Quality competition versus price competition goods: An empirical classification

Transcription

1 HEID Working Paper No7/2008 Quality competition versus price competition goos: An empirical classification Richar E. Balwin an Taashi Ito Grauate Institute of International an Development Stuies Abstract Base on the recent trae moels of the Heterogeneous irms Trae (HT) moel an the Quality Heterogeneous irms Trae (QHT) moel, we classify export goos (at the HS 6-igit level of isaggregation) by quality an price competition. We fin a high proportions of quality-competition goos for the major EU countries an lower proportions for Canaa, Australia an China. However, the overlap of these quality-competition goos is not large, which suggests that characteristics of export goos are substantially ifferent across countries at the same HS 6-igit coe. The Authors. All rights reserve. No part of this paper may be reprouce without the permission of the authors.

2 Quality competition versus price competition goos: An empirical classification Richar E. Balwin an Taashi Ito Grauate Institute of International an Development Stuies, Geneva Abstract: Base on the recent trae moels of the Heterogeneous irms Trae (HT) moel an the Quality Heterogeneous irms Trae (QHT) moel, we classify export goos (at the HS 6-igit level of isaggregation) by quality an price competition. We fin a high proportions of quality-competition goos for the major EU countries an lower proportions for Canaa, Australia an China. However, the overlap of these quality-competition goos is not large, which suggests that characteristics of export goos are substantially ifferent across countries at the same HS 6-igit coe. Key wors: Quality vs Price competition, heterogeneous firms trae moel. JEL Classification: INTRODUCTION Recent work on the theory an empirics of firm heterogeneity an trae provies new an wie ranging insights. In the mainstay moel in this new new trae theory the heterogeneous firms trae moel of Melitz (2003) competitiveness of a firm s prouct epens upon price; the cheapest goos are the most competitive. A minor twist on this moel (which was foreshaowe by a footnote in Melitz 2003) turns the stanar heterogeneous firms trae (HT) moel into the quality heterogeneous firms trae (QHT) moel where competitiveness epens upon the quality-ajuste price. If consumers care enough about quality, the highest price goos are the most competitive, so the association between observe price an competitiveness is reverse, i.e. firms with the lowest observe prices are the least competitive. 1 A simple empirical preiction separates the HT an QHT moels in trae ata. Since trae costs rise with istance of the market, the HT moel preicts that proucts with the lowest price get sol in the most istant markets while the opposite hols in the QHT, i.e. the highest price goos travel the furthest. These iametrically oppose implications provie the founation of a test of the moels by Balwin an Harrigan (2006), BH henceforth; that paper, however, pools across all categories of US exports thus implicitly assuming that all US exports are characterise either by a falling price-istant link (HT) or by a rising price-istant link (QHT). Our paper follows up on the BH by estimating the price-istance relationship separately for each prouct using panel ata. Our paper s main value-ae is to establish a list of three types of proucts. Those where competition appears to be base on price, those where it is base on quality, an those that cannot be confiently place in either category. Specifically we use export ata for nine large exporting nations at the HS 6-igit 2 level of isaggregation. Our key finings are: *11 Avenue e la Paix, 1202 Geneva, Switzerlan; Balwin@grauateinstitute.ch, Taashi.Ito@grauateinstitute.ch. 1 See Balwin an Harrigan (2006). 2 In the text below, the terms HS 6-igit an HS6 are use interchangeably. 1

