Tailwinds from the East: how has the rising share of imports from emerging markets affected import prices?

Tailwinds from the East: how has the rising share of imports from emerging markets affected... Abstract This paper quantifies the effect of the rising share of imports from emerging market economies (EMEs) on import price inflation in the UK. Using a panel regression approach that accounts for heterogeneity across industries, we estimate that between 1999 and 2011, the rise in China’s import share of manufactured goods lowered UK import price inflation by around 0.5 percentage points per year—we call this the ‘tailwind’. Rising imports from other EME country groups are not found to have any significant impact. Our approach allows us to decompose this effect: two-thirds arises from the direct impact of switching to lower-cost Chinese goods; the remaining third comes from other exporters lowering their prices in response to stronger competition from China. We find no evidence that higher inflation rates in EMEs has so far reduced or reversed the sign of this tailwind. 1. Introduction The process of globalisation has led to a rapid growth in trade between advanced and emerging market economies (EMEs) over the past 15 years. This has been facilitated by a number of factors, such as the reduction in legal barriers to trade, transitions towards market-oriented policies and relatively low costs of production in EMEs. As Figure 1 shows, price levels in EMEs are typically significantly lower than in advanced economies, but this differential has been diminishing over time. Fig. 1. View largeDownload slide Relative price levels of selected trading partners Source: Penn World Tables, authors’ own calculations. Fig. 1. View largeDownload slide Relative price levels of selected trading partners Source: Penn World Tables, authors’ own calculations. Many policymakers have argued that the rising share of such cheap imports from EMEs, such as China and India, have acted as a positive terms of trade shock, or ‘tailwind’, by pushing down on import price inflation in the developed world (Greenspan, 2005; Dudley, 2011; ECB, 2008; IMF, 2012). But more recently, as inflation and wage rates have increased in EMEs, some policymakers have become concerned that this tailwind may be fading or even reversing in sign (Li et al., 2012; Feyzioğlu and Willard, 2006; Dale, 2011). The focus of this paper is to examine how big the ‘tailwind’ from EME imports has been on import price inflation in the UK. There is a sizeable literature that considers the impact of globalisation on advanced economies, which can be broadly divided into two strands. A first strand explores the impact of globalisation on advanced economy labour markets—including the impact on wages, wage inequality and employment (Freeman, 2005; Feenstra and Kendall, 1997; Feenstra, 2007; Autor et al., 2013). A second strand analyses the impact of globalisation on advanced economy goods markets, including variables such as import prices, firm behaviour and domestic and consumer prices.1 Our paper falls within this second strand, alongside Kamin et al. (2008) and McCoille (2008), who also consider the impact of the rising imports from China on UK import prices. Other papers have looked at the effect of rising EME imports on domestic producer prices, including Auer and Fischer (2010) and Auer et al. (2011) at the industry level, and Melitz and Ottaviani (2008) and Bugamelli et al. (2010) with firm-level data. Several papers have looked at how globalisation has fed through to CPI inflation. For example, Wheeler (2008) examines how the improvement in terms of trade from China has affected overall UK consumer goods prices. While Ball (2006) argues that there is no obvious theoretical reason why relative price shifts should have any connection with overall prices, some policymakers have argued that lower import prices may have an effect on inflation in the short run (e.g. Bean, 2006; Mishkin, 2009; Rogoff, 2006). Elsewhere, other work has looked at the effect of EME growth on commodity prices (e.g., Millard and Lipínska, 2012) and the role of global slack in Phillips curve equations (Borrio and Filardo, 2006; Calza, 2009) Kamin et al. (2008) describe three distinct channels through which rising EMEs exports can affect the import prices of their trading partners. First, as EME exporters gain market share, they will tend to displace similar goods from other countries which have a lower price level. We call this shift in the composition of imports towards cheaper EME producers the ‘switching effect’. Second, faced with increased competition from EMEs, there may be increased pressure on non-EME exporters to lower their prices—which we call the ‘competition effect’. A third, potentially counterveiling, channel is rising prices in EMEs. If the price of imports from EMEs is rising faster than those from advanced economies, then sectors with greater exposure to EMEs will see higher import price inflation. We term this the ‘inflation effect’. We evaluate the size of each of these three channels on UK import prices over the period 1999–2011. We do so using a panel regression approach using highly disaggregated industry level data on volume and value of import prices, from which we can calculate a unit-value-based measure of import price inflation. EMEs are split into three distinct groups—China, the New EU Member States and other low-cost producers. Of these, we find that only China has a statistically significant downward impact on UK import prices, with no significant effect from the two other EME country groups. Chinese imports are estimated to have lowered annual import price inflation by 0.5 percentage points on average each year of our sample. Of this, we estimate that two-thirds arises from the direct switching channel and one-third from the indirect competition channel. We find no evidence of a significant ‘inflation effect’ as yet. Our estimates are broadly in line with previous UK work based on an accounting approach (McCoille, 2008). But they are larger than Kamin et al. (2008), who report an average effect of –0.25pp for a sample of developed economies, but zero for the UK. Their smaller estimates can partly be explained by their earlier sample period (1993–2001), which predates the years in which China’s market share grew most rapidly in the UK. We contribute to the existing literature in three ways. First, while previous studies assume the tailwind—or impact of EME exports on import price inflation—is homogeneous across industries, we allow for heterogeneity and find it to be important. Specifically, we show that failure to account for heterogeneity generates biased (and smaller) estimates of the size of the tailwind. Second, our panel regression approach allows us to bootstrap standard errors and therefore assess if our estimate of the China tailwind is statistically significant. Third, while the previous literature had either estimated the combined impact of the switching and competition effects, or just captured the switching effect, the disaggregated nature of our data allows us to devise a method of estimating each of these two effects separately, as well as considering the potential headwind from the inflation effect. The remainder of this paper is set out as follows. Section 2 describes the dataset we use, noting the key features of our data that have not been exploited in the literature before. Section 3 sets out our empirical approach. Section 4 sets out main empirical results, including robustness checks. The final section concludes. 2. Dataset We use annual data on imports from the Tradeinfo database, published by the UK’s customs authority—Her Majesty’s Revenue and Customs (HMRC). This records both the value and volume of imports to the UK by trading partner and at industry level according to the Standard International Trade Classification (SITC) system. The key advantage of these data is that they are available at a detailed industry level. In its most disaggregated form, this data covers 3,000 distinct industries, with around 2,000 in manufacturing. The SITC system denotes each industry by a 5-digit code. The first digit corresponds to the broadest sectoral classification; subsequent digits give finer degrees of sectoral disaggregation. In what follows, we refer to a group of industries whose 5-digit codes share the same initial N digits as being in the same ‘N-digit industry’.2 The hierarchical nature of the SITC becomes important later on in the paper when we discuss the role of heterogeneity across in our estimations. To keep our dataset computationally manageable and to avoid possible missing data issues for country-specific control variables, we restrict our attention to 45 of the UK’s largest trading partners.3 Collectively, these countries account for around 90% of total UK imports in each year of our sample and represent around 1.5 million data points. We aggregate the country-specific data to build a panel dataset without a country dimension, but where the cross-sectional unit is the (5-digit) industry. Details of the construction of our variables are outlined briefly below. For more detail and additional descriptive statistics, see Appendix A. We split the EMEs into three groups: China, which has by far the largest share of UK imports of any EME; the New Member States of the EU from Central and Eastern Europe (‘NMS’), who represent a geographically proximate and economically broadly similar group of low-cost producers, with whom the UK has been steadily integrating;4 and ‘other’ low wage countries, which consists of Brazil, India, Indonesia, Mexico, Russia, Turkey, Thailand, Philippines, Pakistan and Vietnam, denoted LWC.5 Over our sample period, China’s share of all imports has increased rapidly, from 2% of all imports in 1999 to just under 9% by 2011 (Figure 2a); with China becoming the second largest single importer to the UK, after Germany. Figure 2b shows that the rising EME market shares have been most noticeable in the manufacturing sector, where China accounts for 13% of imports, and EMEs collectively for just over a quarter. Imports from the New EU Member States have also increased, but at a significantly less rapid pace, and the market share of LWCs has risen by only 1pp over the entire period. Fig. 2. View largeDownload slide UK imports from Emerging Market Economies Fig. 2. View largeDownload slide UK imports from Emerging Market Economies Our dependent variable is the log difference in the unit value of imports for a given industry in a given year, where the unit value is calculated by dividing the total value of imports (across all producers) by the total volume of imports. Import share is defined as the ratio of the value of imports from each EME group divided by the total value of imports.6 To control for the influence of exchange rate fluctuations, we construct an industry-specific exchange rate index. This is defined as the weighted average of bilateral nominal exchange rate changes between year t-1 and t, where the weights are given by each country’s share of imports in year t in a particular industry. Since the weights of each country differ across industries, this index will vary along both the time and industry dimensions. Exchange rates are expressed in the European style, where a rise in the index denotes an appreciation of sterling. 