Brexit and its Impact on Regional Competitiveness in the UK:

(1)

Brexit and its Impact on Regional Competitiveness

in the UK:

How a Depreciated Sterling Affects Value-Added

Real Effective Exchange Rates on a Regional Scale

Author: Supervisor: Second Assessor:

Markus KETTERER Prof. Bart LOS Prof. Sebastian VOLLMER

S3211983 Faculty of Economics Centre for Modern

and Business Indian Studies

m.ketterer90@gmail.com b.los@rug.nl svollmer@uni-goettingen.de Paterswoldseweg 166b Nettelbosje 2 Waldweg 26

9727 BM Groningen 9747 AE Groningen, 37073 Göttingen,

Netherlands Netherlands Germany

A thesis submitted for the degrees of

Master of Science / Master of Arts

(2)

Abstract

In June 2016 the Brexit referendum sent shockwaves around the globe. Though the exact consequences of Brexit are still unknown, researchers predict the UK’s economy to suffer when leaving the EU. Yet, so far there exists no comprehensive assessment on how the sterling’s depreciation after the Brexit referendum affects the international competitiveness of British regions. With cross-regional inequality in the UK already high and on the rise, this paper finds Brexit to further fuel this divergence. This is because changes in competitiveness of comparably richer regions are more pronounced which leaves them even better off compared to poorer regions. (99 words)

(3)

List of Figures

1 Regional British Dependence on International Trade . . . 2

2 Share of British Regions’ Total Gross Goods Exports Going to EU . . . 5

3 A Simple Global Value Chain . . . 7

4 Simplified Regional World Input-Output Table . . . 11

5 Relationship Between UK Trade Weights and Value-Added REERs. . . 28

6 Relationship Between British Regional Income and Value-Added REER . . . . 29

A.1 Imported Intermediates as Share of Total Regional Output in the UK . . . 40

A.2 Trade Balance of UK Regions with EU (£m, 2015 Prices) . . . 41

A.3 Relationship Between Southern Regional Income in the UK and Value-Added REER . . . 42

List of Tables

1 Sterling Depreciation Following Brexit Referendum . . . 21

2 British Intermediate Deliveries in 2010 . . . 22

3 Value-Added REER Weights for UK Regions . . . 24

4 Value-Added REERs for UK Regions . . . 26

5 Value-Added REERs for UK Regions, Equal Elasticities . . . 30

(5)

1 INTRODUCTION 1

1 Introduction

On June 23, 2016, after months of intense political debates which have divided the nation, the British people voted to leave the European Union. As a consequence, the sterling dipped dramatically against other currencies as markets see lower growth potentials for the UK when it leaves the EU. This is bolstered by scholars who emphasize the downside risks of Brexit and state that it will leave the UK worse off (e.g. Dhingra, Ottaviano, Sampson & Van Reenen, 2016; Corsetti & Müller, 2016). On the other hand, the Leave campaign stresses its poten-tial benefits. Dominiczak (2016) states that being a member of the EU undermines the UK’s competitiveness. What is more, the depreciation in the sterling in the aftermath of the Brexit referendum is considered to increase British competitiveness as it leaves British products more competitive on international markets (Bootle & Mills,2016). Depending on the exact outcome of Brexit negotiations, the sterling might even depreciate further against other currencies which is why an assessment of the consequences of this depreciations is important.

In 2015, more than 50% of UK imports originated from the EU, whereas 44% of British exports went to other EU-countries (Office for National Statistics, 2016). This makes the EU the most important trading partner of the UK. Moreover, it was found that the EU-dependence of most British regions increased between 2000 and 2010 but that British regions differ sub-stantially in the extent that they depend on the EU as trading partner (Los, McCann, Springford & Thissen,2017). Figure1adds to this finding by depicting how the use of imported intermedi-ates differs in British regions (panel a). For instance, regional gross output in Eastern Scotland embodies three times less imported inputs than that of Leicesterhire. At the same time, panel b in Figure1reveals a large variation in the extent British regions depend on foreign countries in selling their output. This shows that that the economic dispositions of the different regions differ considerably, which has led to the contention of a North-South divide. In consequence, the UK exhibits one of the highest cross-regional inequalities compared to other EU nations (Charron, 2016). Northern regions, notably Northern Ireland, North England and Wales are economically less developed than Southern regions. These differences suggest that some re-gions will be more affected by Brexit than others. Therefore, this paper sets out to examine how Brexit, and more specifically, the depreciation in the sterling that has happened in the aftermath of the Brexit referendum, affects the competitiveness of British regions on NUTS-2 level which subdivides the UK into 37 regions.

(6)

1 INTRODUCTION 2 Figure 1: Regional British Dependence on International Trade

(a)Imported Intermediates as Share of Total Regional Gross Output

(b)Gross Exports as Share of Total Regional Gross Output

Source: Own computations based onThissen, Lankhuizen & Los(2017).

trade: vertical specialization. Vertical specialization has allowed countries to specialize on specific production steps within value chains and thereby increased the need for and use of imported intermediates (Baldwin & Evenett, 2015). Existing REER measures are unable to account for this pattern as they assume each country to produce one good which is only used as a final good. Thus, these measures cannot differentiate between domestic value added and imported intermediates embodied in a product (Bennett & Zarnic, 2009). As a consequence, Bems & Johnson(2017) have developed a new REER measure – the value-added REER – that corrects for the impact of vertical specialization.

(7)

2 LITERATURE REVIEW 3

non-tariff barriers. This paper finds that a depreciating sterling benefits all British regions because it raises their international competitiveness. Yet, the effects on regional value-added REERs are rather ambiguous. Consistent with existing literature, this paper finds regions that trade their value added relatively more with foreign countries to be more positively affected by Brexit. Finally, this paper paints a dismal picture for most of the already disadvantaged Northern regions in the UK which are found to fall further behind the richer Southern regions in terms of competitiveness. Especially Scottish regions can hardly benefit from the sterling’s depreciation.

This paper’s contribution to existing literature is twofold. First, it adds to the literature that focuses on the effects of Brexit on the British economy. Second, with researchers classifying the UK as a highly heterogeneous country, the paper adds to recent research that finds British regions to have further drifted apart. The remainder of this paper is structured as follows. The next section provides a short literature review on exchange rate assessments and cross-regional inequality in the UK. Section3gives an overview on data and methodology used in this paper. In Section4, the results are presented while Section5presents a conclusion.

2 Literature Review

2.1 Inequality and Economic Heterogeneity of UK Regions

There is ample literature examining the evolution of inequalities in the UK most of which focuses on the development of wage inequality in the UK. Bell & Van Reenen (2014) and Machin (2011) are able to show a substantial increase in wage inequality across the UK as of the late 1970s. Building on the results of Goos & Manning(2007) who find evidence for an increase in job polarization in the UK,Jones & Green(2009) are able to disentangle the in-creasing employment share in low- and high-wage jobs for different UK regions. They find that high quality jobs – characterized by a high median wage – are primarily created in regions such as London and the South-Eastern part of England that already display the highest job quality. Thus, regional inequalities have risen over time. Literature investigating regional inequalities in the UK typically emphasizes the large between-regional inequality (e.g. Blackaby & Manning, 1990; Charron, 2016)1. This partition is notably nurtured by the North-South divide in the UK (Blackaby & Manning, 1990) pointing out the fact that Northern regions including North-ern, Ireland, North England, Wales, and parts of the Midlands are economically disadvantaged against Southern regions like London, Southern and Eastern England. The North-South divide has more recently been bolstered byMcCann(2016).

