Economic exposure to regional value chain disruptions: evidence from Wuhan’s lockdown in China

ABSTRACT Production fragmentation across multiple regions can result in a regional shock propagating along value chains to a wider array of regions. We propose a methodological framework to measure the economic exposure to regional value chain disruptions due to city lockdown during Covid-19. The exposure index is evaluated by applying a hypothetical extraction method to a regionally extended inter-country input–output framework incorporating China’s interregional input–output table. Our methodology can be adapted to conduct disaster impact analyses at city, state and country levels. It provides a tool for the immediate assessment of the economic risks of value chain disruptions, enabling quick policy responses.


INTRODUCTION
The production processes of goods and services in the modern economy are fragmented and conducted in multiple regions and countries (Antràs & Gortari, 2020;Baldwin, 2016;Tian et al., 2022a). Given the fact that an industry in a region requires a variety of different inputs from external regions or countries, interregional and inter-industry linkages play vital roles in the economy. This has been attested by the enormous increase in interregional transactions and trade in intermediate products.
A consequence of geographically fragmented production is that disruptions to the orderly flow of products can cause chain risks. A regional shocksuch as a natural disaster or a terrorist attacknot only causes damage to local industries but also can propagate through input-output (IO) linkages to a wider array of industries outside the region (Carvalho et al., 2021).
In this paper, we provide a methodological framework for quantifying the role of interregional input-output (IRIO) linkages in propagating a regional shock along value chains and exploit the case of Wuhan's lockdown during the early stages of Covid-19. To contain the spread of Covid-19, China imposed a strict lockdown across Wuhan City, the first Chinese city hit by in Hubei province on 23 January 2020. The strict lockdown did not lift until 8 April 2020. During the lockdown, all cross-region transportation was suspended, and very few firms operated normally in Wuhan City. As the capital city of Hubei province, Wuhan plays a substantial role in China's economy, with its gross domestic product (GDP) in 2019 ranking ninth in all cities in China. Wuhan is colloquially known as 'the thoroughfare leading to nine provinces' because it is closely connected to external regions due to its geographical centrality and welldeveloped transportation infrastructure. Thus, Wuhan's stringent lockdown disrupted the relevant value chains seriously.
Therefore, it is necessary to measure the heterogeneous economic exposures of different regions to the Wuhanrelated value chain disruptions. That is, to quantify the extent to which a certain industry in a certain region is exposed to linkages with Wuhan. We adapt the hypothetical extraction method (HEM) (e.g., Dietzenbacher et al., 2019) to calculate the regional GDP in a hypothetical situation where the Wuhan economy is not operational and Wuhan-related value chains are disrupted. We define the difference between the original GDP and hypothetical GDP as the economic exposure to value chain disruptions. The extent of exposure depends on the nature and scale of its direct and indirect interdependencies with the Wuhan economy. The results of this study provide implications for the regions that may directly and indirectly be affected to pre-arrange plans for minimizing the propagated effects spawned by value chain disruptions.
Applying the HEM, this study develops an exposure index upon a flourishing strand of literature using IRIO tables (e.g., Boero et al., 2018;Jiang et al., 2012) and inter-country input-output (ICIO) tables (e.g., Koopman et al., 2014;Los et al., 2016;Tian et al., 2019). Moreover, this study contributes to the existing literature by considering the propagated shock of a regional event from a global perspective. In the era of pronounced regional and international production fragmentation, a regional shock in China may propagate not only to domestic regions within China but also to regions outside China. To measure the exposures of both China's domestic regions and regions abroad, we construct a regionally extended inter-country input-output (REXICIO) table where China is disaggregated into 31 provincial units. The REXICIO table contains IO linkages for China's domestic provincial units and economies outside China. Therefore, our exposure index incorporates all direct and indirect effects due to geographically fragmented production processes within and outside China.
A growing list of studies is assessing the economic impact of Covid-19 on China and the global economy. Various models, including the computable general equilibrium (CGE) model, simulation methods, and a combination of multiple models have been widely employed to evaluate or project the economic cost of Covid-19 (Duan et al., 2021). However, results based on these models are significantly different in different phases of Covid-19. This is because CGE-based and simulation-based methodologies largely rely on scenario set-ups and assumptions to reflect real situations. However, Covid-19 has been continually evolving, making it difficult to accurately predict its duration and severity. In response to the pandemic, cities, regions, and countries have implemented different control measures in different phases. Variations in the strictness of measures and the rapidity with which they are imposed/relaxed reflect divergent economic impacts. Owing to such substantial uncertainties, it is particularly challenging to set realistic scenarios and speculate, in detail, the possible economic impacts of this epidemic.
