Exchange Rates and Prices: Evidence from the 2015 Swiss Franc Appreciation∗ Raphael Auer†, Ariel Burstein‡, and Sarah M. Lein§ September 2020 Abstract We dissect the impact of a large and sudden exchange rate appreciation on Swiss borderimportprices, retailprices, andconsumerexpendituresondomesticandimported non-durable goods, following the removal of the EUR/CHF floor in January 2015. Cross-sectional variation in border price changes by currency of invoicing carries over to consumer prices and allocations, impacting retail prices of imports and competing domesticgoods,aswellasimportexpenditures. Weprovidemeasuresofthesensitivityof retailimportpricestoborderpricesandthesensitivityofimportsharestorelativeprices, which is higher when using retail prices than border prices. JEL classification: F31, F41, L11 Keywords: Large exchange rate shocks, exchange rate pass-through, invoicing currency, expenditure switching, price-setting, nominal and real rigidities ∗We thank Nielsen Switzerland, in particular Ina Banik and Florian Mailaender, and the Swiss Federal Statistical Office, in particular Robert Larocca, for providing and explaining the datasets used in this paper. We thank David Berger, Nathalie Chen, Javier Cravino, Ester Faia, Andrew Filardo, Gita Gopinath, Oleg Itskhoki,FrancescoLippi,OleksiyKryvtsov,AndreiLevchenko,DmitryMukhin,BrentNeiman,DiegoPerez, RobertoRigobon,PhilipSaur´e,GeorgStrasser,andJonathanVogelforveryusefulcommentsanddiscussions. Santiago E. Alvarez, Maximilian Bastian, Lucienne Brunner, and Emma Claggett provided excellent research assistance. The views expressed in this paper are those of the authors and not necessarily those of the Bank for International Settlements. †Bank for International Settlements and CEPR. e-mail: [email protected] ‡UCLA. e-mail: [email protected] §University of Basel, Faculty of Business and Economics, CEPR, and KOF ETHZ. e-mail: [email protected] 1 Introduction In this paper we study how prices and consumer expenditures respond to exchange rate movements based on the large and sudden appreciation of the Swiss franc (CHF) on January 15, 2015. Using data on prices and invoicing currency at the border, as well as Nielsen “homescan” data on retail prices and purchases by Swiss households, we present a range of facts that shed light on the sources of incomplete exchange rate pass-through and the role of nominal rigidities in price adjustment, the extent of expenditure switching by households, and the allocative implications of invoicing currency in international trade. We also provide estimates of the sensitivities of retail prices to border prices and import shares to relative prices, which are important elasticities in open economy models. The Swiss experience provides a unique setting to study the consequences of a large policy-driven change in the nominal exchange rate. On September 6, 2011, after a sharp appreciation of the Swiss franc, the Swiss National Bank (SNB) introduced a minimum exchange rate of 1.20 CHF per EUR. In late 2014 and early 2015, foreign developments such as anticipation of a large-scale quantitative easing program in the euro area raised the perceived cost of sustaining this policy (see e.g. SNB, 2015; Amador et al., 2020), prompting the SNB to unexpectedly abandon the minimum exchange rate on January 15, 2015.1 The subsequent appreciation episode is unique in a number of ways.2 First, it followed a period of remarkable exchange rate stability, with the EUR/CHF exchange rate fluctuating intherangeof1.2–1.22inthelastsixmonthsbeforeJanuary15,2015. Itishenceunlikelythat thepricedynamicsweexaminereflectadjustmentlagsduetopriorexchangeratemovements. Second, the exchange rate movement was large in magnitude relative to standard short-term exchange rate fluctuations in advanced economies, which have been a main focus of the literature.3 EUR/CHF appreciated by more than 20% on the day of the policy change, 14.0% by the end of March relative to January 14, 14.7% by the end of June, and 10.6% by the 1The SNB had reiterated its commitment to the minimum exchange rate throughout late 2014, arguing as late as December 1 that it “remains the key instrument for ensuring appropriate monetary conditions” (see Jordan, 2014). Of 22 economists surveyed between January 9 and 14, 2015, none expected the SNB to get rid of its minimum rate during the course of 2015 (see Bloomberg, 2015). Forward rates the day before the appreciation show that investors expected a flat profile of the exchange rate, as illustrated in Panel (a) in Figure1. Jermann(2017)arguesthatoptionpricesbeforeJanuary15revealedalowprobabilityofabandoning the exchange rate floor. 2Anumberofrelatedpapersalsoexaminethisepisode. Bonadioetal.(2019)documenttheresponseofunit values at the border, Efing et al. (2016) examine the effects on the valuations of publicly listed Swiss firms, and Kaufmann and Renkin (2017, 2019) study the price and employment response of Swiss manufacturing firms and the response of export prices. 