3. Empirical approach Our empirical approach distinguishes between three distinct channels through which EMEs may affect aggregate import prices. For ease of exposition, consider the case where there are just two groups of exporters—advanced economies and emerging markets. Aggregate import prices in each industry i and at time t, P can be written as a weighted average of the price of imports from EMEs ( PEME) and the price of imports from advanced economies ( PADV), with the share of EMEs given by S. The log change in import prices can then be written as: ln Pt−lnPt−1=StlnPtEME−St−1lnPt−1EME+ (1−St)lnPt−1ADV−(1−St−1)lnPt−1ADV (1) Utilising the property that a variable in levels is equal to the sum of its lag and its change, we have: ln Pt−lnPt−1=(St−1+ΔSt)lnPtEME+(1−St−1−ΔSt)lnPt−1ADV−St−1lnPt−1EME−(1−St−1)lnPt−1ADV (2) This can then be rearranged as: where πEME, and πADV denotes the log difference in PEME and PADV, respectively. The first term captures the ‘switching effect’ that arises as an economy starts to import more from EMEs and less from advanced economies, where products are typically more expensive. The larger the price differential between imports from EME and imports from other advanced economies, the larger effect this will have on pushing down prices—hence the negative sign. The second term captures the ‘inflation effect’ stemming from fast export price growth in EMEs—the larger the lagged share imports from EME and the larger the export inflation differential in exports between advanced and emerging economies, the bigger (upward) effect this will have on prices. The final term is simply the rate of import price inflation from advanced economies.7 In a pure accounting sense, this is independent of the share of imports from EMEs. However, economically speaking one might expect this to be influenced by EME’s import shares via a ‘competition effect’—if a higher share of EME imports forces the incumbent advanced economy producers to lower their markups, it will show up via this final term.8 This basic decomposition can be extended to include more than two groups of countries. For K country groups (1…K), we have: lnPt−lnPt−1=∑K(ΔStk(ln Ptkln PtADV)+St−1k(πtk−πtADV))+πtADV (4) Our first set of panel regression models the lagged share of EME imports and the change in the EME import share for each of our three groups of countries—China, New EU Member States, and other Low Wage countries. The coefficient on the lagged share of EME imports will pick up the inflation effect. The coefficient on the change in EME import share picks up the combined effect of the switching and competition effects, which we collectively refer to as the ‘price level effect’. A subsequent set of regressions is used to split this up into its two components. This first set of panel regressions takes the following form: πit=α+β1Sit−1CHINA+β2ΔSitCHINA+γ1Sit−1NMS+γ2ΔSitNMS+ϕ1Sit−1LWC+ϕ2ΔSitLWC+θexchit+μi+λt+ψ'Xit+εit (5) where the dependent variable, πit is log difference in the sterling value of unit values of imports in each period t and for each industry i. Sit−1 denotes the market share of each EME group—China, EU New Member States (NMS) and other low wage countries (LWC). We also include an exchange rate term, exch, defined as the log difference of an industry-specific exchange rate index. Adding this term allows us to capture the average rate of exchange rate pass-through to import prices. Importantly, we remain agnostic about whether exporters use local or producer currency pricing, and estimating this equation at industry levels allows this pricing behaviour to vary according to goods type.9 As is standard in a panel regression, we add time and industry fixed effects, but we also follow Kapetanios et al. (2011) by including the within-period averages of each of the regressors and the dependent variables (by 4-digit industry), denoted in the equation in matrix form as ψ'Xit. Econometrically, these terms allow common correlated effects that are not picked up by other terms in our equation. For example, an industry-specific positive productivity shock that hits domestic producers in manufacturing may induce foreign importers to alter the price of imports. The equation is estimated by ordinary least squares (OLS). Auer and Fischer (2010), who investigate the effect of import competition on domestic producer price inflation (PPI), argue that it is necessary to instrument the change in EMEs market share, and failure to do so could substantially bias estimates. That is because any positive demand shock is likely to increase both producer prices and the share of goods imported from EMEs as a percentage of the domestic market. However, the regression equation estimated in this paper is unlikely to suffer from this same endogeneity problem for three reasons. First, our research question embodies a different independent variable—we consider imports as a share of total imports, rather than as a share of domestic production plus imports. A cyclical demand shock that increases the demand for imports of a particular good is likely to increase demand from all countries proportionately. This would leave the former measure of market share unchanged, but not the latter.10 Second, our independent variable is based on imports to the UK—as opposed to domestically produced goods—which is also less likely to be related to cyclical conditions in the domestic economy than domestic producers would be. Third, the impact of any potential demand side factors that could bias our regression coefficient estimates should be mitigated by the inclusion of the 4-digit industry averages of our dependent and explanatory variables, which seek to capture the effects of any industry-specific shocks (including demand shocks). a. Quantifying the ‘price level’ effect A key goal of the analysis is to estimate the overall size of the price-level effect, as opposed to merely providing estimates of the coefficient. We can compute the overall size of the price-level effect using the coefficients from the regression above using the following expression: Price level effect=∑i=1Iwit⋅β^2i⋅ΔSitCHINA (6) where ΔSitCHINA is the actual observed change in market share in industry i, β^2i is the estimated coefficient of the change in China share,11 and wit is the weight of sector i in total imports at time t. b. Quantifying the ‘inflation effect’ Similarly, the inflation effect is picked up the coefficient on the lagged China share in the regressions. The overall size is given by: Inflation effect=∑i=0Iwit⋅β^1i⋅Sit−1CHINA (7) 4. Results 4.1 Heterogeneity across industry The estimation results are presented in Table 1. The first column shows that when we pool over all industries (regression I), the changes in the market share of China, NMS and other LWCs are for the most part both significant and negative, although the size of the coefficient is two to three times larger for China. The lagged market share variables are insignificant, suggesting that higher inflation in EMEs has not fed through to UK import prices. Table 1. Baseline regression results [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 The dependent variable in each regression is the log change in import prices. Note: Coefficients for variables in the vector X are not reported here for space reasons. *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. Standard errors are in parentheses. View Large Table 1. Baseline regression results [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 The dependent variable in each regression is the log change in import prices. Note: Coefficients for variables in the vector X are not reported here for space reasons. *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. Standard errors are in parentheses. View Large However, these results mask considerable variation across industry groups. For food- and commodity-based products (regression II), none of the coefficients on changes in market share is significant. By contrast, in chemicals (regression III) and manufacturing (regression IV), both China and other LWCs do appear to exert a significant downward effect on prices via gaining market share. That said, while the EME group market shares in the chemicals sector have been relatively constant over our sample period, the shares for manufacturing have risen rapidly (see Appendix B). Therefore, and in keeping with most previous studies, we restrict the focus of this paper to the manufacturing sector.12 Splitting the manufacturing sector into its three separate 1-digit industries also reveals considerable differences. For machinery, our estimates suggest that ceteris paribus a 1% rise in Chinese market share is associated with a fall in prices of 0.82% (regression VII), compared to a fall of only 0.47% in materials (regression VI). The downward pressure exerted by NMS is only significant in manufactured articles, whereas other LWCs are significant in materials and machinery. Again, the lagged market share for all EME groups is generally insignificant. This analysis at finer levels of disaggregation raises the important question of what is the appropriate level of disaggregation. To explore this, we estimate regressions at the 2-digit level, where we continue to find variation across the 26 industry groups (see Appendix D).13 As Table 2 shows, when we test the implied restriction of pooling across 1- and 2-digit industries, we find a clear rejection of the hypothesis of equal coefficients at either the 1- or the 2-digit industry levels. Table 2. Tests for pooling Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 View Large Table 2. Tests for pooling Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 View Large We compute the estimated total China price level effect under three different specifications using the methodology set out in equation (4). The results from this exercise are shown in Figure 3. It shows that the estimated China effect is much larger when the equations are estimated at the 2-digit industry level. In other words, failure to account for coefficient heterogeneity reduces the estimated size of the China effect by around a third. Fig. 3. View largeDownload slide Price-level effects of China under different pooling assumptions Fig. 3. View largeDownload slide Price-level effects of China under different pooling assumptions Given both the economic and statistical significance of coefficient heterogeneity, our preferred specification is to estimate separate regressions for each 2-digit industry. This baseline suggests that the tailwind from China is around –0.72pp per annum over our sample period. Since manufacturing accounts for around two-thirds of all UK imports, this is equivalent to a stand-alone effect on all import prices of around –0.49pp. Looking at the profile of this effect over time, there is no obvious sign of a trend, suggesting that the price-level effect of China has not waned over time. To check for sensitivity of our results to different specifications, we performed a variety of robustness checks. First we computed the China price-level effect for alternative specifications. Given that the NMS and LWC market share variables were insignificant, we dropped them to see how they would change the specification. There was little difference (see Appendix E). To check for autocorrelation, we included a lagged dependent variable. And to check for the importance of common correlated effects, we also dropped the 4-digit industry averages of all variables. But our estimated China price-level effect was very similar to our baseline case. We also checked if the results were driven by insignificant coefficients—that is, we repeated the calculations by re-coding the 12 insignificant industry coefficients (out of 26) to zero. They produce a very similar estimate of the China price level effect. 