1_{More recently, studies also found substantial within-regional inequalities (e.g.}_Dickey_,₂₀₀₇_;_{Lee, Sissons &}

(8)

As early as in the 1990s, Blackaby & Manning(1990) examined the alleged North-South divide in the UK for the period between 1975 and 1982 and found earnings to vary substantially across regions. Because the authors use micro-level data for their study on wage differentials, they are able to attribute some of the regional variation to the attributes of regional labor forces. Duranton & Monastiriotis(2002), who focus on the time period between 1982 and 1997, also find support for the British North-South divide and state that regions have even diverged over time as average earnings per region diverged. Rice & Venables (2003) contend that regional differences measured by GDP per worker are not only persistent over time, they also find them to be larger than in other EU countries or the United States. Similar toBlackaby & Manning (1990) they find the characteristics of the labor force to be a main driver of that pattern. Most recently,Charron(2016) has investigated regional wealth inequality in EU countries focusing on the sample period from 1995 to 2008. He uses the Gini-Index as a measure of regional wealth dispersion and finds that along with Ireland, the UK was the sole EU-152 country that went through a significant regional divergence. Overall, Hills et al.(2010) find regional wage and wealth inequalities in the UK to be higher than in almost all other OECD countries.

In a study published by theSheffield Political Economy Research Institute (SPERI)(2016), the authors focus on regions within the UK and their differing trade relationships with the EU to assess how Brexit will influence the UK and which region(s) will be influenced the most. They state that the substantial disparities in the economic composition of British regions not only cause regional inequalities but also lead to different trade relationships of British regions with the EU. Figure2reveals that in 2015, Northern Ireland, the North East and the South West of England were more dependent on goods exports to the EU than any other region3. Export dependency on the EU is higher than that on the rest of the world for all of these three regions. According to the Sheffield Political Economy Research Institute (SPERI) (2016), Northern Ireland, the North East and the South West are the UK’s poorest region, England’s poorest region, and Southern England’s poorest region, respectively. The authors find these already disadvantaged regions unlikely to benefit from Brexit as it will probably restrict trade with the EU which these regions are highly dependent on. Some of the UK’s richest regions such as London or the South East depend much less on goods exports to the EU, thus, making them less prone to higher trade restrictions with the EU. However, service exports are not considered in the analysis of theSheffield Political Economy Research Institute (SPERI)(2016). If Brexit were to restrict trade in services, London and the South East might not be that much less af-fected as other regions as Figure2suggests, because the service industry is highly concentrated in these regions.

2_{The EU-15 comprises the following countries: Austria, Belgium, Denmark, Finland, France, Germany,}

Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, United Kingdom.

3_{The study of the}_{Sheffield Political Economy Research Institute (SPERI)}₍₂₀₁₆_{) does only account for trade}

(9)

2 LITERATURE REVIEW 5 Figure 2: Share of British Regions’ Total Gross Goods Exports Going to EU

Source:Sheffield Political Economy Research Institute (SPERI)(2016)

In addition to research on regional wage inequalities, there have already been attempts to assess regional competitiveness in the UK (Huggins,2003;Huggins & Thompson,2016). Hug-gins(2003) establishes a new single index to gauge regional competitiveness. He uses data that are supposed to mirror the relationship between performance on a macro-economic level and innovativeness in business conduct. Therefore, his competitiveness index is rather broad as it captures six parameters: GDP per capita, average earnings, business density, knowledge-based business, economic activity rates, and unemployment. He finds his index to vary considerably across regions and his results reveal once more the persistence of the North-South divide. More precisely, dividing the UK into its regions, he finds London, the South East and East to perform above UK average and all other regions to perform below this average. Using a refined version of the competitiveness index created byHuggins(2003), Huggins & Thompson(2016) assess regional competitiveness in the UK on NUTS-3 level4for the year 2016. Confirming assertions that the city of London has detached itself from the rest of the country (McCann, 2016), they find nine out of the ten most competitive localities to be located in the London region. More-over, in line with the results of Huggins(2003), they find the Southern regions, i.e. London, South East, East, and South West together with the West Midlands to be the most competitive regions in the UK.

4_{NUTS = Nomenclature of Territorial Units for Statistics. For an exact listing of NUTS-3 regions in the UK}

(10)

The large regional discrepancies in the UK make it very likely that the sterling’s depre-ciation following the Brexit referendum impacts regions differently. This is why this paper assesses movements in regional instead of national competitiveness, as the latter would conceal that some regions benefit more than others.

2.2 Real Effective Exchange Rates and Competitiveness

In order to assess the value of a nation’s currency against that of any other single currency, nominal exchange rates have been used since the 1940s (Hirsch & Higgins, 1970). Bilat-eral or nominal exchange rates (NER) are typically expressed in "indirect quotation", giving the amount of a foreign currency that is worth one unit of the home currency (Ellis, 2001): S_indirect = units of foreign currency_{unit of home currency} . The other, less frequently used way to compute NERs is the "direct quotation". It gives the amount of domestic currency that has to be paid for one unit of the foreign currency: S_direct = _{unit of foreign currency}units of home currency. Thus, the direct quotation is the inverse of the indirect quotation. NERs are bilateral in nature as they compare only two currencies. It was not until 1970 that Hirsch & Higgins (1970) established the nominal effective exchange rate (NEER). In contrast to bilateral measures, effective exchange rates capture the exchange rate of a given home currency and a basket of multiple other currencies (Hirsch & Higgins, 1970). Each currency contained in that basket is then assigned a weight that is supposed to reflect each country’s importance for the home economy. According toHirsch & Higgins(1970), the weights that sum up to one should give the relative trade share of each country with the home country. Darvas(2012) defines the NEER as

NEER_t=

N

∏

i=1

S(i)_tw(i). (1)

The equation gives the NEER as a geometric average of bilateral exchange rates S(i)t between

the home country and its N trade partners, where the subscript t denotes a given year. Each trade partner is assigned a weight w(i)giving the relative importance of a trading partner for the home country (Ellis, 2001). Thus, home’s bilateral exchange rate with major trading partners weighs relatively more than that with smaller partners. S(i)t is defined in indirect quotation.

(11)

differentials delivers the real effective exchange rate (REER).Darvas(2012) defines this as:

REERt=

NEERt∗CPIt

CPI_t( f oreign),

(2)

where the CPI denotes the consumer price index in a given country. Typically, the REER is computed for a longer time period with a given year t serving as base period in which the REER is set equal to 100. Then, starting from this base period, changes in the REER of a certain country can be tracked over time. An increase in the REER is interpreted as an appreciation, which leaves domestic products less competitive on the international market (Bems & Johnson, 2017). Establishing a measure that directly computes changes in the REER between year t and t+1,Bems & Johnson(2017, p. 6) define the conventional REER as follows:

"Given a set of heterogeneous bilateral relative price changes, the index mea-sures the average change in home versus foreign prices, where the average is constructed to summarize the impact of bilateral price changes on demand for a country’s output."

Thus, the index describes how a country’s international competitiveness in selling its out-put changes in response to relative price changes. Depending on the exact research question, there are different ways to compute REERs (Maciejewski, 1983; Rosensweig, 1987; Chinn, 2006; Schmitz et al., 2012) and scholars face several conceptual choices e.g. how to assign and update trade weights of trading partners included in the index, or how to handle price differentials. Yet, despite a plethora of possible approaches, Patel, Wang & Wei (2014) find that organizations providing REER indices such as the International Monetary Fund (IMF) or the Bank of International Settlements (BIS) make the same, oversimplifying presumptions. For instance,Patel, Wang & Wei(2014) emphasize that conventional REERs assume that countries only trade final products which exclusively embody domestic inputs.

Figure 3: A Simple Global Value Chain

Source: Own depiction based onLos & Timmer(2015).