Instead of relying on scenario set-ups and assumptions, we opt for a simple and convenient methodological framework based on an IO table, which is normally available before a disaster. Thus, our method can, ex ante, evaluate the economic exposure to value chain disruptions immediately. Once a lockdown decision is made in response to an emergent event, our methodology can be used to provide immediate information on the industries and regions that are highly exposed to the lockdown. Instant assessment is vital for policymakers to respond quickly to sudden value chain disruptions. For example, if the results indicate that downstream car manufacturing in Guangdong province is highly exposed to purchasing intermediate products from Wuhan, producers and policymakers in Guangdong can respond instantly by, say, switching their purchasing sources from Wuhan to other regions.
The remainder of the article is structured as follows. Section 2 reviews the methods for disaster impact analysis. Section 3 explains our construction of the exposure index. Section 4 outlines how the REXICIO table was attained. Section 5 provides the results. Section 6 concludes.

REVIEW OF THE METHODS FOR AN ECONOMIC IMPACT ANALYSIS OF THE DISASTER
A disaster is a sudden calamitous event that brings great damage, loss or destruction. It can be the consequence of natural or man-made hazards, such as earthquakes, extreme weather, pandemics and so forth. These hazards not only destroy human lives and the living environment but also lead to interruptions in economic activities. The economic impact in the damaged area can potentially propagate to firms/industries in other areas via forward and/or backward linkages. This type of propagation effect is called the intersectoral or higher order effect.
Estimations of the economic loss of a disaster are roughly twofold: pre-and post-hazard estimations (Okuyama & Santos, 2014). The pre-hazard estimation is to evaluate the economic impact of potential disasters ex ante and can be used to evaluate preparedness and mitigation policies. For example, the estimation results can help evaluate whether pre-hazard mitigation policies are cost-saving compared with post-hazard recovery and reconstruction. The post-hazard estimation appraises the economic losses of a disaster compared with the situation without the disaster, conducive to post-hazard policymaking targeting at mitigating economic losses (e.g., Carvalho et al., 2021).
The IO model is one of the most widely used methods for analyzing the economic impact of a disaster (Okuyama, 2007) and its application to disaster analysis dates back to the studying of strategic bombing during the Second World War (Rose, 2004). One of the advantages of the IO model is its ability to capture higher order effects of a disaster via interregional/inter-industry linkages. However, the IO model also has a set of limitations, including its linearity, a lack of response to price changes, and its rigid structure between domestic and imported inputs and primary and intermediate inputs (Rose, 2004). Some extensions have attempted to address these limitations by adapting the standard IO framework. Hallegatte (2008) proposed an adaptive IO model to incorporate production capacity, price changes, and changes in labour demand. Extensions also considered the duration of the hazard occurrence. For example, a sequential inter-industry model (SIM) was introduced to investigate the dynamic process of the disaster impact paths and recovery process while maintaining the IO economic structure (Okuyama et al., 2004). Yet, SIM has not been widely adopted in empirical analysis, mainly because of the difficulties in obtaining the required data and partly due to the limitations of the IO framework.
The social accounting matrix (SAM) has also been used to conduct disaster impact analysis (Cole, 1995), while the SAM approach has similar advantages and limitations to the IO model. The lack of recent SAM tables has restricted the application of SAM-based models. There have been no official SAM tables for China since 2000 and researchers need to compile their own SAM tables when necessary. SAM tables, which include China's provincial information, are far from available. It is a huge challenge to compile such tables. Instead, as shown below, we have a world IO The inoperability input-output model (IIM) is another notable IO-based approach to assess disaster impacts (e.g., Haimes & Jiang, 2001;Santos et al., 2014;Santos & Haimes, 2004). This model introduces the concept of inoperability in the traditional IO model. The inoperability index was normalized, ranging between 0 (ideal system state) and 1 (total failure state). Dietzenbacher and Miller (2015) and Oosterhaven (2017) argued that IIM was a straightforward, albeit potentially very relevant, application of the standard IO model with a small tweak. Dietzenbacher and Miller (2015) further linked it to the HEM and argued that IIM can be seen as a variant of hypothetical extraction. In this study, we adopt HEM because we need to hypothetically extract not only parts of final demand but also intermediate demand. However, to the best of our knowledge, most current studies using IIM focused on the 'inoperability' generated by reductions in final demand, and few have captured the 'inoperability' of intermediate demand.
