INVESTIGATING THE DETERMINANTS OF BIG MAC INDEX: A PANEL DATA ANALYSIS

Transcription

1 INVESTIGATING THE DETERMINANTS OF BIG MAC INDEX: A PANEL DATA ANALYSIS Hrvoje JOŠIĆ Unversty of Zagreb, Faculty of Economcs & Busness, Kennedy sq 6, Zagreb, Croata hjosc@efzg.hr Zoran WITTINE Unversty of Zagreb, Faculty of Economcs & Busness, Kennedy sq 6, Zagreb, Croata zwttne@efzg.hr Antea BARIŠIĆ Unversty of Zagreb, Faculty of Economcs & Busness, Kennedy sq 6, Zagreb, Croata abarsc@net.efzg.hr Abstract Noted by several researchers, Law of One Prce (LOOP) and Purchasng Power Party (PPP) do not stand for most of the products n the short and long run. It s mportant to nvestgate what stands behnd t, besdes some specfc addtonal costs that could lead to dfferent results. Our goal s to check whether there are any general determnants that sgnfcantly affect the valuaton. The oldest nformal measure of PPP, Bg Mac Index, created by The Economst n 1986, wll be calculated to show absolute and relatve party valuaton of currences around the world. By usng cross-country panel data analyss on the sample of 25 countres n the perod from 2000 to 2015, man determnants of Bg Mac ndex are nvestgated. The results of the analyss have shown that devatons from the party can be explaned by current account varable whle other ndependent varables representng man determnants of Bg Mac ndex were not sgnfcant. Furthermore, results of the analyss ndcate that relatve PPP does not hold n the long run. Keywords: Bg Mac ndex, valuaton, panel data. JEL code: F31 Introducton Purchasng power party (PPP) s a theory whch states that natonal prce levels should be equal after converted to a common currency. It has orgns n a Law of one prce (LOOP). Karl Gustav Cassel (1918) was the orgnator of dea behnd the PPP theory and the most credted economst n ths feld at hs tme. Unlke PPP theory LOOP consders only one dentcal good. PPP comes n two varetes; absolute and relatve verson. Absolute purchasng power party takes nto account prces of goods n two countres whch should reflect exchange rates between countres. Relatve purchasng power party s a dynamc verson of absolute purchasng power party whch predcts relatonshp between exchange rate dfferental and nflaton rates between two countres n observed tme perod. Lght-hearted measures of Law of One Prce (LOOP) and Purchasng Power Party (PPP) such as Bg Mac Index, KFC Index, Starbucks ndex and lately Apple ndces are promoted to be a way to come accross the fact that consumer basket around the world not always ncludes the same products, as a result of bas toward dfferent goods across countres. The Economst's Bg Mac Index was frst publshed n It s nformal way of measurng the purchasng power party (PPP) between two currences. It compares Bg Mac prces between countres based on the purchasng power party theory usng Bg Mac (McDonald's burger) for comparson. Although Apple ndces as newely emergng nformal measures of PPP show several benefts (Jošć and Baršć, 2017), Bg Mac ndex has one very mportant advantage and we must not put t asde so easly. The product tself s one of the rarest examples of non-changng product over long perod of tme, present n wde range of countres across the globe, makng t possble to check the determnants leadng to ts devatons from the party and thus nvestgatng what could lead the changes. Also, when comparng aforementoned equal product prces not rarely the analyss gves opposng results, pontng to currency 69