(12)

inputs, Figure3shows that the discrepancy between domestic value added and the export value – or output – of a country can be substantial (for country B and C). This argument is further substantiated byKlau & Fung(2006) andDi Mauro, Rüffer & Bunda(2008) who find that trade in intermediates biases conventional REERs as conventional REER indices assign too high a weight to countries that are highly engaged in global value chains. Consequently, countries that heavily rely on the import of intermediate goods and that export goods with a rather low level of domestic value added are assigned too high a weight. When computing a conventional REER index for country D in Figure 3, the weight attached to country C is equal to that of E and, thus, too high as it refers to C’s export value instead of its domestic value added. When the depicted goods flow is the only form of international trade in Figure 3, a conventional REER does not assign any weight to A or B as they do not engage in direct trade relations with D. This has caused Timmer, Los, Stehrer & De Vries (2013) to state that vertical specialization makes conventional competitiveness indicators that draw on gross trade figures uninformative. Instead, the value of C’s exports that can be attributed to imported intermediates should be attached to the country of origin of these intermediates. In this simple example this implies that 10 units of C’s export value should be attributed to A and another 10 units to B. Only then a REER index computed for country D would reflect each country’s contribution to the final good consumed in D.Bems & Johnson(2017) introduce a value-added REER able to account for the effects of vertical specialization on REER indices. The methodology proposed byBems & Johnson(2017) is elaborated in section3.2.

(13)

true competitiveness of a country in this scenario. Overall, similar to Timmer et al. (2013), Di Mauro & Forster(2008) state that using conventional indicators to assess a country’s com-petitiveness becomes more difficult because of globalization as it considerably changes export structures. This is in contrast to a world without globalization where production patterns are rather rigid and movements in competitiveness derive from relative price changes across coun-tries (Di Mauro & Forster,2008).

Although several studies have confronted the different ways to compute REERs as a mea-sure for competitiveness (e.g. Marsh & Tokarick, 1996; Ca’Zorzi & Schnatz, 2007), there exists no consensus on the best way (Neary, 2006). Yet, in order to assess the actual changes in competitiveness of British regions induced by the sterling’s dip, it is necessary to apply a REER that is able to capture the effects of globalization. While it assesses changes in demand for value added, the value-added REER is not only able to account for globalization. It does also implicitly take into consideration the production structure of a country and, thus, gives a broader measure for competitiveness.

2.3 REERs and Regional Heterogeneity

Recent literature has considerably contributed to the existing discussion on exchange rates by proposing two updates to the computation of exchange rate indices. These are meant to account for changes in international trade patterns and to capture exchange rate movements on a more disaggregated level. Most notably, scholars have started to compute REERs on a regional instead of a national level5. Clark, Sawyer & Sprinkle (1999) stress that inferring from a national exchange rate index can be misleading when focusing on particular regions. Computing REERs on a regional level is of high importance when considering a country where regions are highly heterogeneous. Existing literature, for instance, focuses on China which is known for its relatively prosperous regions along the seashore and its comparably poor hinter-lands (Ping, 2011; Yan, Li, Lin & Jie, 2016). As has been laid out above, the UK is a very heterogeneous country which makes regional REERs all the more essential. This is further substantiated by the fact that trade relationships vary considerably across regions in the UK, as has been shown in section2.1.

Klau & Fung(2006) find imports and exports to become complementary goods for countries focusing on activities at the final stage of global value chains, also referred to as downstream ac-tivities. In Figure3country C relies on the import of intermediates from B that are subsequently used to produce a final product which is then exported. Linking this finding to the assessment of a depreciation in the sterling provides an essential concern: The textbook prediction of a

5_{Instead of regional REERs,}_Goldberg₍₂₀₀₄_{) proposed to compute industry-specific REERs. This is useful}

(14)

3 DATA SOURCES AND METHODOLOGY 10

depreciated currency making the country which is home to this currency better off by leading to higher exports and lower imports does not necessarily hold true when allowing for vertical specialization. This is because intermediates from foreign countries might not be substituted by domestic inputs in the short run. Today, final products typically embody a substantial amount of imported intermediates and this has to be accounted for when computing REERs. Thus, the short run effects of a depreciating sterling is likely to differ for UK regions in line with each region’s share of intermediate inputs. Those British regions relying heavily on imported intermediates in their production listed to the right of panel a) in Figure1 might not see their competitiveness rise in the short run. This is because the effects of the sterling’s lower value that fosters British final goods exports and, thereby, increases demand for British final goods, is passed on to foreign suppliers of inputs via input linkages (Bems & Johnson, 2017). Hence, those UK regions’ competitiveness expressed in demand for their value added might stagnate or even fall – expressed in an increasing REER.

Thus, evaluating the effect of Brexit on the UK economy demands an adaptation of conven-tional REER indices. First, as highlighted in section2.1, the regional heterogeneity in the UK’s economy leaves some regions more likely to benefit from a depreciation than others which is why the REER index should be computed on a regional level. Second, vertical specialization increases the use and importance of imported inputs and, hence, biases weights used in con-ventional REERs. In order to account for these two findings, the value-added REER developed by Bems & Johnson (2017) is used on a regional level to assess the effect of a depreciation in the sterling subsequent to the Brexit referendum. This paper contends that British regions are discriminately affected by Brexit. More specifically, it is hypothesized that the sterling’s depreciation leads to a further increase of the North-South divide, which means that regions associated with the poorer Northern regions benefit less from the depreciation. With the North-ern regions unable to catch up with the richer SouthNorth-ern regions, heterogeneity in the UK will increase which will fuel contentions of the UK being a divided nation.

3 Data Sources and Methodology

3.1 The Regional World Input-Output Database

(15)

provide an ideal database when focusing on specific regions. With this paper’s aim being the calculation of changes in regional competitiveness in the UK, the WIOD poses the best data source.

The WIOD lists a total of 40 countries and a composite category containing data for the "Rest of World" (ROW) that altogether make up roughly 85% of global GDP (Timmer et al., 2015). Next to present-day EU members – except for Croatia – major economies such as the US and China are included. The regional extension to the WIOD disaggregates data for all major countries to a regional level and gives euro-denoted figures based on market exchange rates (Thissen, Lankhuizen & Los, 2017). Data for the regions listed in the regional WIOD are defined at NUTS-2 level6. Although the disaggregated WIOD does only comprise data for the years from 2000 to 2010, it is the most comprehensive and thus best-suited data source for this paper. Yet, in assessing changes in competitiveness of British regions following the sterling’s devaluation after the Brexit referendum, this paper has to draw on input-output data from 2010.

Figure 4: Simplified Regional World Input-Output Table

Source:Los et al.(2017)

Figure4depicts a simplified version of a regional WIOT where only two countries A and B and a "Rest of World" (ROW) category are captured. A and B are then further disaggregated into two regions, namely, A1, A2, B1, and B2, each of which comprises two industries (I1 and I2). The columns in Figure4 give the buying entities; the rows list those that sell. Thus, matrix Z7incorporates intermediate inputs from all row-wise listed regions and countries to all column-wise listed regions and countries. Similarly, matrix F contains final goods shipments to end-users in regions or countries. The regional WIOT disentangles final demand into four categories. Yet, as specific final demand categories are of no further interest to this paper, aggregated final demand figures are used.