4.2 Computing confidence intervals Whilst the regression coefficients on the China share are statistically significant for most industries, the above estimate of the total China effect does not give any indication as to whether the estimated effect of China and other EMEs is statistically significantly different from zero. To assess this, we used Monte Carlo methods to estimate a confidence interval. Specifically, we take 10,000 draws from the estimated parameter distribution of each βi, and use each draw to compute the China effect based on the values drawn. From this, we compute the 95% confidence interval by discarding the top and bottom 2.5% of the distribution. This confidence interval is shown in Figure 4.14 As a robustness check, we also constructed a confidence interval using a bootstrapping technique based on re-sampling residuals across industry, which unlike the Monte Carlo approach allows for any correlation in residuals across equations.15 This yields very similar results. Fig. 4. View largeDownload slide China price-level effect Fig. 4. View largeDownload slide China price-level effect Over the full sample period, 1999–2011, the 95% confidence interval in Figure 4 never crosses the y-axis, and hence we conclude that China’s impact is significant at the 5% level in each year. A similar exercise for estimating the mean impact and confidence intervals for LWC and NMS shows that their impacts are not statistically different from zero.16 Even though initial regressions suggested that the coefficient on the change in the market share of LWC and NMS countries was significant for some manufacturing industries, when the effect is computed over all manufactured import prices, it is both small and statistically insignificant. For the former group, this likely reflects the twin facts that the growth in import market share was much less strong than for China (Figure 2b) and that the estimated size of the fall inflation for a given gain in market share was much lower for these country groups (Table 1) than for China. The mean inflation effect and 95% confidence intervals are shown in Figure 5. They straddle the zero line, implying that rising inflation in China is not having any statistically significant effect on UK import prices. A similar exercise was carried out to compute the inflation effect of NMS and LWC; the confidence intervals for the inflation channel in these country groups were also very wide and no different from zero (see the figures in Appendix F). Fig. 5. View largeDownload slide Estimates of the China inflation effect Fig. 5. View largeDownload slide Estimates of the China inflation effect 4.3 Decomposing the price-level effect into the switching and competition effects The competition effect captures the response of non-Chinese producers to a change in China’s market share. To isolate this effect, we run a second set of regression where the dependent variable is the log change in import prices (or the unit value of inflation) from all countries excluding China.17 πitEXC=α+β1'Sit−1CHINA+β2'ΔSitCHINA+γ1'Sit−1NMS+γ2'ΔSitNMS+ϕ1'Sit−1LWC+ϕ2'ΔSitLWC+θ'exchit+μ'i+λ't+ψ'Xit+εit (8) The coefficient β'2 captures the response of non-Chinese producers to a 1pp rise in China’s market share. If this coefficient is not statistically different from zero, this indicates no significant pricing response to China gaining market share; if this is significant, it indicates that other producers do respond to Chinese entry. The total size of the competition effect is given by:18 Competition effect=∑i=1Iwit⋅β^'2i⋅ΔSitCHINA⋅(1−SitCHINA) (9) The switching effect is then calculated as the gap between the point estimates of the total price-level effect and the competition effect. Figure 6 below shows the resulting decomposition. Of course, since both point estimates come from separate models, they are each subject to a standard error, but our methods do not allow us to compute a standard error for the difference between them, since we cannot estimate the covariance of the estimates over the two sets of errors—and hence we only have a point estimate of the switching effect. Our results nevertheless point to a qualitatively important role for competition effects. Fig. 6. View largeDownload slide Decomposing the China combined price-level effect Fig. 6. View largeDownload slide Decomposing the China combined price-level effect It suggests that the switching effect accounts for about two-thirds of the total price level effect, with the remaining third attributed to the competition effect. 5. Conclusions In this paper, we quantify the effect of rising import penetration from emerging market economies on UK import prices using a rich panel dataset. This highly disaggregated industry data allow us to account for heterogeneity across industries and across the emerging market economies that export to the UK. We find robust evidence that the rise in China’s share of the markets has acted as a tailwind, lowering manufacturing import price inflation by an estimated 0.7pp on average a year over the period 1999–2011; this is equivalent to a standalone effect of –0.5pp on overall import prices. We find no evidence of a statistically significant effect from the other EME country groups. Constructing the confidence interval of the tailwind by Monte Carlo methods, we find that the China tailwind is indeed statistically significant, but there is no evidence of a significant tailwind from the EU New Member States or other low wage cost economies (including India and Brazil). Finally, this paper finds that around two-thirds of the China tailwind comes via a change in the composition of the import basket, to reflect a greater share of cheaper goods from China. The remaining one-third arises via competition effects, as non-Chinese exporters to the UK lower their prices in response to the increased competition from China. There is no significant impact of the rising inflation rate in China during the latter years of our sample period. Footnotes 1 Auer and Fischer (2010) find that the distribution of price shocks has a strong negative skew, which may interact with menu costs to generate downward pressure on aggregate prices even if the mean of price shocks is zero. 2 For example, ‘Corks and Stoppers of Natural Cork’ has the 5-digit code ‘63311’. Its 1-digit industry (6) is ‘Manufactured articles classified chiefly by material’; its 2-digit industry (63) is ‘Cork and wood manufactures’; its 3-digit industry (633) is ‘Cork manufactures’; and its 4-digit industry (6331) is ‘Articles of natural cork’. 3 The 45 countries are: all EU and OECD members, plus Argentina, Brazil, China, Hong Kong, India, Indonesia, Malaysia, Pakistan, Philippines, Qatar, Russia, Saudi Arabia, South Korea, Taiwan, Thailand and Vietnam. Collectively they account for over 90% of UK imports. The ratio of imports from these countries to total imports from all trading partners is broadly constant over the sample period, and so our country choice does not result in the exclusion of groups of countries which have also seen a significant rise in their overall market share. 4 The New EU Member States (NMS) group of countries includes Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, Slovenia and Romania. 5 All these countries have lower relative price levels than the UK’s main advanced economy trading partners (see Appendix D). 6 Throughout the paper we use the term ‘total’ to refer to summing across the 45 major countries the UK imports from. 7 One could also write the decomposition as: ln Pt−lnPt−1=ΔSt(ln PtEMEln PtADV)+St−1(πtEME)+(1−St)πtADV, but we use an alternative specification as it facilitates discussion of the three channels. 8 Another possible channel is if EME competition drives out the most expensive advanced producers, leaving cheaper advanced economy producers in the market, then the measured πtADV will be lower due to the shift in composition of advanced economy exporters. 9 We use the contemporaneous rather than lagged exchange rate because the literature (e.g. Gopinath et al., 2010) finds that pass-through from the exchange rate to import prices usually takes place within one year. That said, in Section 4 we investigate how sensitive our results are to different assumptions and find that they are robust to additional lags of the exchange rate. 10 This shock, with a proportional increase in demand for imports from all countries, would result in a rise in the ratio of EME or Chinese imports to the overall size of the domestic market. 11 In what follows, we also run regressions into samples split by one- and two-digit industry. That means our estimated coefficient used in equation (4) can differ across industry groups. We therefore include an i subscript on β̂ in equation (4). 12 As a further check, we also repeated our analysis for the non-manufacturing industries. We find that the price-level effect on inflation each year was +0.04pp for China, +0.08pp per year for New Member States and -0.02pp per year for Low Wage Countries. We therefore conclude that none of the EME groups exerted a meaningful effect on non-manufacturing import prices over the sample period. 13 If we go down to the 3-digit level, we run into the problem that some 3-digit industries contain only a single 5-digit industry and hence panel estimation cannot be used. 14 The mean estimate from this exercise is almost identical to our estimate of the tailwind reported in Figure 3. 