YearnyClosei$ indicates the interaction terms between year dummies and the CLOSE dummy. $Xmy$ is the vector of control variables: the log of the Swiss GDP, the log of the Italian GDP, and the log of the Lombardy unemployment rate. Finally, $mm$ and $fi$ indicate month-fixed effects and traffic-counting station-fixed effects, respectively. The graph also shows the value of the EUR/CH exchange rate over time.

Source: Sezione Mobilità del Canton Ticino (2005–2015).

Open in new tabDownload slide