2 undervaluaton and overvaluaton at the same tme, dependng on the chosen ndex. It s mportant to nvestgate what stands behnd t, besdes some specfc addtonal costs that could lead to dfferent results Our goal s to check whether there are any general determnants that sgnfcantly affect the valuaton. Usng cross-country panel data on the sample of 25 countres n the perod from 2000 to 2015 ths paper analyzes the effects of explanatory varables (gross domestc product per capta, current account balance, general government balance, labor compensaton representng average wages and Index of economc freedom) on Bg-Mac absolute party as dependent varable. In addton, relatve party for the Bg-Mac ndex s calculated whle panel unt root tests are used n order to nvestgate whether relatve party holds n the long run. Paper s organzed n fve chapters. After the ntroducton, the second chapter gves research overvew on Bg Mac ndex determnants. Methodology and data are explaned n the thrd chapter whle fourth chapter dsplays emprcal analyss, results and dscusson. Fnal chapter exhbts concludng remarks. Research overvew of Bg Mac ndex determnants In ths chapter man determnants of Bg Mac ndex are presented and explaned. Some of them can also be looked as lmtatons why PPP does not hold. Pakko and Pollard (1996) explaned devatons from PPP consderng three man explanatons: the exstence of barrers to trade, ncluson of non-traded elements n the cost of a Bg Mac producton and prcng to market. Types of barrers to trade are transportaton costs, trade restrctons and taxes. Transportaton costs drve a wedge between the prces of Bg Mac ngredents n dfferent places. More mportant factor are trade restrctons such as mposton of tarffs and quotas on trade of agrcultural products whch ncrease the prce of mports. Prce of Bg Mac hamburger s nclusve of sales or value added taxes so dfferences n tax systems across countres lead to dfferences n Bg Mac partes. However, prce of Bg Mac does not depend only on prce of ts ngredents but also on the cost of real estate and utltes needed n the process of Bg Mac producton. There s also a servce component or wages payed to workers for preparng the meal and servng the customers whch aslo belongs to non-tradng component. Although, accordng to Balassa (1964), non-traded goods affect the devatons from PPP due to dfferences n productvty across countres, there are unlkely large dfferences n productvty of workers n preparng Bg Macs n dfferent locatons. Government expendtures and current account mbalances can also explan why Bg Mac partes does not hold between countres n varous perods. Another reason for PPP not to hold s prcng to market or chargng dfferent prces on dfferent markets. Pakko and Pollard (2003) explaned that n the presence of mperfect competton prces of traded goods may dffer. Frms can prce to market by lmtng exchange-rate pass-through and the ablty to prce to market depends on safety and polluton standards, warranty restrctons and other factors that affect the possblty to resold goods across natonal borders. Accordng to Hadar (2011) man methodologcal lmtatons of the BMI are demand varablty, product comparablty, exchange rate predctablty, elements non-tradablty, transportaton costs, trade restrcton, taxes, productvty dfferentals, government expendtures and current account defcts and prcng to market. Alessandra and Kabosk (2011) argue that prcng to market s mportant as any local non-traded nputs, such as dstrbuton costs. In explanng the dfferences n tradable prces across countres they propose a model of consumer search that generates prcng-to-market. O'Bren and de Vargas (2015) and Clements and S (2016) hghlghted the mportance of takng GDP nto account when assessng currency valuaton. Ths s done by expressng Bg Mac prces as a functon of GDP per capta and adjustng for dfferences n ncomes of countres by constructng adjusted BMI. Adjusted BMI accounts for prces beng cheaper n poorer countres and vce versa. Aforementoned authors found that adjusted BMI outperforms the Raw (the Economst's) ndex n forecastng future currency values. Bat-Ulz (2016) used human development ndex, consdered as a good measurement of overall welfare of the economy, n evaluatng BMI. Human Development Index had a postve correlaton wth the Bg Mac Index ndcatng that the more the country s developed the more a Bg Mac burger costs at the same tme rejectng the BMI measurement of currency valuaton. Methodology and data Absolute purchasng power party valuaton represents devatons from purchasng power party. If devatons are postve, local currency s overvaluated aganst the USD, otherwse local currency s undervaluated. Absolute valuaton for Bg Mac ndex s calculated usng equaton 1: where 100 * ( P / P )(1/ E) 100 P bm s local prce of Bg-Mac hamburger n domestc currency, bm bm (1) * P bm s USD prce of Bg-Mac hamburger n Unted States whle E s nomnal exchange rate. Relatve purchasng power party s a dynamc 70