HEM has been widely employed to quantify the importance of an industry or set of industries to an economy. The core idea of this technique is to quantify how much the economic output of an economy would decrease assuming a particular industry is not present, the so-called hypothetical extraction. Considering that the effects of changes in intermediate output were not easily rationalized in earlier HEM or traditional IO-based models such as IIM, Dietzenbacher and Lahr (2013) generalized the HEM and proposed partial extraction. For partial extraction, an industry consists of a number of identical establishments. Hence, a proportion of (establishments in) this industry (instead of the whole industry) can be extracted. Dietzenbacher et al. (2019) further generalized the HEM to a global model considering the increased inter-country inter-industry linkages. Los et al. (2017) and Chen et al. (2018) employed global extraction to assess the economic exposures to Brexit and Hu et al. (2021) used it to assess the impacts of US-China trade decoupling. In the next section, we illustrate how we use partial HEM to estimate the economic exposure to the Wuhan lockdown.
Another widely used modelling framework in disaster impact analysis is the CGE model. Unlike IO models, CGE models are not linear in common practice (although some CGE models also use the Leontief production function), can respond to price changes, and incorporate the substitution among inputs from different sources. However, CGE models have been described by some researchers as 'black boxes' because of their complexity, lack of transparency, the large number of key parameters, and the arbitrary choices of the values of these parameters. Some assumptions of optimizing behaviour in CGE models are questionable under disaster situations. Researchers also argue that most CGE models are intended for long-run equilibrium analysis and provide lower impact estimates than IO models which are more suitable for short-term analysis (Rose & Liao, 2005). In empirical practices, the IO and CGE model are also extended or combined with other models to solve relevant research questions (e.g., Duan et al., 2021).
Although quantitative macroeconomic models have proven useful in quantifying the economic impacts of disasters, disaster impact analysis has always been challenged by the availability and quality of basic data and inherent model limitations. So far, no perfect model exists for disaster impact quantification. Consequently, estimates using different models may generate divergent results even with the same event. Nevertheless, when used in the correct direction, the estimates can help create effective policies for disaster preparedness and loss mitigation.

PARTIAL HYPOTHETICAL EXTRACTION METHOD
To measure the economic exposure of a region to Wuhanrelated value chain disruptions, we need to measure its exposure to inter-industry linkages with Wuhan City. We adopt the HEM Los et al., 2016) which calculates the GDP in a hypothetical situation where the Wuhan economy is not operational and thus Wuhan-related value chains are disrupted. We define the difference between the original regional GDP and the hypothetical GDP as the estimate of regional value-added that is exposed to Wuhan-related value chain disruptions.
Conventional inter-country input-output (ICIO) tables enable us to conduct country-level analyses considering inter-country dependencies (Tian et al., 2022b). To meet the further need to consider the linkages between China's domestic regions and the global market, we compile the REXICIO table incorporating China's domestic provincial information into the world IO table (Figure 1). We assume that there are m countries with n sectors in each country. China is geographically disaggregated into 31 provincial units, while other countries are not specifically disaggregated into detailed regions.
Matrix Z gives the values of the intermediate input deliveries. The elements in the block labelled f /F give deliveries of the final products. The last column y gives the value of gross output by each sector in all regions/ countries.
The product market-clearing condition is that the summation of a regional industry's products for intermediate Economic exposure to regional value chain disruptions: evidence from Wuhan's lockdown in China 527 use and for final use equals its gross output. This accounting identity is y = Zu + Fu, where u stands for the summation vector consisting entirely of ones and it summarizes the elements in matrices Z and F in a rowwise fashion. We further define the square matrix A = Z(ŷ) −1 as the direct input coefficients matrix with its typical element a sr ij = z sr ij /y r j , which gives the inputs from sector i in region/country s for intermediate use by sector j in region/country r. This yields y = Zu + Fu = Ay + Fu and the solution is given by the well-known static IO model (Miller & Blair, 2009): (1) Matrix A reflects not only production technology but also international and interregional intermediate trade structures. I is an identity matrix with ones on the diagonal and zeros elsewhere. L ; (I − A) −1 is known as the (global) Leontief inverse with the element b sr ij , which gives the extra output of sector i in region/country s that is required for a one unit increase in final demand of sector j in region/country r.