3 1) Of the HS 6-igit coes that can be clearly classifie as quality or price competition, 50 to 60% of HS 6-igit coes exports of large European nations can be classifie as quality goos, while only 30 to 40% of US an Japanese exports fall into this category. We believe that the ifference may lie in pervasive trae in parts an components stemming from US an Japanese companies offshoring strategies that means nearby customers (the offshore factories) are a ifferent type of buyers than the far away customers (arm s length purchasers). 2) or commoity exporters like Canaa an Australia, the fraction of quality goos is much lower, only 15-25%. The share of quality goos in China s exports also fall in this range. We believe that this fits in with our priories that nations with a comparative avantage in raw materials shoul systematically see a lower incience of quality-type goos in their export mix. Literature One of the pioneering articles on the empirical front to use price as a proxy for quality is Schott (2004). It has ocumente a large ifference in prouct prices within the most isaggregate level of prouct classification. Schott (2008) shows that the US consumers pay less for Mae in China than for Mae in OECD for similar goos. ontagné, Gaulier, an Zignago (2008), analysing unit prices of HS 6-igit proucts of 200 countries, fins that the evelope countries proucts are not irectly competing with the eveloping countries proucts. Especially, because of their proucts superior quality, EU countries have less irect competition with the eveloping countries than Japan or the US oes. These finings have important policy relevance. It suggests that evelope countries can maintain their competitiveness by climbing up the quality laers within the existing inustries rather than moving to a new inustry. But which goos are quality goos? We, consumers, also know that some goos are not competing in qualities but in prices. Commoity goos an raw materials are mostly competing in prices. Thus, it is interesting an useful for policy purposes to sort out proucts by the egree of quality or price competition. Plan of paper Section 2 briefly reviews the theory that structures our empirical exercise. Section 3 explains the ata, estimation equation an results. The final section conclues. 2. THEORETICAL RAMEWORK To structure our empirical analysis, we briefly summarize the price competition an quality competition versions of the heterogeneous firms trae moel, highlighting two simple empirically testable preictions of these moels. The classic HT moel (Melitz 2003) can be thought of as the Dixit-Stiglitz monopolistic competition trae moel where firms have ranomly rawn marginal cost functions an face iceberg trae costs as well as a fixe cost of establishing a beachhea in each market. As usual, the Dixit-Stiglitz structure links the value of sales irectly to operating profits, so the beachhea cost means that only sufficiently competitive firms export. Moreover, istance-linke iceberg trae costs imply that a firm s competitiveness is iminishe in more istant markets, so export status isplays a istance-marginal cost graient. The threshol egree of competitiveness necessary to sell in markets rises with the market s istance, so average competitiveness of firms servicing a particular market rises with istance. The HT an QHT moels iffer only in their eterminants of competitiveness. In HT, price is the sole basis of competition, i.e. market entry threshols can be written in terms of a maximum price. In QHT, competitiveness epens upon quality-ajuste price, so market-entry threshols are efine in terms of quality-ajuste price. In the stanar version of the moel, the lower quality-ajuste prices (unobserve) are associate with higher unajuste (observe) prices. In other wors, firms only export the most expensive goos to the most istant markets. 2

4 The BH test of this preiction foun that the QHT provie a better explanation of the US export ata for That fining, however, pools across all US exports, thus implicitly assuming that all US exports are marke either by quality competition or by price competition. In this paper, we explore the hypothesis that some goos may fit HT preictions while others fit the QHT preictions. The hope is that we can ientify a set of proucts which are for a number of major exporting nations characterise by either HT or QHT. 3. DATA, EMPIRICAL MODEL AND RESULTS Taking the istance-price preiction to the ata requires some care in hanling the theoretical preictions. Suppose we ha the true prouct-level ata, i.e. ata on the prices an sales by particular firms of particular proucts in particular estination markets. If the baseline moel is true, Dixit-Stiglitz mark-up pricing woul imply that the proucer price, i.e. OB price, for each variety woul be insensitive to istance. The point is that the istance-price-graient preiction stems from prouct/firm selection, not from firms pricing behaviour. However, if we work with publicly available trae ata, where the finest isaggregation available for all nations is HS 6-igit for all nations (HS 9-igit or HS 10-igit for some nations) then istance will have a selection effect. We shoul fin that the average export price rises (QHT) or falls (HT) ue to an unobserve shift in the bilateral composition of varieties/firms. or example HS6 category comprises Other electric conuctors for a voltage less than or equal to 80 volts. Insie this category will be some high price varieties an some low price varieties (they coexist in both moels ue to prouct ifferentiation). The empirical lever comes from the way the nation-specific price mix changes with istance of the market. If QHT is correct, Germany s HS export basket to rance will have more low price varieties than its export basket to the US. Empirically, this will show up as a lower unit value inex for Germany s HS exports to rance. If the HT moel is correct, we shoul observe the opposite. We conuct the analysis at the HS 6-igit level because this is the most isaggregate internationally harmonise classification coe. Thus if our classification is to prove useful for a broa range of nations, it must be at the HS 6-igit level. Below, we report some sensitivity analysis at the HS 9-igit level for Japan Empirical moel an results The empirical moel we employ is akin to BH. The main ifferences lie in our use of log-linear istance (instea of istance bans) an our panel imension. BH uses a panel for US exports of ifferent proucts to ifferent estinations but for a single year. Our panel ata looks at a given origin nation s exports of a given HS6 prouct to all the estination countries over a ten years perio ( ). or example, as is shown below, 4,845 HS 6-igit lines have at least 20 observations for the whole perio. We run 4,845 regressions for the US. We o the same for the other 8 exporters. In total, we estimate approximately 40,000 regressions. The regression equation is: ~ p = β + β log DIST + β log GDP + β log GDPCAP + β D + ε t, 0 1 ( ) 2 ( t, ) 3 ( t, ) 4 t, where p t, is the log of the OB unit value inex to estination country at time t, DIST is the bilateral istance from the exporter uner stuy an estination country, an GDP t, is the estination-country GDP at time t ; GDPCAP t, is the corresponing GDP per capita. D is a vector of year ummies. ε t, is an ii error. 3