15 Specifically, we decomposed the data into fitted values and residuals. Taking the residuals by year, we obtain 11 sets of residuals, which formed our sampling population. We then generated a synthetic dataset by adding the fitted values for each year to a randomly chosen residual vector (sampling with replacement). We then calculated the estimate China effect from this synthetic dataset, and repeated the whole procedure 10,000 times. The 95% confidence interval was then given by discarding the upper and lower 2.5% of estimates. See the appendix for charts. 16 See Figures D2 and D3 in Appendix D. 17 This is calculated as the aggregate value of non-Chinese imports in industry i at time t to the aggregate volume of non-Chinese imports in industry i at time t. 18 We multiply by ( (1−SitCHINA), because we estimate our competition effect only over non-Chinese imports, but wish to calculate the effect on the price of all imports in a given industry. Bibliography Auer , R. , Degen , K. and Fischer , A . 2011 . Low-wage import competition, inflationary pressure and industry dynamics in Europe , European Economic Review , vol. 59 , no. C , 141 – 66 Auer , R. and Fischer , A . 2010 . The effect of low-wage import competition on U.S. inflationary pressure , Journal of Monetary Economics , vol. 57 , 491 – 503 Google Scholar CrossRef Search ADS Autor , D. , Dorn , D. , Hanson , H. and Song , J . 2013 . The China syndrome: local labour market effects of import competition in the United States , American Economic Review , vol. 103 , no. 6 , 2121 – 68 Google Scholar CrossRef Search ADS Ball , L . 2006 . ‘ Has Globalization Changed Inflation ?’, NBER Working Papers 12687 Bean , C . 2006 . Globalisation and inflation , Bank of England Quarterly Bulletin , Q4 , 468 – 75 Borrio , C. and Filardo , A . 2006 . ‘ Globalisation and inflation: New Cross Country Evidence on the Global Determinants of Domestic Inflation ’, BIS Working Paper 227 Broda , C. and Romalis , J . 2009 . ‘The Welfare Implications of Rising Price Dispersion’ , mimeo , Chicago Booth University Bugamelli , M. , Fabiani , S. and Sette , E . 2010 . ‘ The Pro-Competitive Effect of Imports from China: An Analysis of Firm-Level Price Data ’, Bank of Italy , Working Paper No. 737 Google Scholar CrossRef Search ADS Calza , A . 2009 . Globalization, domestic inflation and global output gaps: evidence from the Euro area , International Finance , vol. 12 , no. 3 , 301 – 20 Google Scholar CrossRef Search ADS Dale , S . 2011 . ‘ MPC in the Dock ’, speech at the National Asset-Liability Management Global Conference , 24 March Dudley , W . 2011 . ‘US Economic Policy in a Global Context’ , Remarks at the Foreign Policy Association Corporate Dinner , New York , http://www.newyorkfed.org/newsevents/speeches/2011/dud110607.html European Central Bank . 2008 . Globalisation, trade and the Euro area macroeconomy , Monthly Bulletin , January, 75–88 Feenstra , R . 2007 . ‘ Globalization and Its Impact on Labour ’, Vienna Institute for International Economic Studies Working Paper No. 44 Feenstra , R. and Kendall , D . 1997 . Pass-through of exchange rates and purchasing power parity , Journal of International Economics , vol. 43 , 237 – 61 Google Scholar CrossRef Search ADS Feyzioğlu , T. and Willard , L . 2006 . ‘ Does Inflation in China Affect the United States and Japan ?’, IMF Working Paper No. 06/36 Freeman , R . 2005 . Are your wages set in Beijing ?, Journal of Economic Perspectives , vol. 9 , 15 – 32 Google Scholar CrossRef Search ADS Gopinath , G. , Itskohiki , O. and Rigobon , R . 2010 . Currency choice and exchange rate pass-through , American Economic Review , vol. 100 , 304 – 36 Google Scholar CrossRef Search ADS Greenspan , A . 2005 . ‘Globalization: Remarks to the Council on Foreign Relations’ , 10 March, http://www.federalreserve.gov/boarddocs/speeches/2005/20050310/default.htm IMF . 2012 . ‘World Economic Outlook’ Kamin , S. , Marazzi , M. and Schindler , J . 2008 . The impact of Chinese exports on global import prices , Review of International Economics , vol. 14 , no. 2 , 179 – 201 Google Scholar CrossRef Search ADS Kapetanios , G. , Pesaran , H. and Yamagata , T . 2011 . Panels with non-stationary multifactor error structures , Journal of Econometrics , vol. 160 , 326 – 48 Google Scholar CrossRef Search ADS Koske , I. , Pain , N. and Sollie , M . 2008 . Globalisation and OECD consumer price inflation , OECD Economic Studies , vol. 44 Li , H. , Li , L. , Wu , B. and Xiong , Y . 2012 . The end of cheap Chinese labour , Journal of Economic Perspectives , vol. 26 , no. 4 , 57 – 74 Google Scholar CrossRef Search ADS McCoille , C . 2008 . The impact of low-cost economies on UK import prices , Bank of England Quarterly Bulletin , Q1 , 58 – 63 Melitz , J. and Ottaviani , P . 2008 . Market size, trade, and productivity , Review of Economic Studies , vol. 75 , no. 1 , 295 – 316 Google Scholar CrossRef Search ADS Millard , S. and Lipinska , A . 2012 . Tailwinds and headwinds: how does growth in the BRICs affect inflation in the G-7 ?, International Journal of Central Banking , vol. 8 , 227 – 66 Mishkin , F . 2009 . Globalization, macroeconomic performance and monetary policy , Journal of Money, Credit and Banking , vol. 41 , no. S1 , 187 – 96 Google Scholar CrossRef Search ADS Rogoff , K . 2006 . ‘Impact of Globalization on Monetary Policy’, paper prepared for symposium on ‘The New Economic Geography: Effects and Policy Implications’ , http://kansascityfed.org/publicat/sympos/2006/pdf/19Rogoff.pdf Wheeler , T . 2008 . ‘ Has Trade with China Affected UK Inflation ?’, External MPC Unit Discussion Paper 22, Bank of England Appendix A. Data Exchange rate data Let V denote the value of imports of country j in industry i at time t. The weight of each country is given by: wijt=Vijt∑j=0IVijt The exchange rate, e, is the average annual bilateral nominal exchange rate, extracted from the Thompson datastream. Our index of exchange rate changes is defined as: exchit=∑j=0Iwijt(ejtejt−1−1) Exchange rates are defined in the European style, so a positive value of exch corresponds to an appreciation in sterling. For countries which adopted the euro during our sample period, we use the exchange rate between sterling and the legacy country, which during the post-euro adoption period is calculated by multiplying the official conversion rate with the sterling euro exchange rate. The figure below shows the mean, maximum, minimum and inter-quartile range of the exchange rate across all 5-digit industries in each year of the sample period. Global wage and inflationary pressure indices Let p denote the Consumer Price Index as measured as the annual rate of CPI inflation, as reported in the IMF’s World Economic Outlook. The global inflationary pressure index is given by: Infit=∑j=0Iwijt(pjtpjt−1−1) Similarly, denoting nominal wages with W, the global wage index is given by: Wagesit=∑j=0Iwijt(WjtWjt−1−1) Market share of low-cost producers We calculate the market share of a subset of K countries, as follows: Sikt=∑j∈KVijt∑j∈JVijt For the three variables, K is defined as follows: China: China New Member States (NMS): Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, Slovenia and Romania Low Wage Cost (LWC): Brazil, Mexico, Russia, Turkey, Thailand, Indonesia, India, Philippines, Pakistan and Vietnam Appendix B. Decomposing the growth in each group’s market share by industry group View largeDownload slide View largeDownload slide Appendix C. Comparison with ONS data Our analysis of import prices is based on unit-value data, which does not adjust for product varieties or quality differences of goods within each 5-digit industry. This might be particularly relevant when looking at EMEs, as Broda and Romalis (2009) show Chinese imports tend to be concentrated in lower-quality varieties of the same product class. To check for this source of bias in our results, we compare results obtained from our data with official import price indices (quality adjusted) published by the Office for National Statistics. But since the ONS data are only available for a selection of 2-digit industries, a side-by-side comparison of regression results for all available industries is presented in Table C1. When the regression is run over all sectors, the ONS data doesn’t yield a significant coefficient on the change in China’s market share, or that of NMS; but the LWC share is significant, albeit with the ‘wrong’ sign. However, the ONS import price indices predominantly cover non-manufacturing industries. Restricting the sample to manufacturing industries, we find that the coefficient on the change in China’s market share is very similar for both measure of import price inflation. The change in NMS share is insignificant in both, and the change in LWC share is significant only when unit-value data is used, which may reflect the lack of quality adjustment in these economies. Table C1. HMRC vs. ONS data [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 View Large Table C1. HMRC vs. ONS data [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 View Large Appendix D. Regression coefficients at the 2-digit level 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 N.E.S: ‘not elsewhere specified’; *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. View Large 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 N.E.S: ‘not elsewhere specified’; *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. View Large Appendix E. Robustness checks Fig. E1. View largeDownload slide China market share effect under different specifications Fig. E1. View largeDownload slide China market share effect under different specifications Fig. E2. View largeDownload slide China price-level effect Fig. E2. View largeDownload slide China price-level effect Appendix F. Additional charts Fig. F1. View largeDownload slide Relative price levels Fig. F1. View largeDownload slide Relative price levels Fig. F2. View largeDownload slide The combined ‘price-level’ effect and 95% confidence interval Fig. F2. View largeDownload slide The combined ‘price-level’ effect and 95% confidence interval Fig. F3. View largeDownload slide The ‘inflation’ effect and 95% confidence interval Fig. F3. View largeDownload slide The ‘inflation’ effect and 95% confidence interval © The Author(s) 2017. Published by Oxford University Press on behalf of the Cambridge Political Economy Society. All rights reserved. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Cambridge Journal of Economics Oxford University Press

Tailwinds from the East: how has the rising share of imports from emerging markets affected import prices?

Loading next page...
 
/lp/ou_press/tailwinds-from-the-east-how-has-the-rising-share-of-imports-from-X5437NQWGi
Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press on behalf of the Cambridge Political Economy Society. All rights reserved.
ISSN
0309-166X
eISSN
1464-3545
D.O.I.