Let v ′ = w ′ (ŷ) −1 be the row vector that provides valueadded coefficients. Its typical element v r j = w r j /y r j provides the value-added generated in sector j in country r per unit of output in this sector. This yields: Equation (2) shows that world GDP is determined by value-added coefficients, intermediate input coefficients, and world demand levels for the final products. Subsequently, the GDP of a specific region r can be attained by replacing vector v in equation (2) with vector v r . The new vector has the same length, but only the value-added coefficients for the sectors in region r are retained, whereas all other elements are set to 0. We thus have: This study aims to measure the extent to which a specific economy r is exposed to the disruption of Hubei-related value chains due to the Wuhan lockdown. The disruption reduces not only the deliveries of final products and intermediates from Hubei but also part of Hubei's demand met by regions outside Hubei. Then, following Los et al. (2016) and Chen et al. (2018), we calculate the hypothetical GDP by hypothetically extracting a part of the corresponding flows of both intermediate and final products: with where the overbar indicates the case of a partial extraction. The use of A * and F * hypothetically extracts a proportion of product flows between Hubei and other regions, simulating the situation where both Hubei's demands from other regions and other regions' demands from Hubei are severed due to the lockdown. Equation (4) yields the value-added in region r in a hypothetical situation. The index for regional economic exposure is given by the 528 Kailan Tian et al. difference between the original and hypothetical GDP divided by the original regional GDP: In the following empirical analysis, we also report the results for the sectors in region r. The results are obtained by replacing vector v r withv r , in which the circumflex denotes a diagonal matrix.
Regarding partial extraction, we follow the assumptions of Dietzenbacher and Lahr (2013) and Dietzenbacher et al. (2019) that an industry consists of a number of identical establishments. Some establishments in the industry cease to be operational due to lockdown, so the industry's capacity declines. Assuming both industry k's deliveries to (intermediate and final) use and deliveries to industry k decrease by a percentage b k , the element in submatrix A hh can be rewritten as The same decrease applies to elements in the submatrices A rh , A ch , A hr , and Similarly, the Hubei-related final product flows are adjusted.
We argue that, in our case, the partial extraction ratio b k can be used to trace the duration of the lockdown. The longer the duration, the more establishments in the industry cease to be operational, and the larger b k is. In the empirical case study below, we need to set the value of b k for all k = 1, · · · , n. We let b k be the ratio of industry k's operating revenue in Wuhan city to the same industry's operating revenue in Hubei province (Table 1). As for final demand, we assume b k is the ratio of Wuhan's final demand to Hubei's final demand. This implies a situation where production in Wuhan crashed entirely and Wuhan's final demand reduced close to zero except for daily necessities due to the strict Wuhan lockdown. Below, we test the results by considering the real scenario in which substantial agricultural products and medical supplies were delivered to Wuhan during the lockdown. Here, we notice that this setting could be an extreme case. Nevertheless, the results can inform us of the importance of the Wuhan economy to other regions and provide policymakers with information on regional economic exposures to the collapse of the Wuhan economy. If ex post data are available, our approach can be used by modifying the value of b k to simulate the actual regional economic losses caused by the Wuhan lockdown.
To conclude, our approach splits value-added in a region into two parts: a part that is embodied in the inter-industry linkages between the region itself and Hubei and the other part that has neither direct nor indirect linkages with Hubei. The former part is thus exposed to the disruption of Hubei-related value chains.
It should be noted that our approach estimates the economic exposure from the demand side due to the demand-driven nature of the underlying Leontief model. Therefore, the results are determined by the extent to which Hubei demands (intermediate and final) products from other regions, as well as the extent to which the other regions demand products from Hubei. Supply-side bottlenecks are disregarded in this model. For example, a firm in Guangdong province that produces a final product (e.g., a car) requires intermediate inputs (e.g., chips) from Hubei. If such core intermediate components are not available from Hubei and no substitution from anywhere else in the world is possible, the firm in Guangdong cannot produce cars at all. The impacts of such supply-side bottlenecks could be captured by future studies that may develop a model from the supply side. However, the development of such a model is beyond the scope of the present study.  Table A1 in the Appendix in the supplemental data online for the sector classifications). All transactions are converted to values in dollars using the market exchange rates. The compilation of the REXI-CIO tables follows Meng and Yamano (2017). The key procedure is as follows.