5 3.2. Data The export ata we use is for 9 exporters (the worl s top 8 exporters plus Australia) for 1997 to These are taken from the UN COMTRADE atabase. The istance ata are from CEPII; the GDP an population ata are from the Worl Development Inicators of the Worl Bank. or the nine exporting nations we work with, Table 1 shows the worl export rank an the number of HS 6-igit lines with at least 20 observations. Table 1: Global export rank an numbers exporte HS6 lines US Germany Japan China rance UK Italy Canaa Australia Rank Lines 4,845 4,550 4,150 4,557 4,674 4,751 4,664 3,465 3,942 Source: UN COMTRADE, Author's calculation Results Table 2 summarises the results. 3 The first column shows the number of HS6 proucts for which the istance coefficient is statistically significant at least at the 5% level. The secon an thir columns show the breakown between the lines with positive an negative coefficients on istance, respectively. The first three lines show the results for the US, Germany an Japan. Consier the US numbers. Out of the US s 4,845 HS 6-igit proucts, 1,957 proucts have statistically significant istance coefficients with 1,667 of these being positive. This suggests that about 34% of US HS6 exports are quality goos, about 6% are price goos. The remainer, almost 60%, cannot be classifie, perhaps because the statistical groupings bunle together some price an some quality proucts. The results for Japan are broaly consistent with those for the US, with 38% of the HS6 lines isplaying positive istance coefficients (suggesting quality competition) an 9% isplaying negative coefficients. inings for the four large EU exporters, Germany, rance, Britain an Italy, reveal a much higher share of quality goos. About 50% of German, rench an British exports are classifie as quality goos, while the figure for Italy is almost 60%. We also note that the share of unclassifiable lines is much lower for these nations, with the share range from 33% for Italy to 48% for Britain, while rance an Germany both close to 40%. An informal test is given by contrasting these results for highly inustrial nations with nations who are known to be more epenent on commoity exports where we expect price to matter more than quality. Australia an Canaa are the nations we chose as they are both large exporters, heavily reliant on primary goo exports an have excellent ata. or these two nations the share of quality goos is only 15% (Canaa) an 27% (Australia). The share of price goos is 7% for Canaa, an 5% for Australia. China presents an interesting case. Its exports are ominate by inustrial goos, but it is wiely perceive to be an export of varieties where low prices are the key to their success. What we fin is that about 21% of its exports are quality-goos by our measure an 17% of its exports are price goos. 3 See the separate file for the whole list of quality an price competition goos by country. 4

6 Table 2: Numbers of quality-competition an price-competition goos Number HS6 lines where istance is significant at 5% Number with positive coefficient (quality competition) Number with negative coefficient (price competition) USA 1,957 1, Germany 2,531 2, Japan 1,947 1, Overlap (J, D, US) rance 2,749 2, UK 2,464 2, Italy 3,136 2, Canaa Australia 1,255 1, China 1, Source: UN COMTRADE, Authors' calculation Overall among US, German an Japanese quality goos The next question is whether these goos classifie at HS 6-igit level are common to the countries in stuy. The fourth row of Table 2 shows the number of HS 6-igit lines that overlap among the US, Japan an Germany, i.e. the number of proucts which show statistically significant coefficient estimates of the same sign for the istance variable for all the three countries. Out of 3,905 HS 6-igit prouct lines which were common for all the three countries, 510 show statistically significant coefficients for all the three countries. Out of the 510 prouct lines, 322 show statistically significant positive coefficient estimates for all the three countries. We woul like to suggest that these HS coes coul be consiere as marke by quality goos. The 322 figure represents just 8% of the 3,905 HS 6-igit prouct lines exporte (about 10% in terms of export value). Only one HS 6-igit coe has statistically significant negative coefficient estimate for all the three countries. There are 157 proucts that show two positive signs an one negative sign, so maybe we coul call these quasi-quality competition goos. 30 proucts show one positive sign an two negative signs. The proportion of quality an quasi-quality competition goos is much higher than price an quasi-price competition goos. The rather small proportion of the overlap of prouct coes across the three countries implies that the prouct mixture in terms of quality/price competition for a particular HS 6-igit coe is ifferent across countries More etaile isaggregation act that less than one out of ten lines can be unanimously classifie as quality or price goos suggests that there may be a great eal of heterogeneity (across exporters) among the basket of goos inclue in each HS6 category. 5