3Therearemanypapersthatresorttolargedevaluationsindevelopingcountries;see,forexample,Burstein et al. (2005), Alessandria et al. (2010), Cravino and Levchenko (2017), and Gopinath and Neiman (2014). However, these episodes tend to be accompanied by other major macroeconomic developments that can confound the effects of exchange rate movements. Cavallo et al. (2015) use micro data on prices to show how a large monetary shock in a non-crisis context – Latvia’s euro area accession – impacts international relative prices. 1 Figure 1: The 2015 CHF appreciation 1.2 .05 1.15 0 1.1 −.05 1.05 −.1 EUR/CHF EUR/CHF IPI core 1 forward schedule CPI domestic Effective ER −.15 CPI imports 12/1/2014 1/1/2015 2/1/2015 3/1/2015 4/1/2015 2013m1 2014m1 2015m1 2016m1 date month (a) EUR/CHF and forward schedule (b) Official CPI and IPI Notes: Panel (a) shows daily nominal EUR/CHF exchange rates and effective CHF nominal exchange rates (Switzerland’s59maintradingpartners)betweenDecember1,2014andApril30,2015,andforwardexchange rates on January 14, 2015 (overnight 1 week, 1, 2, and 3 months). Panel (b) shows monthly EUR/CHF nominalexchangerate,coreimportpriceindex,andconsumerpriceindexforimportsandfordomesticgoods andservices,allrelativetoDecember2014. Sources: BankforInternationalSettlements(2016),SwissNational Bank (2016), Datastream (2015). end of December 2015 (see panel (a) in Figure 1).4 The CPI-based bilateral real exchange rate followed a similar path to the EUR/CHF nominal exchange rate, as shown in Figure A.1 in the online appendix (referred to as appendix from here on). The real appreciation was prolonged, with the EUR/CHF real exchange rate returning to its December 2014 level only by the end of 2017. Third, the appreciation occurred against the backdrop of a stable Swiss economy—TableA.1intheappendixshowsthatSwisseconomicaggregateswereremarkably stable in 2012-2016 — and reflected a policy response to foreign events.5 Following the 2015 CHF appreciation, there was a large decline in average import prices — more so at the border than at the consumer level — and a muted response in average prices of Swiss-produced goods (which we refer to interchangeably as domestic goods), as shown in panel (b) in Figure 1 using aggregate price indices from the SNB and the Swiss Federation Statistical Office (SFSO). To examine in more detail the response of these prices, as well as consumer expenditures, we combine several micro-level data sources, described in Section 2. Information on border prices and invoicing currency is from the good-level survey underlying the calculation of the official Swiss import price index. The transaction-level information on non-durableretailpricesandexpendituresisfromtheSwissNielsenhomescandata, whichwe 4TheSwissfrancappreciatedlessmarkedlyagainstothercurrenciessuchastheyenorthepoundsterling, as is evidenced by the effective exchange rate index shown in panel (a) in Figure 1. 5The price movements we focus on are unlikely to be the lagged result of safe-haven capital inflows while theminimumratewasinplace. Foreignsafe-havendemandforCHFwaslargelychannelledthroughbranches of foreign banks and invested in sight deposit accounts at the SNB (see e.g. Auer, 2015). Moreover, the CHF real exchange rate did not appreciate much in that period, and when it did in 2015, the growth rate of real GDP and real consumption fell slightly relative to 2014 (see Table A.1 in the appendix). 2 augment with data on the origin of the purchased goods. We exploit variation across product categories in currency of invoicing to trace the role of invoicing from border prices to retail prices and, further, to expenditure allocations. We start our analysis in Section 3 by documenting the response of border prices in the aftermathoftheappreciationandhowthisresponsevariesacrossgoodsbyinvoicingcurrency. The decline in border prices was much larger for EUR-invoiced goods than for CHF-invoiced goods,evenconditioningonnon-zeropricechanges,consistentwithfindingsinGopinathetal. (2010) for border prices in the United States. However, estimated differences in conditional price changes attenuate over time and become statistically insignificant about one year after the CHF appreciation. These patterns are qualitatively consistent with models of endogenous invoicing (e.g. Gopinath et al., 2010). We perform simple accounting exercises to quantify the impact on border prices of hypothetical changes in the currency of invoicing from CHF to EUR and changes in the degree of nominal price stickiness. We conclude from these exercises that over short horizons (during which border price stickiness in the currency of invoicing is quantitatively relevant), counterfactual shifts in the currency of invoicing have larger effects on border prices than do counterfactual shifts in the degree of nominal price stickiness. InSection4weexaminetheresponseofretailprices. Afterdocumentinginthehomescandata a decline in the retail price of imports relative to Swiss-produced