The core of constructing the REXICIO table is to merge the Chinese domestic IRIO table into WIOD by combining the structure of the existing IO table with the provincial trade (international import and export) data from customs statistics. First, we give initial values for China's provincial imports/exports of intermediate and final products from the country of origin/ destination in the REXICIO table. China's domestic IRIO table provides information about provincial trade but does not show the origins or destinations of the traded products. Provincial customs trade data provide information about the provinces' trade partners. Thus, we use this information and China's domestic IRIO table to proportionally split China's corresponding total import/export value in the WIOD into the provincial level. Using the broad economic categories (BEC) defined by the United Nations Statistics Division, we distinguish the provincial data of the goods trade into trade for intermediate use and final use. Due to the lack of provincial statistics of services trade, we use the structure of provincial goods trade as a proxy to attain the initial values for services. The underlying assumption is that if more goods are traded between a specific Chinese province and an economy abroad, more services may also be traded between these two economies. An alternative is to use provincial services trade data taken from China's domestic IRIO table as a proxy. However, it should be noted that the provincial services trade flows in the Chinese domestic IRIO table are also estimated.
Next, we use cross-entropy models (Robinson et al., 2001) to estimate and balance international inter-industrial transactions between China's provinces and economies abroad, so that China's domestic IRIO table can be consistently merged into the WIOD. The programming processes are conducted by minimizing the entropy distance between the prior transaction and the new estimated transaction under constraints posed by control totals. Control totals are usually taken from prior information which is a crucial constraint for balancing the IO table. For example, China's provincial customs import/export statistics are important control totals in the balancing procedure. Using the minimization process, we can attain the REXICIO table with balanced rows, columns, international, interregional, and inter-industry relationships. Prior information can also be preserved as much as possible.
To the best of our knowledge, we are among the first to provide such a database, disaggregating China into 31 provincial units within global IO tables. Our empirical results in the next section are based on the 2012 table, which is the latest table available. Considering that the international fragmentation of production processes has been rather modest in recent years (Timmer et al., 2021), we argue that our empirical results are comparable to what we would find for the more recent years if the data would have been available.

RESULTS AND DISCUSSION
Our results show that China's national level of economic exposure to the collapse of Wuhan-related value chains was 2.5% of the GDP. Wuhan is one of the most important cities in China; moreover, Wuhan's industries are closely connected with other regions in China, and these connections have increased between 2012 and 2020. Consequently, China's national exposure to the Wuhan lockdown in the first quarter of 2020 could be larger than the estimation provided by the REXICIO table for 2012. Hubei, the province where Wuhan is located, exhibited the highest level of exposure (52.5% of its GDP). As the provincial capital and the economic centre of Hubei province, Wuhan plays a significant role in Hubei's economic system. Wuhan's GDP accounted for approximately 35.7% of the province's GDP in 2019. Therefore, the Wuhan lockdown brought tremendous risk to Hubei's economy. Regarding other provincial units excluding Hubei, 0.5% of their GDP (US$39,100 million) was at risk.
The exposure rates of countries outside China were relatively lower than those of China's domestic regions. The level of exposure of all economies abroad was 0.02% of their GDP (US$11,843 million). Japan and South Korea were the two most exposed foreign countries, 0.02% of GDP (US$1202 million). The EU (i.e., EU

530
Kailan Tian et al. countries presently including the UK) and the United States were the next most exposed economies. The EU's level of exposure was 0.01% of GDP (US$1752 million), the United States' was 0.005% of GDP (US$765 million), and the rest of the world's was 0.03% of GDP (US$8124 million). The results suggest that Wuhan is more likely to be a domestic-oriented economy; therefore, it plays a more important role in China's domestic regional value chains than global value chains. Therefore, China's domestic regional value chains are more vulnerable to the Wuhan lockdown than global ones.

Regional heterogeneities in economic exposure
The economic exposures of China's four aggregated regions and 31 provincial units to the collapse of Wuhan-related value chains are depicted in Figures 2  and 3, respectively. The classification of the four regions and the detailed provincial results are reported in Table  A2 in the Appendix in the supplemental data online. It can be observed that China's economic exposures show substantial regional variations. For central regions (excluding Hubei) and western regions, the exposure rates were 0.8% and 0.7% of GDP, respectively, which were higher than those in the eastern and north-eastern regions. The most exposed provinces include Qinghai, Shaanxi, Shanxi and Ningxia in the western and central regions, with exposure levels ranging from 1% to 3% of provincial GDP. Heterogenous exposure levels for different regions can be attributed to their divergent linkages with Hubei's economy which were under great risk from the Wuhan lockdown. Accounts based on the REXICIO table show that the most exposed regions have a higher economic dependency on Hubei's economy. For instance, Shaanxi, Shanxi, and Henan were the top three origins of Hubei's intermediate inflows, with 12.0%, 7.9% and 7.9% of Hubei's total intermediate inflows originating from the three provinces, respectively. Meanwhile, Hubei was the third-, sixth-and seventh-most important destination for the intermediate outflows of Qinghai, Shanxi and Ningxia, with 9.8%, 5.2% and 4.1% of their total intermediate outflows flowing to Hubei, respectively. Henan Figure 3. Provincial shares of local gross domestic product (GDP) exposed to Wuhan value chain disruptions. Note: For the detailed data, see Table A2 in the Appendix in the supplemental data online.
Figure 2. Regional shares of local gross domestic product (GDP) exposed to Wuhan value chain disruptions.
Economic exposure to regional value chain disruptions: evidence from Wuhan's lockdown in China 531 was the principal origin of Hubei's trade of final products, with 29.3% of Hubei's total domestic inflows of final products coming from Henan. A closer look into the data on interregional flows reveals that Hubei imported more intermediates (e.g., natural resources, raw materials, parts, and components) from the western and central regions but exported more final products to the eastern regions. In other words, Hubei played an important role as the 'production node' where products were produced with intermediate inputs from the western regions and then delivered to the eastern regions for final use. Consequently, the downstream stages of the western production chains were more vulnerable to the economic risk of Hubei resulting from the Wuhan lockdown.

Sectoral heterogeneities in economic exposure
The economic exposures to Wuhan value chain disruptions are determined by both the sectoral partial extraction ratio and the interregional inter-industry linkages between Hubei and the other regions. Table 1 presents the sectoral exposure values. In China, the fundamental sectors supporting the economic system by providing requisite elements for regional production and living activities like the 'Electricity, gas, and water supply' and the 'Mining and quarrying' sectors, exhibited relatively higher levels of exposure 3.6% and 3.3% sectoral GDP, respectively. Manufacturing sectors had generally high exposure levels (approximately 2.7% of sectoral GDP), among which 'Manufacture of machinery and equipment' was the most exposed manufacturing sector. As one of the mainstay industries in Hubei, this sector had pronounced interregional economic integration, with its outflows to other regions accounting for 16.4% of Hubei's aggregated outflows of all manufacturing sectors. This sector had a large extraction ratio and tight cross-regional inter-industry linkages. Thus, it suffered the highest level of economic damage. The extraction ratios for the services sectors were also high, but many service sectors had weak cross-regional inter-industry linkages. As a result, the exposure levels of service sectors were relatively lower than those of the secondary industries, with the exposure levels of 'Wholesale, transport, and other services relevant to production activities' and 'Remaining services' being 2.5% and 2.1% of sectoral GDP, respectively. However, because of their close relevance to the production processes of secondary industries, the values of services related to production activities become tradable among regions via indirect embodiment in manufactured products. Therefore, the exposure levels of such services were higher than those of the remaining services to the disruption of industrial production chains.
We further specified the differentiated features of sectors to investigate their spatial distribution patterns of economic exposure. We selected three representative sectors according to their 'upstreamness' throughout the production chains. We followed the definition of upstreamnessaverage distance from final useby Antràs et al. (2012), and selected 'Mining and quarrying', 'Manufacture of machinery and equipment' and 'Services' as three representative up-, mid-and downstream sectors, respectively (Table 2).
For most provincial units, the exposure levels of the upstream sector, 'Mining and quarrying' were higher than those of the midstream and downstream sectors. The exposure values for most provincial units exceeded 0.5% of sectoral GDP. Moreover, the regions that are rich in mineral resources had high exposure levels in the 'Mining and quarrying' sector. Qinghai, Shaanxi, Ningxia, and Shanxi had the highest exposure levels in the sector, with values of 7.9%, 6.9%, 5.7% and 4.2% of sectoral GDP, respectively. The 'Mining and quarrying' sector in the four provinces depended heavily on the production demand of Hubei. Shaanxi and Shanxi were the top two origins of Hubei's intermediate inflows of 'Mining and quarrying', accounting for 35.0% and 26.0% of Hubei's intermediate imports of mining and quarrying products, respectively. On the other hand, for Qinghai, Shaanxi, Ningxia, and Shanxi, Hubei was ranked the second-, second-, fifthand sixth-most important destination for the intermediate outflows of mining and quarrying products, with 24.3%, 11.7%, 8.9% and 6.5% of the total amount of mining and quarrying products flowing into Hubei. In summary, Hubei imported substantial mining and quarrying products from western and central regions as intermediate inputs. The Wuhan lockdown disrupted downstream production and brought high economic exposures to the upstream sector.
Regions with high exposure levels in the sector, 'Manufacture of machinery and equipment', were concentrated in some central provinces geographically adjacent to Hubei and some eastern provincial units. The 'Manufacture of machinery and equipment' sector in these regions integrated more profoundly with Hubei's economy. Henan and Jiangsu were significantly exposed to the disruption of manufacturing chains of machinery and equipment, with values of 1.2% and 0.6% of sectoral GDP, respectively. Hubei was among the most important destinations for the final products outflows of 'Manufacture of machinery and equipment' from the two regions, with 44.5% (Henan) and 15.3% (Jiangsu) of their total final products outflows flowing into Hubei. Meanwhile, Hubei's final products inflows of 'Manufacture of machinery and equipment' mostly came from Jiangsu and Henan, with their aggregated shares exceeding 75% of Hubei's total final products inflows of this sector. As for the sector 'Manufacture of machinery and equipment', Hubei tended to be located at the midstream and downstream position of the production chain, and its lockdown led to high economic risk for its main providers and consumers.
The economic exposures of sector 'Services' were dispersedly distributed. Except for Hubei itself, the provinces with relatively high economic exposures in the sector of 'Services' included Qinghai, Shanxi, Ningxia and Shaanxi, which overlapped with the regions that also had high exposure levels in the aforementioned upstream and midstream sectors. As mentioned previously, given that the value-added of the services sector (particularly services 532 Kailan Tian et al. relevant to production activities) could be impacted indirectly through the transmission of secondary industries, the results for services are unsurprising.

DISCUSSION
Our results demonstrated that domestic regions (specifically the western and central regions) exhibited higher levels of risk than economies outside China. However, one may find from the National Bureau of Statistics that economies abroad, such as the United States and EU, exhibited heavier economic damage from Covid-19 than China, and the first-quarter GDPs of China's eastern regions were worse than those of western regions. These statistics seem to be inconsistent with the results of this study. This is because our approach only accounts for the economic losses that are attributed to the disruption of Wuhan-related value chains. It is not our approach, nor our purpose, to estimate the total economic costs of Covid-19 for every region or country. Besides the impacts propagating through value chains, the total economic costs of a region/country depend heavily on the severity and duration of the epidemic in that region or country, and the containment responses to Covid-19. China's eastern regions, for instance, had higher risks of Covid-19 transmission because of the higher population density and human mobility, and, therefore, were under more serious attacks by the pandemic. Consequently, the first-quarter GDP showed that the eastern regions suffered higher levels of economic damage than the western regions. Although our results indicated that the impacts of the Wuhan lockdown on economies abroad were relatively Economic exposure to regional value chain disruptions: evidence from Wuhan's lockdown in China 533 weak, it is still necessary to investigate impacts from a global perspective. The motivation of our compilation of a REXICIO table is to facilitate the assessment of the propagated economic impact along regional and global value chains. We provide a methodological framework in which we can link a small region (a province or city) to the global economy. Instead of Hubei, which is a domestic-oriented economy, we can apply the methodological framework to calculate the exposures to other exportoriented economies. If Wuhan is replaced with Guangdong, an export-oriented economy in China, the results derived from this methodological framework show that the exposure levels of economies abroad would be high, with the exposure values of Japan and South Korea, EU, and the United States being 1.0%, 0.4% and 0.2% in 2012, respectively. Considering that the annual GDP growth rates of these developed economies are relatively low (e.g., less than 2% in some years for the United States), such exposure levels are fairly high. Compared with Hubei, Guangdong exerts more demand for both intermediate and final products from economies abroad, and foreign economies also demand more products from Guangdong. Therefore, a lockdown of Guangdong would exert higher economic exposure on economies abroad. Due to data limitations, we did not split the sectors in the REXICIO table into processing trade industries and regular industries. About half of China's foreign trade fell under the processing trade regime in the past decades. Although its share has been declining in recent years, processing trade still accounted for 25.2% of China's total trade in 2019. In Hubei, processing trade accounted for a lower proportion (22.1% of its total trade in 2019). However, for the sectors 'Manufacture of machinery and equipment' (35.2% in 2019) and 'Manufacture of electronic product and electrical instrument' (37.6% in 2019), processing trade accounted for a relatively higher share. Processing trade mainly relies on China's labour and is commonly conducted by foreign-invested enterprises. It consumes less domestic intermediate inputs than regular trade and other production and thereby generates less value-added for China (Chen et al., 2019). This implies that our results would have changed slightly if a split of processing trade was made. The exposure levels of economies abroad might increase slightly because China's processing trade industries generate higher value-added content for foreign economies.