6_Figure_A.1_{in the Appendix lists all the British regions included in the WIOD.}

7_{As is standard in input-output notation, this paper denotes matrices in bold capital letters (e.g. Z), whereas}

(16)

Moreover, as this paper does not focus on particular industries and follows the approach of Bems & Johnson(2017), industry figures as given by I1 and I2 in Figure4are aggregated such that each region is home to only one single industry. In REER literature, this is referred to as the standardArmington(1969) assumption which states that each region produces only one type of good which can be differentiated from the goods produced in other regions (Chinn,2006;Klau & Fung, 2006). Regardless of the region of origin,Armington(1969) assumes that in a given market there is a constant elasticity of substitution between all competing product pairs which is the same for all competing product pairs in that market. Building their research on WIOTs allowsBems & Johnson(2017) to assume that each good can be used both as intermediate and final product which represents a major advantage over conventional REER indices. Yet, though assuming that each country – or region – produces only one product is common standard when assessing REERs, this gives rise to the possibility of aggregation errors.

Coming back to the matrices embodied in a WIOT, Z and F can be denoted as

Z =       z₁₁ z₁₂ · · · z1n z₂₁ z₂₂ · · · z2n .. . ... . .. ... z_n1 z_n2 · · · znn       and F =       f₁₁ f₁₂ · · · f_1n f₂₁ f₂₂ · · · f_2n .. . ... . .. ... f_n1 f_n2 · · · f_nn       ,

where a typical element z12gives the intermediate deliveries from region 1 to 2. The dimensions

of both matrices are determined by the number of regions/countries listed in the WIOD. These amount to 261 such that the matrices Z and F are of dimension 261x2618. The row vector w’ in Figure4gives the value added generated in each region. Due to the fact that input-output tables apply double-entry bookkeeping, gross output determined by x can be obtained in two ways. Adding all elements in one row, i.e., summing up over all intermediate inputs and final goods deliveries of one region (=x) delivers the same value as aggregating all intermediate inputs used by one region and that region’s value added (=x’). The latter method reflects the way gross output is computed in Equation3where gross output is denoted as the sum of imported intermediates, inputs from within the country, and value added in a given country.

Gross Output_Home= IntermediatesHome+ IntermediatesForeign+ Value AddedHome (3)

8_{Because the}_{Organization for Economic Co-operation and Development}₍₂₀₁₇_{) does not provide data for}

(17)

Finally, vectors w’ and x’ can be denoted as

x0= [x1, x2, ..., xn] and

w0= [w1, w2, ..., wn],

where both vectors are of dimension 1x261.

3.2 The Value-Added REER

Conventional REER indices as the one given in Equation 2are used to track changes in a country’s competitiveness by measuring the influence of relative prices on demand for the gross output of that country. To do so, conventional REERs as that in Equation2are computed for a time series which allows to compute the movements in the REER over time. However, not only haveTimmer et al.(2013) called to shift attention to value added instead of gross trade figures when assessing a country’s competitiveness to ensure informative outcomes and account for vertical specialization. Bems & Johnson(2017) also stress that, contrary to demand for gross output, demand for value added can influence macroeconomic policies. This is because these policies are usually formulated on a national level and therefore require data reflecting domestic performance. As Equation3shows, this is not satisfied by gross output data as it also includes foreign economic activities. Thus, Bems & Johnson (2017) define a value-added REER that derives the changes in a country’s competitiveness from global demand for its value added. The authors define their value-added REER as

\ REERva_i ≡

_∑

j6=i −Ti j Tii ( ˆpva_i − ˆpva_j ). (4) The hats in this equation denote log changes. This means that, contrary to Equation4which requires further computations to track changes in competitiveness,Bems & Johnson’s (2017) added REER directly gives changes in competitiveness. The log change in the value-added REER of region i is defined as bilateral relative value value-added price changes between region i and all other regions included in the data (as j 6= i) denoted by ( ˆpva_i − ˆpva

j ), weighted

(18)

populate the relative price changes ( ˆpva_i − ˆpva

j ) in Equation 4. This is because they base their

REER index on demand for value added which requires a measure for the price of real value added. As the GDP deflator deflates the nominal GDP into the real GDP, it can be seen as a measure for the price of real value addedBems & Johnson(2017).

The weights given in Equation 4 denote elasticities. Ti j reflects the cross-price elasticity between demand for value added from region i in relation to the relative value added price with region j. Thus, Ti j reflects by how much demand for value added in i changes given a change in relative value added price with region j. The denominator of the weights, Tii, gives own-price elasticities. Similar to the weights in Equation 2 those given in Equation 4 also capture the importance of each trading partner for region i which means that a region that trades intensively with i is assigned a higher weight. This derives from the fact that Ti j and Tii are populated by input-output data. Hence, demand for value added from region i is more sensitive to relative price changes with a region it trades intensely with. In line with standard REER indices, the weights sum up to one as ∑jTi j = 0. This equation follows from

the homogeneity of degree zero of demand for value added in value added prices which are given by the GDP deflators (Bems & Johnson, 2017)9. Hence, i’s own price elasticity (Tii) which is always negative offsets all cross-price elasticities (Ti j) which are typically positive but can also be negative. A positive (negative) cross-price elasticity implies that region i’s value added is a substitute (complement) for j’s value added, such that an increase in j’s value added price leads to an increase (fall) in demand for value added from i. Relating this finding back to Figure3, country A’s value added can be seen as complement to B’s value added, whereas B’s value added is a complement for C’s value added. However, assuming that the countries produce an Armington(1969) differentiated good, E’s value added is a substitute for that of C. Hence, it appears that allowing value added from different countries to be complements is essential when global value chains are involved. Unlike the above mentioned theoretical construct where relative value added prices of British trading partners change by a common constant t, the Brexit referendum led to varying movements in relative value added prices as the sterling depreciated by different ratios vis-à-vis British trading partners. This means that demand for value added of British regions changed in response to the depreciation.

Txwith element Txi j being the Ti jelement from Equation4subscripted by x can be defined

as

Tx≡ σ Tσ+ ρTρ+ γTγ,

whereBems & Johnson(2017) define the three matrices Tσ, Tρ and Tγ as10

9_{Homogeneity of degree zero holds if a simple function f (x, y) multiplied by a constant t returns f (tx,ty) =}

t0f(x, y) where t0_{= 1.}

10_{Matrix denotations in this paper differ in parts from those given in}_{Bems & Johnson}_{(2017). This is because}

(19)

Tσ ≡ [I − SzM2]−1Sf(M1− M2Wf)[I − ΩΩΩ0]−1[diag(svai )],

Tρ ≡ [I − SzM2]−1Sz(M1− M2Wz)[I − ΩΩΩ0]−1[diag(svai )],

Tγ= I + [[I − SzM2]−1SzM2(Wz− I) − I][I −ΩΩΩ0]−1[diag(sva_i )],

(5)

where all T-matrices are of dimension NxN. With the number of regions covered in this paper amounting to 261, all T-matrices are of dimension 261x261. Before the meaning of these matrices is explained in chapter 3.4, the subsequent chapter explains how the components of these matrices can be derived using data from the WIOD.

3.3 Deriving Value-Added REER Weights

As has been highlighted in section 2.2, researchers face conceptual choices when calcu-lating REERs. Ellis (2001) dedicates a lot of attention to the different ways to assign trade weights. When computing conventional REERs weights – denoted as w(i) in Equation 1 – institutions such as the IMF or the BIS rely on total trade figures (Bems & Johnson, 2017). Contrary to value-added REER weights, however, they are unable to differentiate between gross output and value-added as well as between intermediate inputs and final goods (Bems & Johnson, 2017). Value-added REER weights are derived from a set of matrices that can be populated using data provided by the WIOD11. As can be seen in Equations5, every T-matrix comprises either Wf or Wz. These matrices are denoted as

Wf=        fii fi fji fi · · · fni fi fi j fj fj j fj · · · fn j fj .. . ... . .. ... fin fn fjn fn · · · fnn fn        and Wz=        zii zi zji zi · · · zni zi zi j zj zj j zj · · · zn j zj .. . ... . .. ... zin zn zjn zn · · · znn zn        .