10.1093/cje/bex062
Publisher site
See Article on Publisher Site

Abstract

Abstract This paper quantifies the effect of the rising share of imports from emerging market economies (EMEs) on import price inflation in the UK. Using a panel regression approach that accounts for heterogeneity across industries, we estimate that between 1999 and 2011, the rise in China’s import share of manufactured goods lowered UK import price inflation by around 0.5 percentage points per year—we call this the ‘tailwind’. Rising imports from other EME country groups are not found to have any significant impact. Our approach allows us to decompose this effect: two-thirds arises from the direct impact of switching to lower-cost Chinese goods; the remaining third comes from other exporters lowering their prices in response to stronger competition from China. We find no evidence that higher inflation rates in EMEs has so far reduced or reversed the sign of this tailwind. 1. Introduction The process of globalisation has led to a rapid growth in trade between advanced and emerging market economies (EMEs) over the past 15 years. This has been facilitated by a number of factors, such as the reduction in legal barriers to trade, transitions towards market-oriented policies and relatively low costs of production in EMEs. As Figure 1 shows, price levels in EMEs are typically significantly lower than in advanced economies, but this differential has been diminishing over time. Fig. 1. View largeDownload slide Relative price levels of selected trading partners Source: Penn World Tables, authors’ own calculations. Fig. 1. View largeDownload slide Relative price levels of selected trading partners Source: Penn World Tables, authors’ own calculations. Many policymakers have argued that the rising share of such cheap imports from EMEs, such as China and India, have acted as a positive terms of trade shock, or ‘tailwind’, by pushing down on import price inflation in the developed world (Greenspan, 2005; Dudley, 2011; ECB, 2008; IMF, 2012). But more recently, as inflation and wage rates have increased in EMEs, some policymakers have become concerned that this tailwind may be fading or even reversing in sign (Li et al., 2012; Feyzioğlu and Willard, 2006; Dale, 2011). The focus of this paper is to examine how big the ‘tailwind’ from EME imports has been on import price inflation in the UK. There is a sizeable literature that considers the impact of globalisation on advanced economies, which can be broadly divided into two strands. A first strand explores the impact of globalisation on advanced economy labour markets—including the impact on wages, wage inequality and employment (Freeman, 2005; Feenstra and Kendall, 1997; Feenstra, 2007; Autor et al., 2013). A second strand analyses the impact of globalisation on advanced economy goods markets, including variables such as import prices, firm behaviour and domestic and consumer prices.1 Our paper falls within this second strand, alongside Kamin et al. (2008) and McCoille (2008), who also consider the impact of the rising imports from China on UK import prices. Other papers have looked at the effect of rising EME imports on domestic producer prices, including Auer and Fischer (2010) and Auer et al. (2011) at the industry level, and Melitz and Ottaviani (2008) and Bugamelli et al. (2010) with firm-level data. Several papers have looked at how globalisation has fed through to CPI inflation. For example, Wheeler (2008) examines how the improvement in terms of trade from China has affected overall UK consumer goods prices. While Ball (2006) argues that there is no obvious theoretical reason why relative price shifts should have any connection with overall prices, some policymakers have argued that lower import prices may have an effect on inflation in the short run (e.g. Bean, 2006; Mishkin, 2009; Rogoff, 2006). Elsewhere, other work has looked at the effect of EME growth on commodity prices (e.g., Millard and Lipínska, 2012) and the role of global slack in Phillips curve equations (Borrio and Filardo, 2006; Calza, 2009) Kamin et al. (2008) describe three distinct channels through which rising EMEs exports can affect the import prices of their trading partners. First, as EME exporters gain market share, they will tend to displace similar goods from other countries which have a lower price level. We call this shift in the composition of imports towards cheaper EME producers the ‘switching effect’. Second, faced with increased competition from EMEs, there may be increased pressure on non-EME exporters to lower their prices—which we call the ‘competition effect’. A third, potentially counterveiling, channel is rising prices in EMEs. If the price of imports from EMEs is rising faster than those from advanced economies, then sectors with greater exposure to EMEs will see higher import price inflation. We term this the ‘inflation effect’. We evaluate the size of each of these three channels on UK import prices over the period 1999–2011. We do so using a panel regression approach using highly disaggregated industry level data on volume and value of import prices, from which we can calculate a unit-value-based measure of import price inflation. EMEs are split into three distinct groups—China, the New EU Member States and other low-cost producers. Of these, we find that only China has a statistically significant downward impact on UK import prices, with no significant effect from the two other EME country groups. Chinese imports are estimated to have lowered annual import price inflation by 0.5 percentage points on average each year of our sample. Of this, we estimate that two-thirds arises from the direct switching channel and one-third from the indirect competition channel. We find no evidence of a significant ‘inflation effect’ as yet. Our estimates are broadly in line with previous UK work based on an accounting approach (McCoille, 2008). But they are larger than Kamin et al. (2008), who report an average effect of –0.25pp for a sample of developed economies, but zero for the UK. Their smaller estimates can partly be explained by their earlier sample period (1993–2001), which predates the years in which China’s market share grew most rapidly in the UK. We contribute to the existing literature in three ways. First, while previous studies assume the tailwind—or impact of EME exports on import price inflation—is homogeneous across industries, we allow for heterogeneity and find it to be important. Specifically, we show that failure to account for heterogeneity generates biased (and smaller) estimates of the size of the tailwind. Second, our panel regression approach allows us to bootstrap standard errors and therefore assess if our estimate of the China tailwind is statistically significant. Third, while the previous literature had either estimated the combined impact of the switching and competition effects, or just captured the switching effect, the disaggregated nature of our data allows us to devise a method of estimating each of these two effects separately, as well as considering the potential headwind from the inflation effect. The remainder of this paper is set out as follows. Section 2 describes the dataset we use, noting the key features of our data that have not been exploited in the literature before. Section 3 sets out our empirical approach. Section 4 sets out main empirical results, including robustness checks. The final section concludes. 2. Dataset We use annual data on imports from the Tradeinfo database, published by the UK’s customs authority—Her Majesty’s Revenue and Customs (HMRC). This records both the value and volume of imports to the UK by trading partner and at industry level according to the Standard International Trade Classification (SITC) system. The key advantage of these data is that they are available at a detailed industry level. In its most disaggregated form, this data covers 3,000 distinct industries, with around 2,000 in manufacturing. The SITC system denotes each industry by a 5-digit code. The first digit corresponds to the broadest sectoral classification; subsequent digits give finer degrees of sectoral disaggregation. In what follows, we refer to a group of industries whose 5-digit codes share the same initial N digits as being in the same ‘N-digit industry’.2 The hierarchical nature of the SITC becomes important later on in the paper when we discuss the role of heterogeneity across in our estimations. To keep our dataset computationally manageable and to avoid possible missing data issues for country-specific control variables, we restrict our attention to 45 of the UK’s largest trading partners.3 Collectively, these countries account for around 90% of total UK imports in each year of our sample and represent around 1.5 million data points. We aggregate the country-specific data to build a panel dataset without a country dimension, but where the cross-sectional unit is the (5-digit) industry. Details of the construction of our variables are outlined briefly below. For more detail and additional descriptive statistics, see Appendix A. We split the EMEs into three groups: China, which has by far the largest share of UK imports of any EME; the New Member States of the EU from Central and Eastern Europe (‘NMS’), who represent a geographically proximate and economically broadly similar group of low-cost producers, with whom the UK has been steadily integrating;4 and ‘other’ low wage countries, which consists of Brazil, India, Indonesia, Mexico, Russia, Turkey, Thailand, Philippines, Pakistan and Vietnam, denoted LWC.5 Over our sample period, China’s share of all imports has increased rapidly, from 2% of all imports in 1999 to just under 9% by 2011 (Figure 2a); with China becoming the second largest single importer to the UK, after Germany. Figure 2b shows that the rising EME market shares have been most noticeable in the manufacturing sector, where China accounts for 13% of imports, and EMEs collectively for just over a quarter. Imports from the New EU Member States have also increased, but at a significantly less rapid pace, and the market share of LWCs has risen by only 1pp over the entire period. Fig. 2. View largeDownload slide UK imports from Emerging Market Economies Fig. 2. View largeDownload slide UK imports from Emerging Market Economies Our dependent variable is the log difference in the unit value of imports for a given industry in a given year, where the unit value is calculated by dividing the total value of imports (across all producers) by the total volume of imports. Import share is defined as the ratio of the value of imports from each EME group divided by the total value of imports.6 To control for the influence of exchange rate fluctuations, we construct an industry-specific exchange rate index. This is defined as the weighted average of bilateral nominal exchange rate changes between year t-1 and t, where the weights are given by each country’s share of imports in year t in a particular industry. Since the weights of each country differ across industries, this index will vary along both the time and industry dimensions. Exchange rates are expressed in the European style, where a rise in the index denotes an appreciation of sterling. 