goods, we provide evidence that variation across goods in invoicing currency at the border has a sizable impact on retail price changes faced by consumers. According to our estimates, in the first two quarters after the appreciation, retail import prices in product categories that are (hypothetically) fully invoiced in foreign currency fell by roughly 7 percentage points more than in product categories (hypothetically) fully invoiced in CHF. While previous evidence on the role of invoicing currency is based on import and export price changes at the border (see e.g. Gopinath et al., 2010; Fitzgerald and Haller, 2014; Gopinath, 2016), our results establish that differences in border price changes associated with the currency of invoicing carry over to consumer prices.6 We estimate the sensitivity of import prices at the retail level with respect to changes in border prices, leveraging heterogeneity in border price changes induced by variation in pre-appreciation EUR invoicing shares. These estimates imply that, after two quarters, a 1 percentage point larger reduction in import prices at the border resulted in a roughly 0.55 percentage point larger price reduction for imported products at the retail level.7 6The invoicing currency and response of border and consumer prices to exchange rate movements is an importantingredientofoptimalexchangeratepolicy(seee.g.Engel,2003;DevereuxandEngel,2007;Egorov and Mukhin, 2020). 7Bergeretal.(2012)usethemicropricedataunderlyingtheofficialUSimportandconsumerpriceindicesof theUSBureauofLaborStatisticstomatchindividualidenticalitemsattheborderandretaillevels,estimating the evolution of good-specific distribution shares. For related work studying pass-through at different layers of the distribution chain, see e.g. Nakamura and Zerom (2010) and Goldberg and Hellerstein (2013). 3 EventhoughtheresponseofretailpricesofSwiss-producedgoodswasonaverageverymuted, weshowthatpricesfellmoreinborderproductcategoriesinvoicedinEURrelativetothosein CHF,conditioningontheexpenditureshareofcompetingimportedgoodsinthesameproduct category. Relatedly, prices of domestically produced goods fell by more in product categories with larger declines in retail prices of imported goods conditioning on import shares. We argue that, under a certain exclusion restriction, these observations point to the presence of pricing complementarities that imply that domestic producers react to changes in prices of competing imported retail products.8 We further examine the response of the extensive margin of adjustment of retail prices. We show that the average decline in retail import prices in 2015 was partly accounted for by an increase in the fraction of nominal price changes, which can in turn be decomposed into a large increase in the frequency of price reductions and a smaller decline in the fraction of price increases. We provide aggregate time series evidence as well as cross-product evidence exploiting variations in invoicing currency and in the magnitude of changes of border prices. Specifically, theincreaseinthefrequencyofpricereductionswaslargerforimportedproducts with a larger share of EUR invoicing and with larger price reductions at the border. That is, differences in border price changes associated with the currency of invoicing carry over to consumer prices not only for average changes but also for the extensive margin of price adjustment.9 Finally, in Section 5 we examine the extent of consumer expenditure switching in response to the appreciation. On average during the year following the appreciation, expenditure shares of imported goods rose by roughly 4% (or by 1 percentage point, from 0.26 to 0.27). Import shares rose substantially even at short horizons after the appreciation. Leveraging cross-sectional variation along the invoicing dimension, we show that expenditure shares on imported goods increased by more in product categories in which imports are invoiced in EUR than in those categories invoiced in CHF. Hence, differences in invoicing currency at the border matter also for consumer allocations.10 To estimate the sensitivity of import expenditure shares with respect to changes in relative prices, we instrument import price 8TheseresultscomplementevidenceofstrategiccomplementaritiesinGopinathandItskhoki(2011),Auer and Schoenle (2016), and Amiti et al. (2019), using retail price data and in the context of a well-identified exchangerateshock. Relatedly,Cavalloetal.(2020)andFlaaenetal.(2020)showthatUSdomesticproducers increased retail prices in response to the recent increase in US tariffs on competing Chinese imports. 9For related work documenting the role of the extensive margin of price adjustment in response to large aggregateshocks,seee.g. Gagnon(2009)inthecontextofMexico’s1994devaluation,KaradiandReiff(2019) in the context of VAT changes in Hungary, and Gopinath et al. (2012) in the context of the trade collapse during the 2008 Great Recession. 10Differences in currency of invoicing at the border also carry over to allocations in the export side. In the context of the CHF appreciation, Auer et al. (2019) show that export growth in 2015 was larger in industries withhigherEURinvoicingofexportborderprices. Cravino(2017)usesdataonChileanexportstoestimatethe differentialresponseofexportstoexchangerateshocksaccordingtotheinvoicingcurrencyofthetransaction. Amitietal.(2018)studythedifferentialresponseofBelgianexportsacrossheterogeneousfirmswithinsectors. 4 changes across product categories using EUR invoicing shares at the border. Estimated price elasticities of import shares are close to 1 based on border-level measures of import prices, and much higher (ranging between 2 and 5) based on retail-level measures of import prices, but also less tightly estimated given large idiosyncratic movements in consumer prices. The large gap in estimated elasticities based on the measure of import prices is partly explained by lower exchange rate pass-through into retail prices compared with border prices.11 2 Data description Inthissectionweprovideanoverviewoftheborderandretaildatathatweuseinouranalysis. We provide additional details in Appendix B. In the replication package we provide contact information to obtain the proprietary data. 2.1 Import prices at the border We base the analysis of border prices on the microdata used by the SFSO to calculate the SwissImportPriceIndex(SwissFederalStatisticalOffice,2016). Thedataareasurvey-based panel of Swiss import prices similar to the US import price data studied in Gopinath and Rigobon (2008). The survey asks firms12 to quote the price and invoicing currency of the goods accounting for the firm’s highest volume of imports.13 Since most consumer goods are surveyed on a quarterly basis, we focus on this time horizon. Surveys are carried out by the SFSO in the first two weeks of each quarter. In the exposition, we refer to the last pre-appreciation quarterly observations (first two weeks of January 2015) as 14Q4, and to the first post-appreciation quarterly observations (first two weeks of April 2015) as 15Q1. Since we observe weights by product categories only starting in December 2015 (after a major resamplingofproducts), ourbaselineborderpriceregressionsareunweighted. Forregressions that use the subset of categories matched to the retail data, we weight according to consumer expenditures. 11Our estimates based on retail prices are on the high range of elasticity estimates in the literature based on time-series variation and using border prices to measure import prices (see e.g. Feenstra et al., 2018, and references therein). 12The SFSO data contain an importing firm identifier, which we use in sensitivity analysis of border price pass-through. However, since we do not observe firm characteristics of Swiss importers or foreign exporters, wedonotstudythefundamentalsthatdriveheterogeneousinvoicingpatternsasine.g. Devereuxetal.(2017) and Amiti et al. (2018). 13Foreachgoodinvoicedinforeigncurrency,wehaveinformationonthepriceexpressedinforeigncurrency andthepriceexpressedinCHF.Giventhatforsomeobservationstherearelargedisparitiesbetweenexchange rates implied by these two prices and official exchange rates (that are likely due to errors by contractors performing the conversion), we perform robustness exercises in which we convert foreign currency prices into CHF using official exchange rates. 5 Table 1: Border data summary statistics Number of observations % CHF-invoiced % EUR-invoiced % USD-invoiced 2013 14,666 68.5 28.7 2.4 2014 14,789 65.8 31.3 2.4 2015 17,381 56.1 38.1 4.7 2016 17,976 51.5 42.0 5.2 Notes: ThistableshowsthenumberofobservationsandtheshareofobservationsinvoicedinCHF,EUR,and USD for various years in the non-commodity border price sample. Table 1 displays, for the sample of non-commodity goods (commodities include agricultural products, coal, petroleum, metals, electricity and gas), the number of border price observations and the share of observations by currency of invoicing per year between 2013 and 2016. The share of observations invoiced in either CHF or EUR is close to 95% over the whole period, with CHF accounting for the highest share but falling over time.14 USD invoicing is quite limited (the US accounted for 7% of Swiss goods imports in 2015).15 In our baseline regressions, we exclude goods invoiced in foreign currency other than EUR because other currencies fluctuated vis-´a-vis CHF and EUR before January 15, 2015. The SFSO assigns imported goods to industries based on the industry of the importing firm using a classification similar to the 4-digit North American Industry Classification System (NAICS) code in the US. Our sample of non-commodity products covers 188 such product categories, of which 43 are consumer good categories that can be matched to retail categories asdescribedbelow. Forouranalysistracingcurrencyofinvoicingatthebordertoretailprices and expenditures, we calculate a pre-appreciation measure of invoicing intensity by border product category. We define the EUR invoicing share by product category as the fraction of border prices invoiced in EUR (relative to those invoiced in CHF or EUR) across all four quarters in 2014. In Table B.5 in the appendix we report the list of matched border product categories and retail product categories, as well as the EUR invoicing share of each category. Given our prior that EUR invoicing shares by category are less tightly inferred for categories with a low number of border price observations, we exclude from our baseline analysis 6 (out of 43) border product categories for which we observe 7 or fewer border prices per quarter 14The rise over time in the share of EUR-invoiced goods is largely due to entry of new goods into the sample that are invoiced in EUR. For continuing products, the fraction that switches invoicing currency between quarters is very low, on average roughly 0.5% per quarter in 2015 (see Figure B.1 in the appendix). 15As reported in Table B.1 in the appendix, invoicing shares are very similar if we weight border product categoriesusingNAICStwo-digitweightsinDecember2015,whichisthefirstperiodtheSFSOreportsweights. We note that Bonadio et al. (2019) and Federal Customs Administration (2015) report invoicing shares for imports based on more comprehensive customs data, allowing transactions to be weighted by import volume. In Federal Customs Administration (2015), import invoicing shares in 2014 are 31.6% for CHF, 54.9% for EUR, and 10.6% for USD. 6 on average in 2014.16 Across the baseline sample of 37 border categories, the EUR invoicing share in 2014 varies between 0 and 0.74, with a median of 0.13 and a mean of 0.25. 2.2 Retail prices and expenditures The analysis of retail prices and expenditures is based on Nielsen homescan data covering a demographically and regionally representative sample of around 3,000 households in Switzerland in the period January 2012 to June 2016 (Nielsen Switzerland, 2016). Participating households record purchases in supermarkets and drugstores, scanning goods such as food, non-food grocery items, health and beauty aids, and selected general merchandise. Individual products are classified into one of 256 product classes (which are narrower than border product categories) such as apple juice, shampoo, and toilet paper.17 In the raw data, an observation is a transaction including the household identifier, barcode (European Article Number, or EAN) of the product purchased, quantity purchased, price paid (net of good-specific discounts due to e.g. coupons), date of the shopping trip, and the name of the retailer. In the three months after the CHF appreciation, we observe on average 85 transactions per household. The data include 17 distinct retail stores. Since we do not observe the location of the retailer in a transaction, we assign it to one of 23 regions where the household lives (for more details, see appendix). We exclude purchases made in other countries via cross-border shopping. We augment these data with information on the country of production of individual goods. Whereas EAN codes provide information on the country in which a product has been registered, in many instances this is not the country in which the product has actually been produced. However, that information is disclosed in the label of each product. We collect labelinformationfromcodecheck.info,aSwisshealthinformationportalwithalargedatabase of products sold in supermarkets, drug stores, and pharmacies (Codecheck, 2016). Coverage is not complete and notably excludes goods that are only occasionally sold in grocery stores, such as toys, clothing, or household electronics. We drop observations for which we do not 16The6categorieswedropaccountforroughly12%ofretailexpendituresonimportedgoodsin2014. Inthe sensitivityanalysis,weconsideramorerestrictivesamplethatdrops9categorieswith8orfewerobservations perquarter,andalessrestrictivesamplethatdrops2categorieswith4orfewerobservationsperquarter. We alsodiscusswhichresultsarerobusttokeepingallbordercategories,includingthosewithonly2observations per quarter. 17In the Appendix we describe additional adjustments we make to the data, such as dropping newspapers, magazines, and non pre-packaged fresh fruits and vegetables products, and dropping transactions with errors in the entered price. 7 Table 2: Nielsen data summary statistics Summary Statistics Nielsen Samples Non-balanced Balanced yearly Balanced monthly No. of Imported Goods 4,545 2,682 937 No. of EU Imported Goods 4,134 2,362 794 No. of Domestic Goods 3,865 3,748 2,189 Expenditure share imports 2014 27 26 23 Expenditure share EU imports 2014 23 22 19 No. Product classes 233 217 172 No. Product classes (imports) 215 188 132 No. of Transactions - Imports 803,273 762,331 598,423 No. of Transactions - Domestic 2,396,208 2,390,273 2,106,375 Notes: The ‘non-balanced’ sample consists of EAN goods with information on country of origin (imports or domestic) that can be matched to border product categories with more than 7 border prices per quarter in 2014 (which we use in the baseline regressions). The ‘balanced yearly’ sample is a subsample of the first one that only includes goods observed each year between 2013 and 2015. The ‘balanced monthly’ sample is a subsample of the first one that only includes goods observed every month from mid-2013 to mid-2016. No. of Imported Goods and Expenditure share imports are, respectively, the number of imported goods and the expenditureshareofimportedgoodsintotalexpendituresin2014. WereportseparatelyimportsfromtheEU. No. product classes and No. product classes imports are the number of unique Nielsen product classes with positive expenditures on imports or domestic goods, and only on imports, respectively. No. of transactions – imports and No. of transactions – domestic are the number of underlying transactions at the household level over imports and domestic goods, respectively. observe the country of origin.18 Table2providesbasicsummarystatisticsoftheNielsendata, forthreedifferentsamples. The first sample (non-balanced) consists of goods with information on country of origin (imports or domestic) that can be matched to border product categories with more than seven border price observations per quarter in 2014. The second sample (balanced yearly) is a subsample of the first one that only includes goods observed each year between 2013 and 2015. The third sample (balanced monthly) is a subsample of the first one that only includes goods observed in each of the 18 months before and after the appreciation. We use the first and second samples in our analysis of expenditure allocations. We use the third sample in our analysis of retail prices. For each sample we provide the number of unique imported and domestic products, product classes, transactions, and import shares in 2014. The share of expenditures on imports relative to expenditures on all goods for which we observe country of 18Weaccessedcodecheck.infobetweenOctober2015andMarch2016,searchingforallgoodsintheNielsen data. We also cross-checked the results from codecheck.info with information on websites of the various retailers. To get a sense of coverage, there are 5,444 unique goods in the Nielsen dataset that are observed in eachofthe18monthsbeforeandaftertheappreciation. Wefound3,481ofthesegoodsontheweb,accounting for 72% of all expenditures in this balanced sample of goods in 2014. 8 origin is 27% in the non-balanced sample (and 23% in the monthly balanced sample).19 The import share is 23% in the non-balanced sample if we restrict the sample to goods imported from the European Union (EU). In our baseline results we include all imports because we do not observe the country of origin of imports in the border price data. 3 Exchange rate pass-through to border prices In this section we report the impact of the 2015 CHF appreciation on border prices, first at the level of individual goods and then at the level of product categories. We then document the extent of price flexibility and exchange rate pass-through by currency of invoicing, conditioning and not conditioning on nominal price changes. Finally, we perform simple accounting exercises to quantify the impact on border prices of counterfactual shifts in invoicing from CHF to EUR and changes in the degree of nominal price stickiness. 3.1 Changes in average border prices by currency of invoicing We first document the differential response of average changes in border prices by currency of invoicing after the CHF appreciation. We denote by pbor the log of the border price (in it CHF) of imported good i in quarter t. Keeping in mind our date convention, we refer to the period prior to the CHF appreciation as 14Q4. We consider panel regressions of the form (cid:88) pbor = β ×I ×EURinv +α +λ +ε , (1) it s s=t i t i it s(cid:54)=14Q4 over the period t = 13Q1,...,16Q2, where I is the time period indicator function, s=t EURinv = 1 (= 0) if product i is invoiced in EUR (CHF) in quarter 14Q4, α is a time i t fixed effect, and λ is a product fixed effect.20 Observations are equally weighted since we do i not observe import values per product. Standard errors are clustered at the level of border product categories. Panel(a)ofFigure2displaysestimatesofα andα +β between2013and2016, representing t t t average cumulative changes, relative to 14Q14, in CHF-invoiced and EUR-invoiced border prices, respectively. CHF- and EUR-invoiced goods display similar price dynamics before 19Forcomparisonpurposes,theshareofimportsintotalconsumptionreportedinSFSO(2014)is26.7%in 2014. Since services are mostly locally sourced, this means that the import share in our sample is lower than in overall consumption of goods. 20Weconsiderabalancedpanelofproductswithpricedataeveryquarterinthetwo-yearperiod13Q4-15Q3. Wedonotinclude15Q4inthebalancedpanelbecausetheSFSOconductedamajorre-samplingofproductsin December2015. Moreover,foreveryquarterweexcludeasmallnumberofobservationsforwhichthecurrency of invoicing differs from 14Q4. 9