Both matrices are of dimension NxN, where N denotes the amount of regions and, if not disaggregated to a regional level, countries included in the WIOD. Elements fji and zji

de-note the final goods and intermediate inputs purchased by region i from region j, respectively, whereas elements fiand zidenote the aggregated expenditure on final goods and intermediates

in i, respectively. The typical element fji/ figives the ratio of final goods purchased in i from j

and total expenditure on final goods in i. The element zji/zigives the intermediates purchased

from j in i as a share of total expenditures on intermediates in i. Thus, the typical element in both matrices gives the importance of region j for region i as a source of final goods (Wf) and

as source of intermediate inputs (Wz).

Equations5also reveal that each T-matrix contains either Sz(Tρ and Tγ) or Sf and Sz(Tσ).

paper introduces price indices only if they are actually needed for computations.

11_{A simple numerical example of the derivation of value-added REER weights with a WIOT restricted to only}

(20)

Sf and Szare computed as

Sf≡     sf₁ 0 · · · 0 sf₂ · · · .. . · · · . ..     and Sz≡     sz₁ 0 · · · 0 sz₂ · · · .. . · · · . ..     ,

where sf_i= [s_i1f, ..., s_iNf ] , s_{i j}f = fi j/xiand sz_i = [sz_i1, ..., sz_iN] , sz_{i j}= zi j/xi. Thus, each element given

in either of these matrices is a vector of dimension 1xN. This implies that both matrices take the dimension NxN2. Thus, a typical vector captured as element in Sf or Sz gives the ratio of

final goods or intermediate inputs sold by one region to all regions (including itself) and gross output in the selling region. For instance, element s_{i j}f gives final goods purchased in region jfrom region i as a share of i’s gross output and element sz_{i j} gives the corresponding share of intermediates sold by i to j and gross output in i. This means that each element of a given vector sf_i (sz_i) gives the importance of region j as purchaser of final goods (intermediate inputs) from region i. Adding up over the rows of both matrices delivers the vector 1Nx1 as final goods and

intermediates sold by one region add up to that region’s gross output (see Figure4).

Next, the vectors sva_i and sx_i are computed. sx_i does not appear in Equations5, but is required to compute ΩΩΩ0. The vectors are defined as sva_i ≡ vai/xiand sxi ≡ zi/xiand both have dimension

1xN. Vector sva_i provides shares of each region’s value added with respect to its gross output whereas sx_i gives the shares of each region’s cumulated expenditure on intermediate inputs to its gross output. Similar to the rows in the matrices Sf and Sz, an element-wise addition of sva_i and

sx_i delivers a vector 11xN as each region’s total use of intermediate inputs and its value added

sum up to the region’s gross output (see Equation3). This implies that a region that does not use intermediate inputs extensively (a region with a high value added), reveals a high sva_i and a low sx_i. While differentiating between imported intermediates and intermediates from within the UK, FigureA.1in the Appendix shows that the Southern regions such as London or regions in Scotland reveal a lower total use of intermediates and, therefore, have a higher ratio of value added to gross output. As these regions also expose a lower use of imported intermediates than other regions, the depreciation in the sterling which causes rising import prices is less important for them. Finally, making use of the vector sx_i and the matrix Wz, the matrix ΩΩΩ0can be derived.

This is given as the product of the diagonalized vector sx_i and Wz, i.e.

      sx₁ 0 · · · 0 0 sx₂ · · · 0 .. . ... . .. ... 0 0 0 sx_n       ∗        zii zi zji zi · · · zni zi xi j zj zj j zj · · · zn j zj .. . ... . .. ... zin zn zjn zn · · · znn zn        = ΩΩΩ0.

With element sx_i being defined as zi/xi, the ii element of ΩΩΩ0equals zii/xiwhich is defined as

(21)

from region i which is required for one unit of output in that region. In standard input-output notation, input coefficients are captured by the A-matrix12. Input-output analyses assume these coefficients to be constant in the period under consideration such that an increase in production in region j by factor t implies an increase in the use of all inputs used in that region by the exact same factor: no input can be substituted by any other input. Yet, this assumption is close to reality where substituting inputs is hard in the short run. However, ΩΩΩ0 is not defined to be identical to the A-matrix but instead is its transpose, such that ΩΩΩ0= A0. Each row in ΩΩΩ0provides the use of intermediates of the purchasing region from all regions (again, including itself) as a share of its own gross output. This means that ΩΩΩ0gives ij elements reflecting the intermediates used by region j purchased from region i as a share of j’s gross output.

Eventually, from all matrices contained in the T-matrices, only M1 and M2 remain to be

explained. These are generic matrices that do not require any data input. The matrices are defined as M1≡ INxN⊗ 1Nx1 and M2 ≡ 1Nx1⊗ INxN, where the ” ⊗ ” reflects the use of the

so-called Kronecker product. Thus, the matrices can be denoted as

M₁≡     1Nx1 0 · · · 0 1Nx1 · · · .. . · · · . ..     and M₂≡     INxN INxN .. .     .

Thus, M1and M2are of dimension N2xN. M2captures N identity matrices whereas M1is

populated by N 1Nx1-vectors.

3.4 Meaning of T-matrices

After all components of the three T-matrices given in Equations5 have been derived, the meaning of the T-matrices themselves is now investigated.

As has been laid out before, conventional REER indices neither account for the use of imported intermediates nor do they account for the value added of a region, but instead rely on gross output figures. This is different in Bems & Johnson’s (2017) value-added REER index which captures each region’s input-output and trade structure in the three T-matrices. When aggregating the three T-matrices into a matrix Tx,Bems & Johnson(2017) weigh each matrix

with a parameter, namely σ , ρ, and γ. These parameters indicate elasticities of substitution among final goods from different regions (σ ), between intermediate inputs from different re-gions (ρ), and between intermediate inputs and value added (γ). In line with the standard Armington (1969) assumption, these elasticities are constant and identical for all competing product pairs which follows from the fact that each elasticity parameter is multiplied with every element in the respective T-matrix. For instance, if σ is set equal to one, then any competing

(22)

pair of final products reveals an elasticity of substitution equal to one. These elasticities also serve as weights that are attached to the respective T-matrices. As σ rises relative to ρ and γ, more weight is placed on the substitution between final products.

The workhorse model of this paper draws on the center stage model of Bems & Johnson (2017) which assumes "Leontief production", implying that ρ = γ = 0. This model, which is referred to as the "low elasticity" model is motivated by the idea of limiting the substitution of intermediate inputs. As has been mentioned above, the substitution of intermediate inputs in reality is absent in the short run. Bems & Johnson(2017) relate this finding to the rigidity of production chains in the short run which means that companies struggle to find adequate substitutes for their suppliers in the short run. In Figure3this implies that country C is unable to respond to a relative price increase vis-à-vis B and cannot substitute for B’s inputs. In this case, the value-added REER weights are exclusively determined by matrix Tσ. As has been

mentioned before, a typical ij value-added REER weight is high, when trade relations between region i and j are pronounced. Similarly, the typical ij element in Tσ is high when trade in final

products between region i and j is high. Thus, if lower weight is placed on trade in intermediates relative to trade in final goods (implying that ρ and γ decrease relative to σ ), value-added REER weights attached to regions’ final products trade partners rise relative to the weights attached to intermediates trade partners. Thus, in what follows, only the matrix Tσ, which was given as

Tσ ≡ [I − SzM2]−1Sf(M1− M2Wf)[I − ΩΩΩ0]−1[diag(sva_i )], is further described.