3. Empirical approach Our empirical approach distinguishes between three distinct channels through which EMEs may affect aggregate import prices. For ease of exposition, consider the case where there are just two groups of exporters—advanced economies and emerging markets. Aggregate import prices in each industry i and at time t, P can be written as a weighted average of the price of imports from EMEs ( PEME) and the price of imports from advanced economies ( PADV), with the share of EMEs given by S. The log change in import prices can then be written as: ln Pt−lnPt−1=StlnPtEME−St−1lnPt−1EME+ (1−St)lnPt−1ADV−(1−St−1)lnPt−1ADV (1) Utilising the property that a variable in levels is equal to the sum of its lag and its change, we have: ln Pt−lnPt−1=(St−1+ΔSt)lnPtEME+(1−St−1−ΔSt)lnPt−1ADV−St−1lnPt−1EME−(1−St−1)lnPt−1ADV (2) This can then be rearranged as: where πEME, and πADV denotes the log difference in PEME and PADV, respectively. The first term captures the ‘switching effect’ that arises as an economy starts to import more from EMEs and less from advanced economies, where products are typically more expensive. The larger the price differential between imports from EME and imports from other advanced economies, the larger effect this will have on pushing down prices—hence the negative sign. The second term captures the ‘inflation effect’ stemming from fast export price growth in EMEs—the larger the lagged share imports from EME and the larger the export inflation differential in exports between advanced and emerging economies, the bigger (upward) effect this will have on prices. The final term is simply the rate of import price inflation from advanced economies.7 In a pure accounting sense, this is independent of the share of imports from EMEs. However, economically speaking one might expect this to be influenced by EME’s import shares via a ‘competition effect’—if a higher share of EME imports forces the incumbent advanced economy producers to lower their markups, it will show up via this final term.8 This basic decomposition can be extended to include more than two groups of countries. For K country groups (1…K), we have: lnPt−lnPt−1=∑K(ΔStk(ln Ptkln PtADV)+St−1k(πtk−πtADV))+πtADV (4) Our first set of panel regression models the lagged share of EME imports and the change in the EME import share for each of our three groups of countries—China, New EU Member States, and other Low Wage countries. The coefficient on the lagged share of EME imports will pick up the inflation effect. The coefficient on the change in EME import share picks up the combined effect of the switching and competition effects, which we collectively refer to as the ‘price level effect’. A subsequent set of regressions is used to split this up into its two components. This first set of panel regressions takes the following form: πit=α+β1Sit−1CHINA+β2ΔSitCHINA+γ1Sit−1NMS+γ2ΔSitNMS+ϕ1Sit−1LWC+ϕ2ΔSitLWC+θexchit+μi+λt+ψ'Xit+εit (5) where the dependent variable, πit is log difference in the sterling value of unit values of imports in each period t and for each industry i. Sit−1 denotes the market share of each EME group—China, EU New Member States (NMS) and other low wage countries (LWC). We also include an exchange rate term, exch, defined as the log difference of an industry-specific exchange rate index. Adding this term allows us to capture the average rate of exchange rate pass-through to import prices. Importantly, we remain agnostic about whether exporters use local or producer currency pricing, and estimating this equation at industry levels allows this pricing behaviour to vary according to goods type.9 As is standard in a panel regression, we add time and industry fixed effects, but we also follow Kapetanios et al. (2011) by including the within-period averages of each of the regressors and the dependent variables (by 4-digit industry), denoted in the equation in matrix form as ψ'Xit. Econometrically, these terms allow common correlated effects that are not picked up by other terms in our equation. For example, an industry-specific positive productivity shock that hits domestic producers in manufacturing may induce foreign importers to alter the price of imports. The equation is estimated by ordinary least squares (OLS). Auer and Fischer (2010), who investigate the effect of import competition on domestic producer price inflation (PPI), argue that it is necessary to instrument the change in EMEs market share, and failure to do so could substantially bias estimates. That is because any positive demand shock is likely to increase both producer prices and the share of goods imported from EMEs as a percentage of the domestic market. However, the regression equation estimated in this paper is unlikely to suffer from this same endogeneity problem for three reasons. First, our research question embodies a different independent variable—we consider imports as a share of total imports, rather than as a share of domestic production plus imports. A cyclical demand shock that increases the demand for imports of a particular good is likely to increase demand from all countries proportionately. This would leave the former measure of market share unchanged, but not the latter.10 Second, our independent variable is based on imports to the UK—as opposed to domestically produced goods—which is also less likely to be related to cyclical conditions in the domestic economy than domestic producers would be. Third, the impact of any potential demand side factors that could bias our regression coefficient estimates should be mitigated by the inclusion of the 4-digit industry averages of our dependent and explanatory variables, which seek to capture the effects of any industry-specific shocks (including demand shocks). a. Quantifying the ‘price level’ effect A key goal of the analysis is to estimate the overall size of the price-level effect, as opposed to merely providing estimates of the coefficient. We can compute the overall size of the price-level effect using the coefficients from the regression above using the following expression: Price level effect=∑i=1Iwit⋅β^2i⋅ΔSitCHINA (6) where ΔSitCHINA is the actual observed change in market share in industry i, β^2i is the estimated coefficient of the change in China share,11 and wit is the weight of sector i in total imports at time t. b. Quantifying the ‘inflation effect’ Similarly, the inflation effect is picked up the coefficient on the lagged China share in the regressions. The overall size is given by: Inflation effect=∑i=0Iwit⋅β^1i⋅Sit−1CHINA (7) 4. Results 4.1 Heterogeneity across industry The estimation results are presented in Table 1. The first column shows that when we pool over all industries (regression I), the changes in the market share of China, NMS and other LWCs are for the most part both significant and negative, although the size of the coefficient is two to three times larger for China. The lagged market share variables are insignificant, suggesting that higher inflation in EMEs has not fed through to UK import prices. Table 1. Baseline regression results [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 The dependent variable in each regression is the log change in import prices. Note: Coefficients for variables in the vector X are not reported here for space reasons. *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. Standard errors are in parentheses. View Large Table 1. Baseline regression results [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 [I] [II] [III] [V] [V] [VI] [VIII] Sample (1-digit industries) (0–9) (0–4) (5) (6–8) (6) (7) (8) All Food, beverage, fuel, commodities, oils Chemicals Manufacturing Materials Machinery Manufactured articles Lagged China share –0.030 (0.056) –0.015 (0.120) –0.211 (0.177) –0.002 (0.066) –0.036 (0.110) 0.100 (0.168) –0.089 (0.095) Change in China share –0.474*** (0.076) 0.003 (0.117) –0.848*** (0.297) –0.550*** (0.093) –0.393*** (0.136) –0.780*** (0.179) –0.588*** (0.183) Lagged NMS share –0.010 (0.065) –0.151 (0.161) –0.051 (0.180) 0.019 (0.075) –0.011 (0.091) –0.125 (0.127) –0.453** (0.195) Change in NMS share –0.184* (0.094) –0.349* (0.209) 0.054* (0.182) –0.173 (0.117) –0.011 (0.140) –0.300 (0.186) –0.948** (0.421) Lagged other LWC share –0.025 (0.045) 0.114 (0.084) –0.094 (0.101) –0.072 (0.060) –0.011 (0.073) –0.105 (0.091) –0.083 (0.182) Change in other LWC share –0.212*** (0.061) –0.039 (0.138) –0.352** (0.146) –0.264*** (0.080) –0.185** (0.089) –0.646*** (0.181) –0.051 (0.240) Exchange rate –0.282*** (0.105) –0.121 (0.134) –0.076 (0.328) –0.472*** (0.165) –0.230 (0.206) –0.565 (0.377) –0.797** (0.361) N (no. of obs.) 35,351 7,983 5,462 21,902 9,408 7,435 5,059 I (no. of industries) 3,152 724 480 1,947 836 668 443 R2 (overall) 0.278 0.343 0.218 0.273 0.261 0.296 0.257 The dependent variable in each regression is the log change in import prices. Note: Coefficients for variables in the vector X are not reported here for space reasons. *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. Standard errors are in parentheses. View Large However, these results mask considerable variation across industry groups. For food- and commodity-based products (regression II), none of the coefficients on changes in market share is significant. By contrast, in chemicals (regression III) and manufacturing (regression IV), both China and other LWCs do appear to exert a significant downward effect on prices via gaining market share. That said, while the EME group market shares in the chemicals sector have been relatively constant over our sample period, the shares for manufacturing have risen rapidly (see Appendix B). Therefore, and in keeping with most previous studies, we restrict the focus of this paper to the manufacturing sector.12 Splitting the manufacturing sector into its three separate 1-digit industries also reveals considerable differences. For machinery, our estimates suggest that ceteris paribus a 1% rise in Chinese market share is associated with a fall in prices of 0.82% (regression VII), compared to a fall of only 0.47% in materials (regression VI). The downward pressure exerted by NMS is only significant in manufactured articles, whereas other LWCs are significant in materials and machinery. Again, the lagged market share for all EME groups is generally insignificant. This analysis at finer levels of disaggregation raises the important question of what is the appropriate level of disaggregation. To explore this, we estimate regressions at the 2-digit level, where we continue to find variation across the 26 industry groups (see Appendix D).13 As Table 2 shows, when we test the implied restriction of pooling across 1- and 2-digit industries, we find a clear rejection of the hypothesis of equal coefficients at either the 1- or the 2-digit industry levels. Table 2. Tests for pooling Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 View Large Table 2. Tests for pooling Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 Restricted model Unrestricted model p-value Full pooling Separate regressions for each 1-digit industry 0.000 Full pooling Separate regressions for each 2-digit industry 0.000 Separate regressions for each 1-digit industry Separate regressions for each 2-digit industry 0.000 View Large We compute the estimated total China price level effect under three different specifications using the methodology set out in equation (4). The results from this exercise are shown in Figure 3. It shows that the estimated China effect is much larger when the equations are estimated at the 2-digit industry level. In other words, failure to account for coefficient heterogeneity reduces the estimated size of the China effect by around a third. Fig. 3. View largeDownload slide Price-level effects of China under different pooling assumptions Fig. 3. View largeDownload slide Price-level effects of China under different pooling assumptions Given both the economic and statistical significance of coefficient heterogeneity, our preferred specification is to estimate separate regressions for each 2-digit industry. This baseline suggests that the tailwind from China is around –0.72pp per annum over our sample period. Since manufacturing accounts for around two-thirds of all UK imports, this is equivalent to a stand-alone effect on all import prices of around –0.49pp. Looking at the profile of this effect over time, there is no obvious sign of a trend, suggesting that the price-level effect of China has not waned over time. To check for sensitivity of our results to different specifications, we performed a variety of robustness checks. First we computed the China price-level effect for alternative specifications. Given that the NMS and LWC market share variables were insignificant, we dropped them to see how they would change the specification. There was little difference (see Appendix E). To check for autocorrelation, we included a lagged dependent variable. And to check for the importance of common correlated effects, we also dropped the 4-digit industry averages of all variables. But our estimated China price-level effect was very similar to our baseline case. We also checked if the results were driven by insignificant coefficients—that is, we repeated the calculations by re-coding the 12 insignificant industry coefficients (out of 26) to zero. They produce a very similar estimate of the China price level effect. 