We used a partial extraction ratio to measure the severity of the value chain disruption during lockdown. We presented the empirical results for the case in which the Wuhan-related value chains completely crashed. This scenario was not perfectly realistic but was not far from reality given the fact that Wuhan City was under strict lockdown for almost three months and very few establishments could operate well during such a long strict lockdown (Duan et al., 2021). We also adjusted the value of the extraction ratio to reflect some real changes. For example, post-event information suggested that substantial agricultural products and medical supplies were delivered to Hubei during the lockdown. Hence, we performed some scenario analyses by remaining Hubei's demands for agricultural products unchanged and increasing its demands for medical products slightly by 10%. The results changed, but the changes were very small. The change in China's (Hubei's) exposure value was less than 0.01% (1%). This is because the increased demand for medicine was very small compared with the huge declines in other industries. The partial HEM can also be adapted to capture other scenarios if the requisite data are available.
Again, we argue that our case study serves as a prehazard estimation of regional economic exposures by extracting the entire Wuhan economy. Some assumptions need further adjustments if future studies would adapt the presented methodology to ex post estimate the economic losses of the Wuhan lockdown. For example, producers and consumers in Hubei and other regions may buy different bundles of products from before and they may buy products from different geographical sources during and after the lockdown. Consequently, such changes would exert different economic impacts across regions. More detailed post-hazard information and key parameters are necessary to estimate post-hazard economic losses. However, as introduced above, this is beyond the scope of this study as we restrain from setting too many arbitrary assumptions or choices regarding the values of massive parameters.

CONCLUSIONS
This study developed a methodological framework to evaluate the economic impacts of a regional shock. The economic exposure index was evaluated by applying the HEM to the REXICIO table incorporating China's domestic provincial information into the world ICIO table. Applying this methodology to the case of Wuhan lockdown, this study investigated the importance of propagation effects through value chains based on the empirical results at national, regional, and sectoral levels.
Our study has several policy implications. First, it is vital for regions that are highly exposed to a lockdown to respond quickly once the lockdown decision is made. The longer the lockdown lasts, the more sever the disruption of the value chains is. If appropriate responses are enacted quickly, it might be possible to strengthen the resilience of value chains and mute economic losses.
Second, it is necessary to narrow the scale of lockdown using scientific and transparent containment measures to minimize economic losses. Our results show that the lockdown of a city would cause not only serious economic costs for the city itself but also substantial propagated effects on other regions. One strategy for balancing epidemic containment and economic development is minimizing the area of quarantine via the timely identification of infected cases and accurate contact tracing. China adopted this strategy for fighting Covid-19; once infected cases were reported, the local government responded quickly by locking down a community or part of the city instead of locking down the entire city.
Third, both the central government and local governments need to strengthen regional coordination to aid a region in combating the epidemic and promote economic recovery. The strict lockdown measures in Hubei successfully mitigated the spread of Covid-19, and they did, somewhat, provide public goods for other regions. Our results showed that the lockdown limited China's national GDP exposure level to 2.5% while put 52.5% of Hubei's GDP at risk. Therefore, more supporting monetary policies and fiscal policies were provided for Hubei's post-disaster economic recovery. Moreover, after the lockdown was lifted, regional coordination was strengthened to restore production linkages with Hubei and recover consuming products from Hubei.
Finally, preferential measures should be taken for upstream industries and manufacturing industries to promote production resumption and strengthen the resilience of production chains. Secondary industries have longer production chains that are geographically fragmented and are highly exposed to the risk of production interruption. Therefore, taking timely measures to recover manufacturing production contributes significantly to reducing the negative impacts propagating along value chains, thus minimizing the economic damage to Hubei, China and the world.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.