Tσ includes the trailing matrix [I − ΩΩΩ0]−1[diag(sva_i )]. With ΩΩΩ0being equal to the transpose

of the A-matrix, [I − ΩΩΩ0]−1represents the Leontief-Inverse, which is denoted as L. The typical element li j of L can be interpreted as the production in region i that is required to allow for

one unit of extra final demand for the product of region j. Thus, the ij element of the matrix [I −ΩΩΩ0]−1[diag(sva_i )] gives the production in region i that is necessary for a given ratio of value added over output in region j. This matrix is designed to capture how changes in value added prices in the UK affect demand for final goods of downstream producers, i.e., regions that use British intermediates in production. Downstream producers will benefit from a devalued sterling as inputs become cheaper for them. In response to that, their prices for gross output decreases and they become more competitive in trading their own goods – and their own value added – on global markets.

[I − SzM2]−1, the leading matrix in Tσ also captures input linkages. Solving the term

from the inside out, hence, multiplying Sz with M2 reshapes Sz such that it becomes equal

to [sz₁; sz₂; ...]. This means that off-diagonal elements, i.e. the zeros of Sz, disappear and the

matrix is now of dimension NxN. The resulting matrix is similar to the A-matrix. The typical element sz_{i j} is given as zi j/xi whereas the typical element ai j is given as zi j/xj. The matrix

[I − SzM2]−1 in Tσ is therefore similar to the Leontief-Inverse but requires a slightly different

(23)

required to create one unit of output in i. Less technically speaking, when consumers switch between final products, input linkages translate this effect into changes in demand for value added from specific countries of origin (Bems & Johnson, 2017). A depreciation in the sterling will not lead to higher demand for British intermediates in the short run, due to rigid production chains. Yet, it will increase demand for final goods from the UK. This benefits all regions – including British regions – producing inputs that are needed for the production of final goods in British regions as they can sell more intermediates to the UK. Hence, where the matrix [I − ΩΩΩ0]−1[diag(sva_i )] implies that downstream producers benefit from a depreciation in the sterling, the matrix [I − SzM2]−1shows that upstream producers to the British economy benefit,

too.

The term Sf(M1− M2Wf) is in the center of Tσ. When again solving this term from the

inside out, the term M2Wf replicates Wf such that the resulting matrix captures N versions of

it and is of dimension N2xN. When subtracting this term from M1 the result can be thought

of as a NxN matrix where the main diagonal elements are vectors of dimension Nx1. Each vector contains the typical ij element that equals 1 − fji/ fi and can be interpreted as i’s total

expenditure on final goods except for expenditure on final goods originated from region j. The typical off-diagonal element is equal to 0 − fji/ fiwhich means that off-diagonal elements give

the negative of the value captured in Wf. Multiplying Sf with (M1− M2Wf) delivers a matrix

that gives both own-price elasticities (main diagonal) as well as cross-price elasticities (off-diagonal) of demand for final products in relation to output prices13 (Bems & Johnson, 2017). Because elasticities in this matrix are dependent on international flows of final products, the cross-price elasticity of any foreign region j regarding price changes for final goods of a British region i due to the sterling’s depreciation not only hinges on the level of explicit bilateral competition between these two regions – thus, the competition of both regions’ final goods in both markets. It also depends on the degree of competition between these two regions on third markets, which is the competition of both regions’ final goods on any other than their own two markets (Bems & Johnson, 2017).

Finally, the element σ Tσi jreveals how demand for value added of British region i embodied

in a final product changes in response to a change in relative value added prices with a non-British region j. A depreciation in the sterling will leave the relative value added price of region jmore expensive, which implies that i’s final goods become cheaper in region i, in region j and in third markets. Clearly, consumers anywhere will then shift from purchasing j’s final goods to purchasing i’s final goods. Thus, they will substitute British for non-British goods. Again, in this "low elasticity" model, only expenditure for final goods is allowed to change – that for intermediates is constant. If this positive effect caused by the sterling’s depreciation is not relativized by spillovers induced by input linkages, then not only demand for i’s final goods

13_{As has been mentioned above, output prices are not included in the notation of this paper which is why they}

(24)

but also demand for i’s value added rises and i becomes more competitive. The typical element σ Tσi j is large when final goods trade is large between i and j. This implies that the British

regions which benefit the most from the sterling’s depreciation are those that have substantial final goods trade relations with regions where relative value added prices decrease the most in response to the depreciation.

3.5 The Use of Value-Added REERs When Examining Brexit

Using the value-added REER bears three major advantages over conventional REER indices for the purpose of this paper. First, the use of value-added REERs allows to assume a limited substitutability of inputs in the short run. In the immediate aftermath of the Brexit referendum, the sterling has decreased in its value which means that British producers were urged to find substitutes for at least some foreign suppliers who have become more expensive. However, producers in the UK probably struggled to find adequate substitutes for certain inputs in their own region. By relaxing the assumption (and thereby allowing ρ and γ to be greater than zero), the value-added REER does also allow to examine Brexit’s impact when allowing producers to be able to substitute for inputs.

Second, similar to conventional REERs, the value-added REER allows to incorporate changes in relative prices. By using elasticities as weights, it is also able to translate these price movements into changes in demand for value added. Moreover, cross-price elasticities (the Ti j) are not bound to be positive. Contrary to the conventional REER that draws on the Armington(1969) assumption where goods from different regions are always substitutes and REER weights are consequently always positive, the value-added REER allows that goods from different regions are complements. Referring back to the findings ofKlau & Fung(2006), this might be essential. In the "low elasticity" model,Bems & Johnson(2017) find value added from two regions to become complements if input linkages between region i and j – as described in the previous chapter – are more important than competition of their final products. If this condition holds, then value-added REER weights become negative. FigureA.1in the Appendix shows that the use of inputs from British regions is substantial for all regions and as a result of its EU-membership, the UK is deeply integrated into the EU market, leading to high amounts of imported inputs from the EU. This implies that allowing value added of two regions to be complements is essential.

(25)

weights, it is vital to distinguish between gross and value added data (Bems & Johnson, 2017) to reflect each country’s contribution in the production of a final good. With the Armington (1969) assumptions not taking into consideration this distinction, conventional REERs that draw on these assumptions use incorrect and misleading weights. Unlike these conventional REER indices, the value-added REER is able to attribute the value added from every produc-tion step to the region of origin. As countries are typically deeply integrated into value chains of the regional blocs they are located in – such as the EU, or North America –, the value-added REER typically attaches lower weights to countries that are located in the same regional bloc than the conventional REER (Bems & Johnson, 2017). This is because within regional blocs we often find many export moments before a final good is completed, but in each production step only a small value is added to the good. This means that in computing value-added REERs for British regions, EU countries get lower weights and countries located further away such as the US or China get higher weights compared to conventional REER weights. Depending on the exact movements of the sterling against the currencies of the UK’s trading partners, there is no doubt that this has positive results for the competitiveness of British regions. Table 1 shows that the sterling has lost more in value against the US dollar and the Chinese renminbi compared to the euro right after the outcome of the referendum. Therefore, attaching higher weights to the US or China is beneficial for the competitiveness of British regions.