4.2 Computing confidence intervals Whilst the regression coefficients on the China share are statistically significant for most industries, the above estimate of the total China effect does not give any indication as to whether the estimated effect of China and other EMEs is statistically significantly different from zero. To assess this, we used Monte Carlo methods to estimate a confidence interval. Specifically, we take 10,000 draws from the estimated parameter distribution of each βi, and use each draw to compute the China effect based on the values drawn. From this, we compute the 95% confidence interval by discarding the top and bottom 2.5% of the distribution. This confidence interval is shown in Figure 4.14 As a robustness check, we also constructed a confidence interval using a bootstrapping technique based on re-sampling residuals across industry, which unlike the Monte Carlo approach allows for any correlation in residuals across equations.15 This yields very similar results. Fig. 4. View largeDownload slide China price-level effect Fig. 4. View largeDownload slide China price-level effect Over the full sample period, 1999–2011, the 95% confidence interval in Figure 4 never crosses the y-axis, and hence we conclude that China’s impact is significant at the 5% level in each year. A similar exercise for estimating the mean impact and confidence intervals for LWC and NMS shows that their impacts are not statistically different from zero.16 Even though initial regressions suggested that the coefficient on the change in the market share of LWC and NMS countries was significant for some manufacturing industries, when the effect is computed over all manufactured import prices, it is both small and statistically insignificant. For the former group, this likely reflects the twin facts that the growth in import market share was much less strong than for China (Figure 2b) and that the estimated size of the fall inflation for a given gain in market share was much lower for these country groups (Table 1) than for China. The mean inflation effect and 95% confidence intervals are shown in Figure 5. They straddle the zero line, implying that rising inflation in China is not having any statistically significant effect on UK import prices. A similar exercise was carried out to compute the inflation effect of NMS and LWC; the confidence intervals for the inflation channel in these country groups were also very wide and no different from zero (see the figures in Appendix F). Fig. 5. View largeDownload slide Estimates of the China inflation effect Fig. 5. View largeDownload slide Estimates of the China inflation effect 4.3 Decomposing the price-level effect into the switching and competition effects The competition effect captures the response of non-Chinese producers to a change in China’s market share. To isolate this effect, we run a second set of regression where the dependent variable is the log change in import prices (or the unit value of inflation) from all countries excluding China.17 πitEXC=α+β1'Sit−1CHINA+β2'ΔSitCHINA+γ1'Sit−1NMS+γ2'ΔSitNMS+ϕ1'Sit−1LWC+ϕ2'ΔSitLWC+θ'exchit+μ'i+λ't+ψ'Xit+εit (8) The coefficient β'2 captures the response of non-Chinese producers to a 1pp rise in China’s market share. If this coefficient is not statistically different from zero, this indicates no significant pricing response to China gaining market share; if this is significant, it indicates that other producers do respond to Chinese entry. The total size of the competition effect is given by:18 Competition effect=∑i=1Iwit⋅β^'2i⋅ΔSitCHINA⋅(1−SitCHINA) (9) The switching effect is then calculated as the gap between the point estimates of the total price-level effect and the competition effect. Figure 6 below shows the resulting decomposition. Of course, since both point estimates come from separate models, they are each subject to a standard error, but our methods do not allow us to compute a standard error for the difference between them, since we cannot estimate the covariance of the estimates over the two sets of errors—and hence we only have a point estimate of the switching effect. Our results nevertheless point to a qualitatively important role for competition effects. Fig. 6. View largeDownload slide Decomposing the China combined price-level effect Fig. 6. View largeDownload slide Decomposing the China combined price-level effect It suggests that the switching effect accounts for about two-thirds of the total price level effect, with the remaining third attributed to the competition effect. 5. Conclusions In this paper, we quantify the effect of rising import penetration from emerging market economies on UK import prices using a rich panel dataset. This highly disaggregated industry data allow us to account for heterogeneity across industries and across the emerging market economies that export to the UK. We find robust evidence that the rise in China’s share of the markets has acted as a tailwind, lowering manufacturing import price inflation by an estimated 0.7pp on average a year over the period 1999–2011; this is equivalent to a standalone effect of –0.5pp on overall import prices. We find no evidence of a statistically significant effect from the other EME country groups. Constructing the confidence interval of the tailwind by Monte Carlo methods, we find that the China tailwind is indeed statistically significant, but there is no evidence of a significant tailwind from the EU New Member States or other low wage cost economies (including India and Brazil). Finally, this paper finds that around two-thirds of the China tailwind comes via a change in the composition of the import basket, to reflect a greater share of cheaper goods from China. The remaining one-third arises via competition effects, as non-Chinese exporters to the UK lower their prices in response to the increased competition from China. There is no significant impact of the rising inflation rate in China during the latter years of our sample period. Footnotes 1 Auer and Fischer (2010) find that the distribution of price shocks has a strong negative skew, which may interact with menu costs to generate downward pressure on aggregate prices even if the mean of price shocks is zero. 2 For example, ‘Corks and Stoppers of Natural Cork’ has the 5-digit code ‘63311’. Its 1-digit industry (6) is ‘Manufactured articles classified chiefly by material’; its 2-digit industry (63) is ‘Cork and wood manufactures’; its 3-digit industry (633) is ‘Cork manufactures’; and its 4-digit industry (6331) is ‘Articles of natural cork’. 3 The 45 countries are: all EU and OECD members, plus Argentina, Brazil, China, Hong Kong, India, Indonesia, Malaysia, Pakistan, Philippines, Qatar, Russia, Saudi Arabia, South Korea, Taiwan, Thailand and Vietnam. Collectively they account for over 90% of UK imports. The ratio of imports from these countries to total imports from all trading partners is broadly constant over the sample period, and so our country choice does not result in the exclusion of groups of countries which have also seen a significant rise in their overall market share. 4 The New EU Member States (NMS) group of countries includes Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, Slovenia and Romania. 5 All these countries have lower relative price levels than the UK’s main advanced economy trading partners (see Appendix D). 6 Throughout the paper we use the term ‘total’ to refer to summing across the 45 major countries the UK imports from. 7 One could also write the decomposition as: ln Pt−lnPt−1=ΔSt(ln PtEMEln PtADV)+St−1(πtEME)+(1−St)πtADV, but we use an alternative specification as it facilitates discussion of the three channels. 8 Another possible channel is if EME competition drives out the most expensive advanced producers, leaving cheaper advanced economy producers in the market, then the measured πtADV will be lower due to the shift in composition of advanced economy exporters. 9 We use the contemporaneous rather than lagged exchange rate because the literature (e.g. Gopinath et al., 2010) finds that pass-through from the exchange rate to import prices usually takes place within one year. That said, in Section 4 we investigate how sensitive our results are to different assumptions and find that they are robust to additional lags of the exchange rate. 10 This shock, with a proportional increase in demand for imports from all countries, would result in a rise in the ratio of EME or Chinese imports to the overall size of the domestic market. 11 In what follows, we also run regressions into samples split by one- and two-digit industry. That means our estimated coefficient used in equation (4) can differ across industry groups. We therefore include an i subscript on β̂ in equation (4). 12 As a further check, we also repeated our analysis for the non-manufacturing industries. We find that the price-level effect on inflation each year was +0.04pp for China, +0.08pp per year for New Member States and -0.02pp per year for Low Wage Countries. We therefore conclude that none of the EME groups exerted a meaningful effect on non-manufacturing import prices over the sample period. 13 If we go down to the 3-digit level, we run into the problem that some 3-digit industries contain only a single 5-digit industry and hence panel estimation cannot be used. 14 The mean estimate from this exercise is almost identical to our estimate of the tailwind reported in Figure 3. 15 Specifically, we decomposed the data into fitted values and residuals. Taking the residuals by year, we obtain 11 sets of residuals, which formed our sampling population. We then generated a synthetic dataset by adding the fitted values for each year to a randomly chosen residual vector (sampling with replacement). We then calculated the estimate China effect from this synthetic dataset, and repeated the whole procedure 10,000 times. The 95% confidence interval was then given by discarding the upper and lower 2.5% of estimates. See the appendix for charts. 16 See Figures D2 and D3 in Appendix D. 17 This is calculated as the aggregate value of non-Chinese imports in industry i at time t to the aggregate volume of non-Chinese imports in industry i at time t. 18 We multiply by ( (1−SitCHINA), because we estimate our competition effect only over non-Chinese imports, but wish to calculate the effect on the price of all imports in a given industry. Bibliography Auer , R. , Degen , K. and Fischer , A . 2011 . Low-wage import competition, inflationary pressure and industry dynamics in Europe , European Economic Review , vol. 59 , no. C , 141 – 66 Auer , R. and Fischer , A . 2010 . The effect of low-wage import competition on U.S. inflationary pressure , Journal of Monetary Economics , vol. 57 , 491 – 503 Google Scholar CrossRef Search ADS Autor , D. , Dorn , D. , Hanson , H. and Song , J . 2013 . The China syndrome: local labour market effects of import competition in the United States , American Economic Review , vol. 103 , no. 6 , 2121 – 68 Google Scholar CrossRef Search ADS Ball , L . 2006 . ‘ Has Globalization Changed Inflation ?’, NBER Working Papers 12687 Bean , C . 2006 . Globalisation and inflation , Bank of England Quarterly Bulletin , Q4 , 468 – 75 Borrio , C. and Filardo , A . 2006 . ‘ Globalisation and inflation: New Cross Country Evidence on the Global Determinants of Domestic Inflation ’, BIS Working Paper 227 Broda , C. and Romalis , J . 2009 . ‘The Welfare Implications of Rising Price Dispersion’ , mimeo , Chicago Booth University Bugamelli , M. , Fabiani , S. and Sette , E . 2010 . ‘ The Pro-Competitive Effect of Imports from China: An Analysis of Firm-Level Price Data ’, Bank of Italy , Working Paper No. 737 Google Scholar CrossRef Search ADS Calza , A . 2009 . Globalization, domestic inflation and global output gaps: evidence from the Euro area , International Finance , vol. 12 , no. 3 , 301 – 20 Google Scholar CrossRef Search ADS Dale , S . 2011 . ‘ MPC in the Dock ’, speech at the National Asset-Liability Management Global Conference , 24 March Dudley , W . 2011 . ‘US Economic Policy in a Global Context’ , Remarks at the Foreign Policy Association Corporate Dinner , New York , http://www.newyorkfed.org/newsevents/speeches/2011/dud110607.html European Central Bank . 2008 . Globalisation, trade and the Euro area macroeconomy , Monthly Bulletin , January, 75–88 Feenstra , R . 2007 . ‘ Globalization and Its Impact on Labour ’, Vienna Institute for International Economic Studies Working Paper No. 44 Feenstra , R. and Kendall , D . 1997 . Pass-through of exchange rates and purchasing power parity , Journal of International Economics , vol. 43 , 237 – 61 Google Scholar CrossRef Search ADS Feyzioğlu , T. and Willard , L . 2006 . ‘ Does Inflation in China Affect the United States and Japan ?’, IMF Working Paper No. 06/36 Freeman , R . 2005 . Are your wages set in Beijing ?, Journal of Economic Perspectives , vol. 9 , 15 – 32 Google Scholar CrossRef Search ADS Gopinath , G. , Itskohiki , O. and Rigobon , R . 2010 . Currency choice and exchange rate pass-through , American Economic Review , vol. 100 , 304 – 36 Google Scholar CrossRef Search ADS Greenspan , A . 2005 . ‘Globalization: Remarks to the Council on Foreign Relations’ , 10 March, http://www.federalreserve.gov/boarddocs/speeches/2005/20050310/default.htm IMF . 2012 . ‘World Economic Outlook’ Kamin , S. , Marazzi , M. and Schindler , J . 2008 . The impact of Chinese exports on global import prices , Review of International Economics , vol. 14 , no. 2 , 179 – 201 Google Scholar CrossRef Search ADS Kapetanios , G. , Pesaran , H. and Yamagata , T . 2011 . Panels with non-stationary multifactor error structures , Journal of Econometrics , vol. 160 , 326 – 48 Google Scholar CrossRef Search ADS Koske , I. , Pain , N. and Sollie , M . 2008 . Globalisation and OECD consumer price inflation , OECD Economic Studies , vol. 44 Li , H. , Li , L. , Wu , B. and Xiong , Y . 2012 . The end of cheap Chinese labour , Journal of Economic Perspectives , vol. 26 , no. 4 , 57 – 74 Google Scholar CrossRef Search ADS McCoille , C . 2008 . The impact of low-cost economies on UK import prices , Bank of England Quarterly Bulletin , Q1 , 58 – 63 Melitz , J. and Ottaviani , P . 2008 . Market size, trade, and productivity , Review of Economic Studies , vol. 75 , no. 1 , 295 – 316 Google Scholar CrossRef Search ADS Millard , S. and Lipinska , A . 2012 . Tailwinds and headwinds: how does growth in the BRICs affect inflation in the G-7 ?, International Journal of Central Banking , vol. 8 , 227 – 66 Mishkin , F . 2009 . Globalization, macroeconomic performance and monetary policy , Journal of Money, Credit and Banking , vol. 41 , no. S1 , 187 – 96 Google Scholar CrossRef Search ADS Rogoff , K . 2006 . ‘Impact of Globalization on Monetary Policy’, paper prepared for symposium on ‘The New Economic Geography: Effects and Policy Implications’ , http://kansascityfed.org/publicat/sympos/2006/pdf/19Rogoff.pdf Wheeler , T . 2008 . ‘ Has Trade with China Affected UK Inflation ?’, External MPC Unit Discussion Paper 22, Bank of England Appendix A. Data Exchange rate data Let V denote the value of imports of country j in industry i at time t. The weight of each country is given by: wijt=Vijt∑j=0IVijt The exchange rate, e, is the average annual bilateral nominal exchange rate, extracted from the Thompson datastream. Our index of exchange rate changes is defined as: exchit=∑j=0Iwijt(ejtejt−1−1) Exchange rates are defined in the European style, so a positive value of exch corresponds to an appreciation in sterling. For countries which adopted the euro during our sample period, we use the exchange rate between sterling and the legacy country, which during the post-euro adoption period is calculated by multiplying the official conversion rate with the sterling euro exchange rate. The figure below shows the mean, maximum, minimum and inter-quartile range of the exchange rate across all 5-digit industries in each year of the sample period. Global wage and inflationary pressure indices Let p denote the Consumer Price Index as measured as the annual rate of CPI inflation, as reported in the IMF’s World Economic Outlook. The global inflationary pressure index is given by: Infit=∑j=0Iwijt(pjtpjt−1−1) Similarly, denoting nominal wages with W, the global wage index is given by: Wagesit=∑j=0Iwijt(WjtWjt−1−1) Market share of low-cost producers We calculate the market share of a subset of K countries, as follows: Sikt=∑j∈KVijt∑j∈JVijt For the three variables, K is defined as follows: China: China New Member States (NMS): Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, Slovenia and Romania Low Wage Cost (LWC): Brazil, Mexico, Russia, Turkey, Thailand, Indonesia, India, Philippines, Pakistan and Vietnam Appendix B. Decomposing the growth in each group’s market share by industry group View largeDownload slide View largeDownload slide Appendix C. Comparison with ONS data Our analysis of import prices is based on unit-value data, which does not adjust for product varieties or quality differences of goods within each 5-digit industry. This might be particularly relevant when looking at EMEs, as Broda and Romalis (2009) show Chinese imports tend to be concentrated in lower-quality varieties of the same product class. To check for this source of bias in our results, we compare results obtained from our data with official import price indices (quality adjusted) published by the Office for National Statistics. But since the ONS data are only available for a selection of 2-digit industries, a side-by-side comparison of regression results for all available industries is presented in Table C1. When the regression is run over all sectors, the ONS data doesn’t yield a significant coefficient on the change in China’s market share, or that of NMS; but the LWC share is significant, albeit with the ‘wrong’ sign. However, the ONS import price indices predominantly cover non-manufacturing industries. Restricting the sample to manufacturing industries, we find that the coefficient on the change in China’s market share is very similar for both measure of import price inflation. The change in NMS share is insignificant in both, and the change in LWC share is significant only when unit-value data is used, which may reflect the lack of quality adjustment in these economies. Table C1. HMRC vs. ONS data [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 View Large Table C1. HMRC vs. ONS data [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 [I] [II] [III] [IV] Dependent variable Unit value (HMRC) Import prices (ONS) Unit value (HMRC) Import prices (ONS) Industry level All available All available Manufacturing Manufacturing Lagged China share –0.030 (0.056) 0.025 (0.104) –0.002 (0.066) 0.051 (0.171) Change in China share –0.474*** (0.076) –0.071 (0.231) –0.550*** (0.093) –0.714*** (0.160) Lagged NMS share –0.010 (0.065) –0.079 (0.139) 0.019 (0.075) –0.457 (0.020) Change in NMS share –0.184* (0.094) 0.115 (0.252) –0.173 (0.117) 0.232 (0.300) Lagged LWC share –0.025 (0.045) –0.080 (0.139) –0.072 (0.060) 0.553*** (0.119) Change in LWC share –0.212*** (0.061) 0.115 (0.509) –0.264*** (0.080) 0.241 (0.154) Exchange Rate –0.282*** (0.105) –0.170 (0.290) –0.472*** (0.165) –0.859*** (0.773) N (no. of obs.) 35,351 564 21,902 144 I (no. of industries) 3,152 50 1947 12 R2 (overall) 0.278 0.052 0.273 0.081 View Large Appendix D. Regression coefficients at the 2-digit level 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 N.E.S: ‘not elsewhere specified’; *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. View Large 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 2-digit SITC Code Industry description Coefficient on lagged level of China share Coefficient on change in China share N Average share of manuf imports, % Average annual gain in China market share, pp 61 Leather, leather manufactures, N.E.S., and dressed fur skins –0.448 –1.442** 246 0.1 0.8 62 Rubber manufactures, N.E.S. 0.083 –0.525 371 1.2 1.3 63 Cork and wood manufactures, other than furniture 0.305 –0.741** 363 0.9 1.6 64 Paper, paperboard and articles of paper pulp, paper or paper board 0.040 0.498 805 3.0 0.3 65 Textile yarn, fabrics, made-up articles, N.E.S., and related products 0.053 –0.541** 2,645 2.4 1.2 66 Non-metallic mineral manufactures, N.E.S. 0.063 –0.057 1,133 2.8 0.7 67 Iron and steel –0.157 –0.521 1,685 2.4 0.4 68 Copper –0.244 0.088 810 2.7 0.3 69 Manufactures of metals, N.E.S. 0.082 –0.841 1,411 3.2 1.2 71 Power generating machinery and equipment –0.569 –4.279** 512 5.1 0.0 72 Machinery specialised for particular industries 1.241** –0.551 1,387 2.7 0.4 73 Metalworking machinery –0.147 –1.164*** 829 0.6 0.4 74 General industrial machinery and equipment, N.E.S., and machine parts N.E.S. 0.041 –0.553* 1,805 5.1 0.6 75 Office machines and automatic data processing machines –0.852 –2.216** 308 8.8 1.9 76 Telecommunications, sound and recording and reproducing apparatus and equipment –0.070 –1.479*** 408 8.2 1.4 77 Electrical machinery, apparatus and appliances, N.E.S., and electrical parts thereof ... 0.575** –0.660** 1,515 8.7 1.1 78 Road vehicles (including air-cushion vehicles) 0.113 –1.326*** 475 17.7 0.1 79 Transport equipment, N.E.S. 1.595* 1.656*** 262 0.5 0.2 81 Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fittings –0.098 –0.208 204 0.8 1.4 82 Furniture and parts thereof; bedding mattresses, mattress supports, cushions and similar stuffed furnishings –0.075 –1.427* 276 2.1 2.3 83 Travel goods, handbags and similar containers –0.472 –4.887 108 0.6 1.0 84 Articles of apparel and clothing accessories –0.520* 0.147 1,144 5.6 2.0 85 Footwear –0.137 –0.928** 218 1.5 2.3 87 Professional, scientific and controlling instruments and apparatus –0.075 –1.931*** 775 3.2 0.3 88 Photographic apparatus, equipment –0.446 –2.247*** 715 1.3 0.3 89 Miscellaneous manufactured articles, N.E.S. –0.349 –1.069 1,666 0.1 0.9 N.E.S: ‘not elsewhere specified’; *, **, *** denote significance at the 10%, 5% and 1% levels, respectively. View Large Appendix E. Robustness checks Fig. E1. View largeDownload slide China market share effect under different specifications Fig. E1. View largeDownload slide China market share effect under different specifications Fig. E2. View largeDownload slide China price-level effect Fig. E2. View largeDownload slide China price-level effect Appendix F. Additional charts Fig. F1. View largeDownload slide Relative price levels Fig. F1. View largeDownload slide Relative price levels Fig. F2. View largeDownload slide The combined ‘price-level’ effect and 95% confidence interval Fig. F2. View largeDownload slide The combined ‘price-level’ effect and 95% confidence interval Fig. F3. View largeDownload slide The ‘inflation’ effect and 95% confidence interval Fig. F3. View largeDownload slide The ‘inflation’ effect and 95% confidence interval © The Author(s) 2017. Published by Oxford University Press on behalf of the Cambridge Political Economy Society. All rights reserved.

Journal

Cambridge Journal of EconomicsOxford University Press

Published: Dec 28, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off