3 verson of the absolute purchasng power party. It s less strct than the absolute verson and states that the percentage change n consumer prces n both countres should mply the same change n the foregn exchange rate at the same tme. In equaton 2 s presented formula for calculaton of relatve PPP calculaton for BMI: * % P % P % E bm Another way of expressng relatve PPP s usng base year for whch the PPP holds (equaton 3). bm * * E E P P P P t E t P t P * (2) (3) Bg Mac affordablty s calculated usng equaton 4. It represents the number of Bg Mac hamburgers that can be daly bought wth country s GDP per capta. GDP per capta and prce of Bg Mac n country j have been expressed n USD. GDPj BM afford. / 365 (4) ( BM ) In order to nvestgate man determnants that affect the absolute valuaton of natonal currences aganst the dollar, panel regresson model s formulated ncludng 25 countres n the perod from 2000 to Explanatory varables n the analyss are gross domestc product, current account, labour compensaton, general government balance and ndex of economc freedom. Other varables such as productvty and barrers to trade (transportatons costs, trade restrcton and taxes) were not ncluded n the analyss. It s unlkely there are larger dfferences n productvty of workers n bakng hamburgers n McDonald's across the world. Furthermore, Bg Mac s a non-traded good so barrers to trade should not have effect the fnal prce of a product 1. Data for regresson are calculated or provded from varous Internet sources. Proposed cross-country panel regresson model s formulated as follows: VALUATION 0 1GDP 2CA 3LABCOMP 4GGB 5 prce j IEF (5) VALUATION - varable denotng the percentage of overvaluaton or undervaluaton of natonal currency relatve to USD. Postve values of varable VALUATION denotes percentage of overvaluaton whle negatve values denotes undervaluaton of natonal currency aganst the USD 2. GDP - Gross Domestc Product per capta (USD constant prces 2010 PPPs) of country. 3 CA - varable representng current account balance expressed as a percentage of GDP country. 4 LABCOMP - labour compensaton per capta country. 5 GGB - general government balance expressed as a percentage of GDP country. 6 IEF - Index of economc freedom country. 7 1 If analyss was conducted on Bg Mac ngredents used n the producton of the fnal product then ths varable would matter. 2 Data are avalable from the web page In the perod from 2012 to 2016 the data are publshed sem-annualy, for January and July, whch was chosen n the analyss. 3 Data are avalable from the web pages and 4 Data are avalable from the web pages and 5 Labour compensaton per capta s calculated usng varable labsh (share of labour compensaton n GDP at current natonal prces) from Feenstra, Robert C., Robert Inklaar and Marcel P. Tmmer (2015), "The Next Generaton of the Penn World Table" Amercan Economc Revew, 105(10), , avalable for download at multpled wth natonal GDP (n USD, 2010 constant prces). Ths varable was chosen nto analyss nstead of varable wages because full data seres for varable wages was not avalable and could not be constructed. 6 Data are avalable from the web pages and 7 Ths varable was chosen to descrbe mperfect competton or unmeasurable effects of prcng to market varable lke warranty restrcton, resale probablty, safety standards, busness regulators, polluton crtera, wholesale, etc. 71

4 GDP, CA and IEF. It s expected that hgher value of gross domestc product per capta s postvely Expected sgns of regresson are postve for varables varables GGB and LABCOMP and negatve for correlated wth Bg Mac prces and consequently overvaluaton of natonal currences aganst the USD. Smlarly, the dfferences between wages (labour compensaton) for workers n dfferent countres may explan why some currences are under/over valuated aganst the USD. Developed economes tend to gve hgher wages to ther workers so ther natonal currences should be overvaluated aganst the dollar as measured by Bg Mac prces. Researchers have generally faled to fnd a statstcal relatonshp between exchange rate movements and current account mbalances. It can go ether way, be postve or negatve so t s dffcult to conclude about drecton of ths effect (Pakko and Pollard, 1996). On the other sde, ncreased government spendng rse the overall prce level and consequently natonal currency wll be overvaluated relatve to ts PPP level. Hgher value of ndex of economc freedom should be assocated wth easer dong busness and therefore lower prces and undervaluaton of natonal currences aganst the USD. Cross-country panel regresson model wll be conducted dfferentatng between Pooled OLS (POLS), Fxed effects (FE) and Random effects (RE). In order to choose between fxed effects and random effects model Hausman test wll be used whle redundant fxed effects (Lkelhood rato) test wll be used to dfferentate between POLS and FE model. In addton, panel unt root tests wll be used to nvestgate f relatve purchasng power party holds n the long run. Results and dscusson In Table 1 (Appendx) s presented calculaton of absolute purchasng power party valuatons for 25 countres (ncludng Euroarea) for the perod from 2000 to It can be seen that the most of the natonal currences were undervaluated aganst the USD. In the year 2000 overvaluated natonal currences were Brtsh pound, Dansh krone, Japanese yen, South Korean won, Swedsh krona and Swss franc whle n 2016 only Swedsh krona and Swss franc were overvaluated aganst the USD. Relatve purchasng power party valuaton for selected countres n the perod from 2001 to 2016, 2000 chosen as the base year, s presented n Table 2 (Appendx). Calculaton s made usng equaton 3. Average country's Bg Mac absolute PPP valuaton s presented n Fgure 1. Fgure 1: Average country's Bg Mac absolute valuaton, Source: Authors' calculatons It can be notced average undervaluaton of natonal currences relatve to USD n most countres n the observed perod. Overvaluaton was present only n Brtan, Denmark, euro area, Sweden and Swtzerland. 72

5 Fgure 2: Average annual currences' absolute valuaton relatve to USD, 2000 to 2015 Source: Authors' calculatons Average annual currences' absolute valuaton relatve to USD from 2000 to 2015 s dsplayed n the Fgure 2. Throughout the observed perod there can be seen average currences' absolute undervaluaton aganst the USD wth maxmum value of % n Conversely, n the years 2008 and 2011 average absolute valuaton s mnmal and PPP nearly holds. Fgure 3: Bg Mac affordablty, 2000 and Bg Mac affordablty, Bg Mac affordablty, 2015 Australa Brazl Brtan Canada Chle Chna Czech Republc Denmark Euro area Hong Kong Hungary Indonesa Japan Malaysa Mexco New Zealand Poland Russa Sngapore South Afrca South Korea Sweden Swtzerland Tawan Thaland Source: Authors' calculatons Australa Brazl Brtan Canada Chle Chna Czech Republc Denmark Euro area Hong Kong Hungary Indonesa Japan Malaysa Mexco New Zealand Poland Russa Sngapore South Afrca South Korea Sweden Swtzerland Tawan Thaland Fgure 3 dsplays Bg Mac affordablty n years 2000 and In the year 2000 the most Bg Mac hamburgers could be daly bought n Sngapore (59), Hong Kong (56), Australa (46) and Canada (41) whle customers from Chna and Indonesa could purchase only 6 Bg Macs daly. In the year 2015 alongsde Sngapore (68) and Hong Kong (62), the most hamburgers s affordable daly n Tawan (50) whle Brazl s at the end of the ladder wth only 9 hamburgers to purchase daly. Table 3 exhbts descrptve statstcs of varables absolute PPP valuaton (expressed n percentages, %), gross domestc product (n USD), current account balance (n % of natonal GDP), general government balance (n % of natonal GDP), labour compensaton and Index of economc freedom (annual ndex n range of maxmum value 100 and mnmum value 0). 73

6 Table 3: Descrptve statstcs of varables, 25 countres, Varable Apsolute PPP valuaton,% Gross domestc product p.c. Current account balance, % General government balance, % Labuor compensaton Index of economc freedom Mean , , Medan , , Maxmum , , Mnmum , , Std. Dev , , Skewness Kurtoss Jarque-Bera Sum -6, ,779, ,971, , Sum Sq. Dev. 396, E , E+10 40, Observatons Source: Authors' calculatons Absolute PPP valuaton mean s ndcatng that average absolute PPP valuaton s negatve meanng undervaluaton of natonal currences aganst the USD (as seen on Fgure 2). Medan amounts at -21.7, maxmum value s (for Swtzerland n 2011) and mnmum value s (for South Afrca n 2002). Gross domestc product per capta mean s 26, and medan s 26, Maxmum value of GDP per capta was reported for Sngapore n 2015 (85,382.30) whle mnmum value s 2, (Chna n 2000). Current account mean s postve (2.33), as ts medan (1.45). Hghest current account sufct was reported n Sngapore n 2007 wth percent whle hghest current account defct was present n Hungary n 2004 wth percent of GDP. On the other sde, general government balance mean s negatve (-1.28) as well as medan (-1.30). Hghest general government sufct was present n Sngapore n 2007 wth 12 percent whle hghest defct were recorded n Brtan and Brazl wth percent n 2015 and 2009 respectvely. Labour compensaton mean and medan amount at 14, and 14, Maxmum value was present n Swtzerland n 2014 (40,090.68) and mnmum value was n Chna n 2000 (1,757.93), smlarly as t's GDP per capta values. Country wth hghest value of Index of economc freedom was Hong Kong (90.1 n 2014). On the other sde, lowest Index of economc freedom had Russa (48.7 n 2002). Results of the cross-country panel regresson analyss for 25 countres n the perod from 2000 to 2015 are presented n Table 4. Panel s balanced wth 400 observatons ncluded. Dependant varable s VALUATION whle ndependant varables are LAB_COMP, GDP, GGB, CA and IEF. In order to dfferentate between POLS, Fxed effects and Random effects models, Hausman test for correlated random effects and Lkelhood rato test for redundant fxed effects are used. Table 4: Cross-country panel regresson, 25 countres, Dep. var. VALUATION Independent varables POLS FE RE Constant ( ) ( ) ( ) LAB_COMP *** ( ) ( ) *** ( ) GDP *** ( ) ( ) *** ( ) GGB *** ( ) ( ) ( ) CA *** ( ) ** ( ) * ( ) IEF ( ) ( ) ( ) 74

7 Adjusted R-squared S.E. of regresson Prob. (F-statstc) Mean dependent varable S.D. dependent varable Akake nfo crteron Durbn -Watson stat Observatons Redundant fxed effects (Lkelhood rato) Correlated random effects (Hausman test) Cross-secton F ( ), Prob. (0.0000) Cross-secton Ch-square ( ), Prob. (0.0000) Ch-Sq. Statstc ( ), Prob. (0.0387) Source: Authors' calculatons OLS estmates show Whte heteroskedastcty- consstent standard errors and covarances; t- statstcs n parentheses; sgnfcant at the 1 percent level: ***. at the 5 percent level: **. at the 10 percent level: *. Hausman test s used n order to choose between fxed effects and random effects model. The null hypothess states that preferred model s random effects whle the alternatve hypothess states the opposte. The value of Ch-Square statstc for Hausman test s wth probablty of ndcatng that fxed effect model s preferable over random effects model. Lkelhood rato test's cross-secton Ch-square statstc s and Cross-secton F value s under 0.05 probablty ndcatng that fxed effects model s preferable over pooled OLS. Fxed effects model s not redundant suggestng that POLS hde the heterogenety among the cross-secton data. Statstcally sgnfcant ndependent varable n regresson s current account under 5 percent level of sgnfcance. Other ndependent varables are not sgnfcant n the model under 10% of sgnfcance. Fgure 4: Absolute valuaton vs Bg Mac determnants graph Source: Authors' calculatons Scatter dagram for absolute valuaton versus Bg Mac determnants (GDP, CA, GGB and LABCOMP) graph s presented n Fgure 4. Values of absolute valuaton are dspersed over the dagram so no clear concluson could 75

8 be made about relatonshp between absolute valuaton and ts determnants. Relatve purchasng power party wll be tested wth the help of panel unt root tests whch estmate the statonarty of the process. In Table 5 s gven summary of panel unt root tests. Null hypothess assumes common unt root process whle the alternatve hypothess states the opposte. If there s common unt root process, then relatve PPP holds. Table 5: Summary of panel unt root tests Seres: Relatve party Exogenous varables: Indvdual effects. ndvdual lnear trends Automatc selecton of maxmum lags Automatc lag length selecton based on SIC: 0 to 2 Newey-West automatc bandwdth selecton and Bartlett kernel Method Stat. Prob. Null: Unt root (assumes common unt root process) Crosssectons Levn, Ln & Chu t Bretung t-stat Null: Unt root (assumes ndvdual unt root process) Im, Pesaran and Shn W-stat ADF - Fsher Ch-square PP - Fsher Ch-square Probabltes for Fsher tests are computed usng an asymptotc Ch-square dstrbuton. All other tests assume asymptotc normalty. Source: Authors' calculatons Majorty of panel unt root tests suggested that seres of relatve party s statonary process although Levn, Ln and Chu test ponted out to the opposte concluson. Values of relatve purchasng power party does not bounce back to ther equlbrum levels so relatve purchasng power party does not hold n the long run. It can be concluded that there s no clear evdence n favour of acceptng the valdty of man determnants of Bg Mac ndex. Further nvestgaton should be made takng nto account possble endogenety n some of regressors and constructng dynamc panel data specfcaton model. Obs. Conclusons Goal of the paper was to check whether there are any general determnants that sgnfcantly affect the valuaton of Bg Mac ndex. Frstly, absolute and relatve purchasng power party valuatons for 25 countres (ncludng Euroarea) n the perod from 2000 to 2016 were calculated. Results of the analyss pont out to the concluson that the most of natonal currences were undervaluated aganst the USD n the observed perod. In order to nvestgate man determnants of the Bg Mac ndex, cross-country panel regresson model was used. Fve explanatory varables were taken nto consderaton: labour compensaton, gross domestc product per capta, current account balance, general government balance and Index of economc freedom. Usng Hausman test fxed effects model was chosen as approprate. Statstcally sgnfcant ndependent varable n regresson was current account balance under 5 percent level of sgnfcance. Other ndependent varables were not sgnfcant n the model under 10% of sgnfcance. It can be concluded there s no clear evdence n favour of acceptng the valdty of man determnants of Bg Mac ndex but further nvestgatons should be made. Furthermore, majorty of panel unt root tests suggest that seres of relatve PPP s statonary process although Levn, Ln and Chu test pont to the opposte. Values of relatve purchasng power party does not bounce back to ther equlbrum levels so t can be concluded that relatve purchasng power party does not hold n the long run. 76

9 References 1. Alessandra, G., Kabosk, J. P. (2011) "Prcng-to-Market and the Falure of Absolute PPP", Amercan Economc Journal: Macroeconomcs, 3(1): Balassa, B. (1964) The purchasng power party doctrne: A reapprasal, Journal of Poltcal Economcs, pp Bat-Ulz, T. (2016) Is the Bg Mac Index a relable ndex for currency evaluaton?. Artcle. Avalable at: [Accessed 20 November 2017]. 4. Cassel, G. (1918) Abnormal devatons n nternatonal exchanges, The Economc Journal, 28 (112), pp Clements, K., S, J. (2016) Smplfyng The Bg Mac Index, Journal of Internatonal Fnancal Management & Accountng, [onlne] 28(1), pp Feenstra, R. C., Inklaar R., Tmmer M. P. (2015) "The Next Generaton of the Penn World Table", Amercan Economc Revew, 105(10), [Accessed 20 November 2017] [Accessed 20 November 2017] [Accessed 20 November 2017] [Accessed 20 November 2017] [Accessed 20 November 2017]. 12. Hadar, J. (2011) Currency Valuaton and Purchasng Power Party, World Economcs, [onlne] 12(3), pp Jošć, H., Baršć, A. (2017) Apple ndces as a new nformal measure of Purchasng Power Party, Zbornk EZG, 15(1), O'Bren, T., de Vargas, S. (2016) The Adjusted Bg Mac Methodology: A Clarfcaton, Journal of Internatonal Fnancal Management & Accountng, [onlne] 28(1), pp Pakko, M. R., Pollard P.S. (2003) Burgernomcs: A Bg Mac gude to purchasng power party, The Federal Reserve Bank of St. Lous, pp. 16. Pakko, M.R., Pollard, P. S. (1996) For here or to go? Purchasng power party and the Bg Mac, Federal Reserve Bank of St. Lous Revew, (Jan), pp

10 Appendx Table 1: Absolute purchasng power party valuaton, Country Australa Brazl Brtan Canada Chle Chna Czech Republc Denmark Euro area Hong Kong Hungary Indonesa Japan Malaysa Mexco New Zealand Poland Russa Sngapore South Afrca South Korea Sweden Swtzerland Tawan Thaland Source: Authors' from the data avalable at: 78

11 Table 2: Relatve purchasng power party valuaton, , base year 2000 Country Australa Brazl Brtan Canada Chle Chna Czech Republc Denmark Euro area Hong Kong Hungary Indonesa Japan Malaysa Mexco New Zealand Poland Russa Sngapore South Afrca South Korea Sweden Swtzerland Tawan Thaland Source: Authors' calculatons 79