Table 1: Sterling Depreciation Following Brexit Referendum

Currency _{June 23,2016}GBP/foreign currency_{June 24,2016} %-change (-)

Czech koruna 35.25 33.28 5.59 Euro 1.30 1.23 5.78 Danish krone 9.69 9.12 5.88 Swedish krona 12.09 11.51 4.83 Hungarian forint 409.58 389.52 4.90 Polish zloty 5.68 5.47 3.86 Japanese yen 156.68 139.59 10.91 Brazilian real 4.96 4.62 6.76 Australian dollar 1.95 1.83 6.06 Mexican peso 27.09 25.73 5.03 Russian rouble 95.09 88.72 6.70 Bulgarian lev 2.55 2.40 5.79 Romanian leu 5.89 5.55 5.78 Indian rupee 99.58 92.52 7.09 Indonesian rupiah 19654.52 18285.16 6.97 Canadian dollar 1.89 1.76 6.55 Chinese renminbi 9.73 9.03 7.19

South Korean won 1693.34 1593.67 5.89

Turkish lira 4.23 3.99 5.60

US dollar 1.48 1.36 7.81

Taiwan dollar 47.36 44.17 6.72

(26)

3 D A T A SOURCES AND METHODOLOGY 22

Table 2: British Intermediate Deliveries in 2010

Region North East North West Yorkshire & Humber East Midlands West Midlands East

North East 26026.53 1834.26 1552.50 552.11 698.75 828.23

North West 1847.82 88899.97 5548.14 3238.29 3079.83 2105.01

Yorkshire & Humber 1680.77 5355.68 61716.79 3847.23 1936.52 1561.53

East Midlands 711.70 3323.16 3880.97 50234.03 3274.04 2695.54 West Midlands 748.42 3050.51 1714.99 2992.20 63366.49 1631.83 East 808.29 1802.22 1356.33 3553.35 1584.60 81804.81 London 3766.87 10023.43 7046.68 5217.30 7173.46 7766.77 South East 1744.48 3353.39 2186.10 3007.96 4054.58 3923.98 South West 935.40 1782.11 1018.45 938.67 1756.99 1304.22 Wales 518.52 2032.23 954.02 622.83 2283.09 1148.25 Scotland 1202.83 3060.31 1873.32 1758.65 2266.18 1980.88 Northern Ireland 349.97 906.94 578.34 531.21 623.03 643.89

Region London South East South West Wales Scotland Northern Ireland

North East 1385.56 1240.70 865.54 102.60 1375.57 228.13

North West 3356.59 2992.51 2577.69 1790.67 3291.77 618.73

Yorkshire & Humber 2611.56 1884.18 1605.72 392.89 2113.63 712.50

East Midlands 2360.49 2353.94 1566.57 552.95 1034.09 627.12 West Midlands 2410.97 2498.36 2636.61 5600.68 2259.84 703.16 East 3783.51 3352.08 1624.24 350.55 281.32 631.74 London 192941.77 10883.60 6682.03 7647.14 19288.64 2882.67 South East 5180.02 138774.17 6087.36 1038.48 971.50 822.72 South West 2252.49 3852.89 74291.72 3548.15 389.55 453.10 Wales 1173.44 1770.42 1927.89 23174.19 103.52 158.70 Scotland 2400.20 3334.11 2197.15 296.45 59437.23 601.23 Northern Ireland 789.14 794.02 633.15 77.24 364.81 19738.2

Note:Data are given in millions of euros.

(27)

4 RESULTS 23

4 Results

4.1 Fitting the Data

In order to compute value-added REER weights, this paper draws on data from the region-alized WIOD (Thissen, Lankhuizen & Los, 2017) for the year 2010. UnlikeBems & Johnson (2017) who use data provided by the IMF to compute value-added REERs on a national level, this paper uses regional data provided by the statistics portal of theOrganization for Economic Co-operation and Development (2017). This is motivated by the fact that it is the only source that allows to compute regional GDP deflators and, thus, populate ( ˆpva_i − ˆpva

j ) in Equation4.

The OECD provides regional nominal GDP (Millions National currency, current prices) and regional real GDP (Millions National currency, constant prices, base year 2010) data which allow to derive regional GDP deflators and, thus, a price of real value added. As GDP deflators are defined in national currencies, they are subsequently converted into euro-terms so that they are given in a common currency. Conversions are carried out by using NERs provided by Eurostat(2017a).

The Organization for Economic Co-operation and Development (2017) provides data on NUTS-2 level except for some cases where data are given on NUTS-3 level. In these cases, data from the Organization for Economic Co-operation and Development (2017) had to be aggregated to match the regions given in the WIOT. Regarding the depreciation of the sterling incorporate in the analysis here, this paper draws back on exchange rate developments after the plebiscite in June, 2016. On June 24, the day following the Brexit vote, the sterling depreciated against all British trading partners included in the WIOD (see Table1). This sharp depreciation was a result of markets believing in lower growth potentials for the UK as a result of it leaving the EU (Gourinchas & Hale,2017). This paper uses the sterling’s depreciation against British trading partners on June 24, 2016 and incorporates it into computations. This means that this paper assesses the impact of Brexit on the international competitiveness of UK regions whilst drawing on the sterling’s dip following the Brexit referendum, whereas other potential effects of Brexit are not accounted for. It is also assumed that British production patterns in 2019 were the same as in 2010. In what follows, unless stated differently, all presented results are based on the assumption that the elasticity parameters ρ and γ, giving the elasticity of substitution between intermediates of different regions and between intermediates and value added, respectively, are equal to zero.

4.2 Getting the Weights Right

(28)

4 RESULTS 24

Moreover, theOffice for National Statistics (2016) states that the EU is even more important for the UK as a source of imports. Table3lists all UK regions included in the WIOD and the weight, thus, the Ti j/Tiithat has to be attached to major trading blocs such as the EU, NAFTA, and the BRIIC states14on the one side and the rest of the UK on the other side when computing each region’s value-added REER15. The weight attached to the rest of the UK regions has been included in the table to reflect the importance of intra-British trade for each region. The UK regions are listed according to the weight they place on the EU, starting by the region with the highest weight for the EU. As for Northern Ireland, for instance, 35.13% of demand for its value added is determined by trade with the EU. The table proves UK regions to be highly heterogeneous regarding the importance of their trade partners – even at a highly aggregated level considering only entire regional blocs. Table3shows that the EU receives considerably higher weights than regional blocs including major economies such as the US (NAFTA) or China (BRIIC). Thus, after accounting for vertical specialization, the EU’s importance for the UK is unbroken. Bearing in mind that Table3lists value-added REER weights for a situation where ρ = γ = 0, it follows that the EU is much more important as final goods trade partner than NAFTA or major emerging economies.

Table 3: Value-Added REER Weights for UK Regions

Region EU NAFTA BRIIC UK

Northern Ireland 35.31% 10.99% 8.05% 42.84%

Kent 33.76% 13.23% 7.37% 42.38%

Greater London 33.60% 10.24% 7.57% 45.92%

Cornwall and Isles of Scilly 31.99% 11.92% 7.52% 45.39% Surrey, East and West Sussex 31.98% 10.10% 7.26% 48.04%

Essex 31.77% 12.03% 7.58% 45.54%

Gloucestershire, Wiltshire and North Somerset 31.40% 16.07% 8.24% 39.91%

West Wales and The Valleys 31.40% 11.29% 6.60% 48.38%

Hampshire and Isle of Wight 31.11% 14.78% 7.91% 42.15%

East Wales 30.81% 9.80% 6.58% 50.55%

Cheshire 29.82% 10.31% 6.40% 51.25%

Berkshire Bucks and Oxfordshire 29.74% 10.18% 6.64% 51.01%

East Anglia 29.57% 9.48% 6.84% 51.64%

Leicestershire, Rutland and Northants 29.53% 12.55% 6.80% 47.99%

Devon 29.52% 12.13% 7.36% 47.55%

Dorset and Somerset 29.27% 12.43% 7.27% 47.54%

Lancashire 28.69% 11.63% 6.35% 50.69%

Greater Manchester 28.20% 9.25% 6.41% 53.75%

East Riding and North Lincolnshire 28.19% 10.18% 5.94% 53.49% Northumberland, Tyne and Wear 28.16% 8.64% 6.45% 54.44%

14_{NAFTA = North American Free Trade Agreement, including the US, Canada, and Mexico; BRIIC = Brazil,}

Russia, Indonesia, India, and China.