7 To investigate the hypothesis, we compare the analyses at HS6 results with estimates on HS 9-igit ata for Japan. Specifically we reprouce the above proceure for all the HS9 lines within given HS6 coes. Since the purpose here is a simple check, not an exhaustive analysis, we have conucte an analysis using a ranom sample from HS coes. Japan has approximately 5,000 lines at the HS6 level with roughly 9,100 lines at the HS9 level. Some HS6 coes have only a single HS9 coe, renering our test invali so we iscar all such HS6 coes an then take a 1 % ranom sample of the remaining HS6 coes. or these ranomly sample HS6 coes, we have taken the export ata at the HS9 level. The regression is run for each HS 9-igit coe. The test will come from comparing the istance coefficient estimates for the HS6 aggregate with the more isaggregate estimates for the HS9 coes encompasse by the HS6 coe. Table 3 summarises the results. The left panel shows that HS6 result, with the first column isplaying the HS6 coe, the secon the istance coefficient an the thir the p-value. The fourth column summarise the fining by putting a + or where the istance coefficient is positive or negative at the 5% level respectively. The right panel shows similar statistics for the HS9 coes that make up the HS6 coe liste. The final column summarises the match between the two as a half-match (H), a full-match () or more-than-half-match (M). N inicates no-match. If HS9 heterogeneity is a key source of the low number of goos that can be clearly classifie as quality or price goos, then we shoul see a stark mismatch between the results inicate by the HS9 ata an the more aggregate HS6 ata, i.e. many N in the final column. Out of the total 16 HS 6-igit coes we checke, 7 have full fit an 3 have more than a half fit. Only 3 have no match. Closer examination of the no-match cases is revealing. In two of the cases ( an ), the istance coefficient is positive an significant for the ata poole at the HS6 level, but none of the unerlying HS9 coefficients are significant. We note however that the HS9 coefficients are positive. This suggests that the lack of variation in the HS9 export estinations may account for the lack of statistical significance. The remaining case of no match (901890) is very instructive. Here we see that most of the HS9 coefficients are positive, but one is negative. The aggregate HS6 coefficient is estimate to be negative but not significantly ifferent from zero. This case suggests that the HS6 classification is inappropriate for our purposes in that it pools price an quality goos. If this sort of result were wiesprea, then it woul cast oubt on our HS6 results, but the fact that it occurs in only one of the 16 cases provies some assurance that our HS6 estimates are yieling useful results. While further testing is require, using for example US HS10 an EU HS8 ata, our exploratory investigation suggests that analysis at HS6 level provies a reasonably goo inication of the unerlying situation. It is particularly important to note that in no case i the HS6 result inicate that the line was a quality or price goo when the HS9 ata inicate otherwise. Our fining oes not resolve the issue of the low number of clearly classifiable HS6 coes. We cannot say whether there is something wrong with our empirical framing of the question, or whether the problem lies in heterogeneity (by exporting country) of basket of goos in the HS9 coes. 6

8 Table 3: HS6 vs HS9 estimates HS6 coe Distance coefficient p Sign of significant coefficients HS9 coe Distance coefficient p Sign of the coefficient estimate it H N MH H MH N H MH N 7

9 4. CONCLUSION This seeks to classify HS6 igit proucts as being characterise either by price-competition or by quality-competition goos as suggeste by the HT an QHT moels. We fin that about 8% of HS6 coes can be thus classifie in the sense that results for the US, Germany an Japan all coincie. Of the goos that can be clearly classifie, we fin that the major EU countries are leaing the quality competition race with 50 to 60 % of their exports being goos in which quality seems to matter in the sense that the goos that get sol to the most istant markets are the most expensive ones. or Japan an the US the figures are lower at 38 an 34 % respectively. Canaa, Australia an China have much lower proportions of quality-competition goos. To check for problems with aggregation, we analyse Japan s export ata at the HS 9-igit level an its aggregation into the HS6 scheme. We fin few ifference in the inferences that come from analysis at the HS9 versus HS6 level. REERENCES Balwin, Richar an Harrian, James (2007). "Zeros, Quality an Space: Trae Theory an Trae Evience " CEPR Discussion Paper No Balwin, Richar (1988). "Hysteresis in Import Prices: The Beachhea Effect," American Economic Review, American Economic Association, vol. 78(4), pages , September. ontagné, Gaulier, an Zignago. (2008) "Specialization across varieties an North-South competition", Economic Policy, January 2008, Krugman, P. (1980) Scale Economies, Prouct Differentiation an the Pattern of Trae American Economic Review, 70, Melitz, Marc J. (2003),.The Impact of Trae on Intra-Inustry Reallocations an Aggregate Inustry Prouctivity,. Econometrica, 71:6, pp Schott, Peter. (2008) "The relative sophistication of Chinese exports", Economic Policy, January 2008,