(29)

4 RESULTS 25

Tees Valley and Durham 27.87% 8.88% 6.10% 54.92%

Shropshire and Staffordshire 27.82% 11.88% 6.74% 50.51%

Merseyside 27.78% 8.20% 6.15% 55.63%

Bedfordshire, Herefordshire 27.68% 8.43% 6.45% 55.20%

Cumbria 27.63% 9.32% 5.72% 55.20%

Herefordshire, Worcestershire and Warks 27.44% 11.67% 6.76% 51.10% Derbyshire and Nottinghamshire 27.42% 8.72% 6.32% 55.27%

North Yorkshire 27.33% 10.46% 6.37% 53.08% West Yorkshire 26.90% 8.78% 6.19% 55.88% South Yorkshire 26.24% 8.10% 6.03% 57.45% Lincolnshire 26.20% 9.99% 6.20% 55.03% West Midlands 25.40% 9.00% 6.18% 57.15% Eastern Scotland 24.23% 9.66% 5.20% 58.96%

South Western Scotland 23.29% 10.13% 5.21% 59.34%

North Eastern Scotland 15.96% 7.62% 3.43% 71.50%

Highlands and Islands 12.55% 7.42% 3.01% 75.50%

Note:Percentage shares given in the table do not add up to 100. This is because the residual set of countries not included in either of the regional blocs is not listed.

Source: Own computations based on data fromThissen, Lankhuizen & Los(2017).

(30)

4 RESULTS 26

4.3 Competitiveness in the UK

Table 4presents results for value-added REERs when accounting for the sterling’s depre-ciation as given in Table 1. Bems & Johnson (2017) define value added price changes in Equation 4 as the value added price of region i divided by the value added price of region j in year t+1 over the exact same expression for the year t. Incorporating a depreciation in the sterling then demands to multiply each British region’s relative price in year t+1 by (1-x) where x denotes the depreciation of a given trade partner as given in Table1. Regions listed in Table 4 are sorted according to the magnitude of the log change in their value-added REER. With negative values indicating increases in competitiveness, the results reveal that despite a large variation, all regions in the UK are able to benefit from a depreciation in the sterling as each region’s competitiveness increases. Although some of them make intensive use of imported intermediates as shown in Figure 1, it is notably the regions from the South of England who are amongst the biggest beneficiaries of a depreciation in the sterling. This implies that input linkages are not large enough to offset the effect of competition in final goods among these British regions and their trade partners. Yet, among those regions that see their competitiveness increase the most, there are also regions associated with the Northern regions. Several regions from the Midlands appear on top of Table4and Northern Ireland also seems to benefit consid-erably from a lower value in the sterling. However, most of the Northern regions are listed at the bottom of Table 4. What is more, while the value-added REER only gives a measure on how a region’s competitiveness in selling its value added changes, e.g. that of Gloucestershire, Wiltshire and North Somerset has improved by 0.04564%, it can also be used to compute exact changes in demand for value added (Bems & Johnson, 2017).

Table 4: Value-Added REERs for UK Regions

NUTS-1 Regions NUTS-2 Regions Log Change in Value-Added REER South West Gloucestershire, Wiltshire and North Somerset -45,64 South East Hampshire and Isle of Wight -42,88

South East Kent -38,52

South West Dorset and Somerset -37,51

South West Devon -37,44

South West Cornwall and Isles of Scilly -37,08

East Essex -37,01

East Midlands Leicestershire, Rutland and Northants -35,86

Northern Ireland Northern Ireland -35,39

London Greater London -34,93

West Midlands Shropshire and Staffordshire -34,71 West Midlands Herefordshire, Worcestershire and Warks -34,35 South East Surrey, East and West Sussex -33,55

Wales West Wales and The Valleys -33,24

North West Lancashire -32,66

(31)

4 RESULTS 27

Yorkshire and the Humber North Yorkshire -31,54

East East Anglia -31,28

Wales East Wales -31,24

North West Cheshire -30,65

East Midlands Lincolnshire -30,26

North West Greater Manchester -29,93

Yorkshire and the Humber East Riding and North Lincolnshire -29,29 North East Northumberland, Tyne and Wear -29,11

East Bedfordshire, Herefordshire -28,83

West Midlands West Midlands -28,73

North East Tees Valley and Durham -28,64

East Midlands Derbyshire and Nottinghamshire -28,62 Yorkshire and the Humber West Yorkshire -28,50

North West Merseyside -28,12

North West Cumbria -28,10

Yorkshire and the Humber South Yorkshire -27,27

Scotland South Western Scotland -26,88

Scotland Easter Scotland -26,84

Scotland North Eastern Scotland -18,80

Scotland Highlands and Islands -17,60

Note:Data is given in 1,000s here. For instance, -45.64 has to be read as -45.64∗e−03.

Source: Own computations based on data fromThissen, Lankhuizen & Los(2017), Organization for Economic Co-operation and Development(2017), andEurostat(2017a).

Interestingly, when focusing on NUTS-1 regions, it becomes clear that there is a consider-able variation in how specific NUTS-2 regions located in a common NUTS-1 region perform. For instance, in the East Midlands, Leicestershire, Rutland and Northants appears among the main beneficiaries, whereas Derbyshire and Nottinghamshire appears at the bottom of Table4. Table 4also shows that the Scottish regions are hardly able to increase their competitiveness. When considering the high weights that Scottish regions place on intra-UK trade as given in Table 3, however, this finding is hardly surprising. As a consequence, Scottish regions are those regions benefiting the least from a lower value in the sterling among all British regions. Thus, the "low elasticity" scenario predicts the Scottish regions to lose in competitiveness rel-ative to all other British regions. Generally, the results listed in Table4 speak in favor of the hypothesized increase in the North-South divide.

(32)

4 RESULTS 28

to the Brexit referendum. Correspondingly, Figure5 reveals several outliers. Hampshire and Isle of Wight as well as Gloucestershire, Wiltshire and North Somerset both reveal a higher increase in competitiveness than the fitted line would suggest. On the other hand, Northern Ireland which trades to a similar extent with other British regions as Hampshire and Isle of Wight reveals a much lower increase in competitiveness. These deviations can be explained by the weight these regions place on trade with NAFTA. Table1shows that the sterling has depre-ciated more vis-à-vis the dollar than against the euro. With Hampshire and Isle of Wight and Gloucestershire, Wiltshire and North Somerset placing the highest weights on NAFTA among all British regions, they also benefit the most from the weaker sterling vis-à-vis the dollar.

Figure 5: Relationship Between UK Trade Weights and Value-Added REERs

Source: Own computations based on data fromThissen, Lankhuizen & Los(2017), Organization for Economic Co-operation and Development(2017), andEurostat(2017a).

(33)

4 RESULTS 29

the depreciation of the sterling does not only increase the North-South divide but also leaves the economically most disadvantaged Northern regions in the UK worse off.

Figure 6: Relationship Between British Regional Income and Value-Added REER

Note:Observations are labeled according to NUTS-1 regions instead of NUTS-2 regions.

Source: Own computations based on data fromThissen, Lankhuizen & Los(2017), Organization for Economic Co-operation and Development(2017), andEurostat(2017b).

Brexit and its Impact on Regional Competitiveness in the UK: