Abstract
The COVID-19 pandemic has hit the world hard, costing more than three and half million lives. Governments around the globe are not in a consensus position on the most appropriate response to the pandemic. This study utilizes an economic model to assess choices and compare outcome of public health policies using China as a case study. A lax policy could have costed the country up to 97% of inbound tourism revenue; reduced real gross domestic product by 11% and decreased employment by 15%. Analysis shows that the appropriate prevention and control policy of the Chinese Government have mitigated the impact of COVID-19 significantly for both tourism and non-tourism sectors. Importantly, the article highlights that the substantial negative impact on investment in tourism will slow down the sector’s recovery. The article calls for strong tourism-focused response policies for a speedy recovery.
Introduction
The coronavirus disease in 2019 (COVID-19) struck Wuhan of China and had shifted to many countries across the globe by June 2020. Among all industries, tourism has been hit hard by COVID-19 across countries as a result of the public health policy that led to international border closures and effectively suspended all international travel. Within each country, domestic tourism has also been severely curtailed by lockdowns that aimed to contain then reduce the spread of the virus quickly. The United Nations World Tourism Organization (UNWTO) estimated that the loss due to the pandemic is 8.5 million to 1.1 billion in international tourism arrivals, equivalent to US$910 million to US$1.2 trillion in export revenue and 100–120 million jobs (UNWTO, 2020).
Evidently, the early internal lockdown in Wuhan effectively dealt with COVID-19 and recovered the domestic tourism market for China. Domestic tourism subsequently bounced back shortly afterwards. According to the China Tourism Academy (2020), the Qingming Festival (5–7/4/2020) attracted 43.25 million visitors, generating tourism receipts of 8.26 billion yuan. Similarly, the labour holiday (1–5/5/2020) recorded 48.81 million visitors and 12.28 billion yuan of tourism receipts (Central Government of China, 2020). By mid-July 2020, inter-provincial tourism was then allowed. During the 8-day national holiday of the mid-autumn festival (1–8/10/2020), the country recorded 637 million domestic visitors and 466.56 billion yuan in tourism receipts (Xinhua Net, 2020).
The research on COVID-19 impacts seems to fall into two broad categories. The most common one focuses more on the qualitative side of the pandemic, that brings out the key management lessons for policy makers to prepare for similar crisis in future (e.g. Foo et al., 2020; Gossling et al., 2020; Hall et al., 2020; Ioannides and Gyumothy, 2020; Jaipuria et al., 2020; Sigala, 2020; Ugur and Akbiyik, 2020). The other category involves more quantitative analysis, explicitly adopting economic models to measure and provide insights into the impacts across the whole economy from the pandemic (e.g. Mandel and Veetil, 2020; McKibbin and Fernando, 2020; Pham et al., 2021; Shan et al., 2020; UNCTAD, 2020). This quantitative approach requires explicit inter-industry linkages in an economy captured by an input–output (IO) database, particularly in this case, the computable general equilibrium (CGE) modelling technique given its ability to analyse complex government response policies, so we adopt a CGE modelling approach for this study.
Our article intends to evaluate the two different public health policies in the face of a pandemic. To combat COVID-19, many countries including China implemented local closures and lockdowns on top of the international border closures. However, there has always been a counter argument that challenges the trade-off between lockdowns and economy and between lost lives and lost livelihood associated with the public health policies. Given the more infectious and dangerous delta variant of COVID-19 than the previous variants and its devastating effects in India and other countries, it is imperative for policy makers to be aware of the two possible outcomes while the vaccination is not entirely completed to eradicate the situation. It is, thus, of interest and importance to evaluate the effectiveness of the response of the Chinese government to the health crisis. This article fills this gap.
The second contribution is that we develop a tourism CGE model to assess the role of tourism during pandemic and under the different policy responses. It is important to note that tourism could become a significant factor to aggravate a pandemic and that the tourism sector suffers most during a pandemic and propagates the negative economic effects throughout the economy. With the exception of Pham et al. (2021), the existing CGE studies on COVID-19 do not have a tourism sector in their models, so the role of the tourism sector and its implication for the economy is only implicit in these studies. The insights gained from the tourism modelling results in the present study can enrich tourism theory and provide practical suggestions and re-assurance for crisis management for policy makers and the industry.
The article is organized as follows. The Literature Review section briefly reviews the previous studies on pandemics and tourism. The Model and Data section describes the model and data used for this study. The Modelling COVID-19 and Tourism section explains how the pandemic and tourism are simulated, followed by interpretation of the simulation results in the Investment Shock section. The Conclusions section concludes the article.
Literature review
Due to the significant impact of the COVID-19 pandemic, research on this topic has been prompt and vast. This section organizes reviews by groups of previous to recent studies separately, with an explicit focus on tourism.
Studies on previous epidemics/pandemics
There have been a number of pandemics since 2000: severe acute respiratory syndrome (SARS) in 2002–2003, avian flu in 2005–2006, swine flu in 2009, Middle East respiratory syndrome (MERS) in 2012 and Ebola in 2013–2014. Most tourism studies have focused on SARS. Two studies used the detailed quantitative modelling approach. Hai et al. (2004) used survey data and an input–output model to assess the impact of SARS. They concluded that the tourism sector was hit hardest, with a 50–60% decrease in international tourism revenue, a 10% decrease in domestic tourism revenue and a loss of US$25.3 billion to the Chinese economy. Mao et al. (2009) used a cusp catastrophe model to examine the post-SARS recovery of Taiwan tourist arrivals from Japan, Hong Kong and the USA. They found that the perception of travel risk in the post-SARS period was different for tourists from different countries. They suggested a combined strategy for post-pandemic tourism recovery: a mass media campaign targeting the general public to project the image of safety in the destination country, as well as segmented and individualized marketing to reduce the risk perceived by individual travellers. The other studies that focused solely on SARS were Dombey (2003), Pine and McKercher (2004), Zeng et al. (2005) and Mason et al. (2005). They provided objective accounts of the event and assessment of the impact using data on changes in tourist arrivals, hotel occupation, etc. Zeng et al. (2005) and Mason et al. (2005) also emphasized the role of media in restoring the confidence of tourists in the post-pandemic era.
A few studies have considered other pandemics/epidemics. Kuo et al. (2008) and McAleer et al. (2008) used econometric estimation to compare the impact of SARS and avian flu on tourism demand. They found that SARS had a more significant impact. Also, based on document analysis and in-depth interviews of key stakeholders in the tourism industry, Qiu et al. (2018) compared the impact of SARS and the H7N9 outbreaks in China. They concluded that the social and economic impacts of H7N9 were not as serious as those of SARS. Joo et al. (2019) used a seasonal autoregressive integrated moving average model to estimate the impact of the 2015 MERS outbreak on the tourism industries in South Korea. Their results indicated an approximate loss of US$2.6 billion from international tourists, US$542 million in the accommodation sector, US$359 million in the food and beverage service sector and US$106 million in the transportation sector. Sifolo (2015) considered the impact of Ebola on the South African tourism sector and suggested a multi-sector approach to mitigating possible risks.
A study on foot and mouth disease (FMD) by Blake et al. (2003) used the CGE modelling approach – the same approach used in the current article. Although the FMD that struck the UK in 2001 was less serious to humans, it had a significant impact on cloven-hoofed animals (cattle, pigs and sheep). A micro-regional tourism simulation model was used to estimate the losses in tourist bookings and tourism revenue, which were then aggregated to a national level. The CGE modelling using this data showed that the UK government’s strategies for dealing with FMD caused a reduction in tourism revenue by ₤7.7 billion in 2001 and ₤5.2 billion in 2002. The GDP reduction ranged between ₤3.6 billion for 2001 and ₤1.6 billion for 2002. Based on these large adverse effects, they suggested that policy makers should have considered the implication on the tourism-related sectors when formulating the policy responses to the FMD outbreak.
Research on COVID-19 pandemic
The impact of COVID-19 was quickly felt by the tourism industry. In February, the Australian Tourism and Transport Forum (TTF, 2020) estimated that the pandemic would cause a 90–100% decrease in tourism revenue from China. From March 2020, a number of academic studies and official assessments have been published on this topic.
Most research on COVID-19 has focused on global effects or regional comparisons based on available statistics. McKibbin and Fernando’s (2020) working paper appears to be the first academic paper on COVID-19, with its main aim being to model the impact of COVID-19. The assessment was based on simulations using the DSGE/CGE model developed by Lee and McKibbin (2004) and extended by McKibbin and Sidorenko (2006). The model has six sectors and 24 countries/regions that were primarily adopted from the GTAP database (Aguiar et al., 2019). Due to the uncertainty in the development of the disease at the time of their study, as many as seven scenarios were simulated. A series of shocks are imposed, including labour supply, equity risk premium, cost of production in each sector, household consumption demand and government expenditure. The sizes of these shocks were based on the assumptions of the different infection and mortality rates for China, with the 2003 SARS outbreak being used as a benchmark. The shocks for other countries and regions were obtained by scaling the shock for China by an index of vulnerability or a country risk index. This index is the average of the three indices of governance risk (PRSGroup, 2012), financial risk (Fisman and Love, 2004) and health policy (GHSIndex, 2020). Their simulation results showed COVID deaths of 279,000 to 12,573,000 for China and 279,000 to 68,347,000 for the world. The death tolls for China were substantially overestimated partially due to the use of the SARS outbreak as a benchmark for their shocks. We now know that the death rate from COVID-19 in China is much lower than that from SARS. However, their findings demonstrated the GDP decreases were about 0.4–6.0% for China, 0.1–8.4% for USA, 0.3–9.9% for Japan and 0.2–8.4% for the Euro zone. These results are roughly in the ballpark of current official forecasts by international organizations.
Shan et al. (2020) used a global adaptive multiregional input–output model to estimate the economic and environmental effect of COVID-19 and scenarios of lockdown and fiscal stimulus packages. Their paper concluded that due to COVID-19, global emissions from economic sectors will decrease by 3.9% to 5.6% in 5 years (2020 to 2024). Supply-chain contributed 90.1% of emissions decline from power production in 2020. Fiscal stimuli in 41 major countries increase global 5-year emissions by −6.6 to 23.2 Gt (−4.7 to 16.4%), depending on the strength and structure of incentives.
Vasiljeva et al. (2020) developed a quarterly dynamic CGE analysis by extending his previous global pandemic research and using GTAP modelling framework. In this study, the world economy was stratified into 27 regions, with 30 sectors in each region. The modelling took into consideration risk-modifying behaviour, in terms of preventive measures, in response to the pandemic. The results show the most severe impacts of the world GDP in the second quarter of 2020, with −2.97% loss. Gradual recovery of the world GDP growth is projected for 2021, a modest growth of 0.98% in the third quarter of 2021. For Australia, it is projected to have its largest decline in GDP by −2.97% in the second quarter of 2020, followed by a slow recovery through 2021.
Mandel and Veetil (2020) used a multi-sectoral disequilibrium model with 56 industries and 44 countries to carry out computational experiments to replicate time sequential lockdown of different countries. Results indicate a fall in world output by 7% at the early stage of the pandemic. They also found that supply-chain spillover effects could amplify the effects on the world economy. Using a modified version of the GTAP model, Walmsley et al. (2021) estimated the macroeconomic impacts of mandatory business closures in the US economy and many other countries. Their 3-month closure of businesses scenario predicted 20.3% decline of US GDP (i.e. US$4.3 trillion) on an annual basis. The associated employment loss in the US is projected to be 22.4%.
Pham et al. (2021) models the adverse impacts due to the international border closures on the Australian economy. GDP of the country is estimated to decline between 2 to 2.2% for 2020, adding between 3.2 to 3.45 percentage points of unemployment rates to the economy. This study highlights the fact that job losses not only in the unskilled group but also in the skilled groups such as managers, professional, technician and trades.
The United Nations Conference on Trade and Development (UNCTAD, 2020) also used the GTAP model to gauge the economic impact of COVID-19 and used a range of assumptions regarding inbound tourism expenditure. GTAP database version 10 for 2014 was updated to 2018. ‘Accommodation, food and services’ and ‘recreation and other services’ were used as a proxy for tourism. Three scenarios were examined: a moderate (optimistic) scenario with a one-third reduction in annual inbound tourism expenditure, an intermediate scenario with a two-thirds reduction in inbound tourism and the most severe scenario with a total loss of all inbound tourism for the whole year. The results showed that tourism-oriented countries would experience a drop of around 10% in GDP and of around 15% in unskilled employment even in the most optimistic scenario. Employment could drop as much as 29% in the intermediate scenario and up to 44% in the severe case. The report calls for governments to protect people and maintain a healthy tourism industry.
Skare et al. (2020) used a panel structural vector auto-regression model to estimate the worldwide tourism impact of COVID-19. Their modelling results based on panel data of 185 countries from 1995 to 2019. They suggested that the pandemic would cause a US$4.1–12.8 trillion decline in the contribution of tourism to GDP and a decrease in 164.506–514.080 million jobs in contribution to employment. The loss in inbound tourism receipts was in the range of US$604.8 billion–US$1.9 trillion, and a fall in capital investment was between US$362.9 billion–US$1.1 trillion. They suggested coordinated private and public policy support to assure capacity building and operational sustainability of the tourism sector.
By reviewing relevant studies, Gossling et al. (2020) compared the impact of COVID-19 with previous epidemics/pandemics and other types of global crises. They also assessed the impact of the measures to counteract pandemics (travel restrictions and social distancing) on tourism arrivals and various segments of the tourism sector. They concluded that tourism is especially susceptible to counter-pandemic measures. By investigating the direct and indirect impact of tourism on pandemics, they claimed that tourism has significantly contributed to pandemics. For example, travel has spread the virus while the industrialized food production pattern for tourists has been responsible for repeated outbreaks of the coronavirus. They argued that these effects are similar to the effect of climate crisis on pandemics. Their logic is as follows: an increase in deforestation has interfered with wildlife and resulted in climate change. This has led to human migration and displacement and thus contributed to pandemics. As a result, they concluded that current tourism growth model is not sustainable, and the COVID-19 pandemic may be a catalyst for change in the tourism sector.
Other academic papers on COVID-19 include Sigala (2020), Foo et al. (2020), Ugur and Akbiyik (2020), Williams (2020), and Jaipuria et al. (2020). Using artificial neural networks, Jaipuria et al. (2020) estimated a substantial decrease in foreign tourists’ arrivals and foreign exchange earnings for India. Williams (2020) utilized a 2019 Eurobarometer survey and a probit model to estimate the impact of COVID-19 on the tourism workers in the undeclared economy. He found that 0.6% of all European citizens had undertaken undeclared work in tourism industry. He argued that they were impacted by the COVID-19 heavily and were not covered by government financial support. Following the line of argument that tourism contributes to pandemics put forward by Gossling et al. (2020), Hall et al. (2020), and Ioannides and Gyumothy (2020), Sigala (2020) argued that the COVID-19 pandemic should be viewed as a transformation opportunity for the tourism industry and tourism research. E-tourism and virtual services were viewed as alternatives. The article also questioned the effectiveness of government efforts toward economic recovery (e.g. stimulus packages, tax reliefs and subsidies). Foo et al. (2020), on the other hand, stated that government stimulus could pull the tourism industry in Malaysia through the crisis. Using text-mining techniques, Ugur and Akbiyik (2020) examined the impact of COVID-19 on the tourism sector. They found that the tourism sector is significantly affected by global crises such as pandemics and suggested that tourism insurance packages could reanimate the industry.
The review highlights the appropriateness of the CGE modelling technique for the purpose to examine the effects of response policies by the Chinese government in handling COVID-19. However, it is important to note that COVID-19 tends to affect the economy through tourism then to all other sectors. Existing CGE models do not explicitly capture tourism, with the exception of Pham et al. (2021). As such, the review orientates our approach to using a single-country model with an explicit tourism module so that policy analysis can be more direct.
Model and data
The CGE model used for this study is a single-country model for China, built upon the ORANI-G model (Horridge, 2000). The ORANI-G model is a standard single-country CGE model (including various economic agents in an economy), but it is modified for the Chinese tourism and economy. The tourism demands are tailored to Chinese tourism market and COVID-19. The data and parameters (elasticity values and various shares and ratios) are chosen to reflect the Chinese economy.
There is no tourism demand function in the original ORANI-G model, so we added two types of tourism demand: inbound tourism and domestic tourism. Both demands are modelled through downward-sloping functions of constant elasticity. The inbound tourism demand Dit is negatively related to inbound tourism price Pit and exchange rate ER (defined as the price of RMB in terms of foreign currencies),and positively related to a variable Mit – the motivation to travel. External shocks of safety issues, travel restrictions or new attractive tourism products may negatively or positively affect this variable. The level-form demand function of inbound tourism can be expressed as
Similarly, the domestic tourism demand Ddt is negatively related to domestic tourism price Pdt and positively related to a variable Mdt – the motivation to travel. The level-form demand function of domestic tourism can be expressed as
When modelling COVID-19, the labour supply function is an essential component of the modelling task, as the pandemic has led to quarantine/hospitalization, mortality and morbidity, all of which negatively affect labour supply. In the model, labour supply S is expressed as a positive-sloping function of constant elasticity. It is positively related to real wage W and negatively related to the COVID infection Ra, mortality Rm and lockdown factor LD. Mathematically, we have
The percentage change form of labour supply function can be written as
All other functions in the model are the same as in the original ORANI-G model. For example, production is specified via a combination of the Leontief function and the Constant Elasticity of Substitution (CES) functions. At a more detailed level, intermediate inputs are sourced from either domestic or imported sources, depending on the movements in relative prices. This substitution effect is governed by a CES function that allows industries to minimize their production costs. Industries are also assumed to choose a bundle of primary inputs using the CES nest to reduce production costs as much as possible. Investment demand is modelled using the nested Leontief-CES function. Household demand function is modelled using a linear expenditure system that incorporates income level, commodity prices and the size of the population in the economy. Export demand is a downward-sloping demand of world prices of goods and services. Government demand is linked to household consumption, but also can be determined by fiscal policy when required.
The main data for the modelling include an input–output database for the Chinese economy, tourism expenditure data and various behaviour parameters. The input–output data in this study are the latest 2017 Chinese Input–Output Tables, published by the National Bureau of Statistics of the People’s Republic of China (NBSPRC, 2019). The IO data were updated to 2019 using the RAS procedure, 1 based on the 2019 GDP and consumption data. The original database has 149 sectors and commodities. For the purpose of this study, we aggregate them to 29 sectors: one health sector, 13 tourism-related sectors and 15 other industries in the rest of the economy.
Information for 2019 inbound and domestic tourism is provided separately by the State Council of China (2020). For the inbound segment, the number of tourist arrivals was 145 million visitors, generating US$131.3 billion in tourism revenue (equivalent to RMB 943 billion based on the exchange rate in 2019). For the domestic component, the number of visitors was estimated to be 6 billion, generating RMB 5725 billion. In this article, total tourism expenditure was further disaggregated down to different goods and services. For inbound tourism, the disaggregation was implemented using 2019 foreign tourism expenditure in 2020 China Statistical Yearbook (NBSPRC, 2019).
For the domestic market, the disaggregation is conducted based on the domestic expenditure pattern shown in China National Administration of Tourism (2007). As the source is well before the 2019 IO database, we updated this expenditure pattern based on the 2012 urban household survey data from the NBSPRC (2012), which provides useful data on cultural and recreational services. In the years leading to COVID-19 event, Chinese domestic tourists experienced a general trend of increasing spending on air travel and shopping and a decrease in water transportation and postage services. To reflect this trend, we utilized the only two available most recent tourism survey data. One was provided by the Beijing Tourism Development Committee (Xing and Yu, 2018) and the other by the People’s Government of Hainan Province (2012). We updated the national pattern to reflect these trends. The tourism shopping pattern was supplemented by the findings of Chang et al. (2018), who analysed the data on online tourism commodity shopping in Beijing. The shopping items are classified into three main types: food products (dried fruits, alcohol, tea, meat products, pastry, etc.), gifts (porcelain, textile and leather products) and printing products (books, notebooks, diaries wall planners).
The elasticity of tourism demand is a key parameter in this study. Song et al. (2012) estimated the price elasticities of inbound tourism demand in China from 10 major origins, finding an average of −0.802. We adopt this elasticity value of inbound tourism in our model. For domestic tourism, Yang et al. (2014) estimated a price elasticity of −0.428 for tourists from urban areas and −0.307 for tourists from rural areas. Since the majority of domestic tourists in China come from urban areas, where economic development is growing strongly in relative terms, we use −0.4 in our model for domestic tourism demand. The price elasticity of labour supply is set as 0.06, which is based on the estimation by Qi (2010). The rest of calibration of the labour supply function is discussed in the Modelling COVID-19 and Tourism section. Other elasticity parameters in the model are adopted from the GTAP-E (Burniaux and Truong, 2002), including export demand elasticity and all Armington substitution elasticities for goods and services.
Modelling COVID-19 and tourism
The COVID-19 outbreak in China is examined in two scenarios: (a) an active prevention and control scenario and (b) a no-control scenario. The former scenario tries to mimic the responses and measures used by the Chinese government, while the latter scenario is a comparison case and is used to demonstrate a scenario where government took no action to combat COVID-19. The simulation involves a number of shocks.
Shocks related to labour supply
As discussed in the previous section, COVID-19 affects labour supply in the model through three factors: COVID infection, mortality and lockdown measures. In the prevention and control scenario, the rates and lockdown effects are calculated based on relevant data in China. By 6 October 2020, China had 90,700 confirmed cases and 4,739 deaths (Statista, 2020), so the mortality rate in terms of the confirmed cases is 4739/90,700=5.5%. Based on the Chinese population, we can calculate the infection rate as 0.907 million/1.3 billion=0.007% and the mortality rate in total population as 4739/1.3 billion=3.6*10−6.
It may be argued, however, that because COVID infection and mortality appears to be skewed towards older people, the impact of COVID may not contribute directly to a decrease in labour supply. however, COVID can affect the younger family members of patients through emotional response, stress endured, as well as caring activities. Effectively, COVID-19 infections, even in older patients, can indirectly affect labour supply. To minimise complication of the simulations, we assume that the COVID infection rate and death rate for the whole population are the same for the labour force, and thus, these rates indicates the percentage decrease in labour force.
The elasticity values of labour supply with respect to the infection and mortality rates are estimated as follows. Based on the percentage change form of labour supply function in the Model and Data section, the elasticities for the infection rates and mortality rate show how these rates affect the amount of labour inputs supplied. A person testing COVID-19 positive needs to self-isolate and/or stay in hospital, as well as have some time for post-treatment recovery. We estimate that a COVID patient would lose about 3 months out of one person’s working year. Based on this figure, we assign the value for elasticity related to infection rate as 3/12=0.25. A death will attribute a permanent decrease in population and thus labour supply, so we assign a value of 1 for elasticity related to the mortality rate.
The effect of lockdown on labour supply is directly expressed as the number of lost working days in a year, so the elasticity of lockdown should also be 1. The lockdown effect can be estimated through degree of impact on economic activities and the lockdown duration. The lockdown in China is a soft lockdown at city or province level; that is, there are restrictions between some cities/provinces so people can still go out but cannot exit a hotspot city like Wuhan. Furthermore, not all sectors are affected by the lockdown and the degrees of impact vary. For example, online shopping and post/delivery service activity increases substantially while hotels lose almost all costumers during a lockdown. Considering that the Chinese retail sales decreased by 19% in the first quarter of 2020 with about 2 months lockdown period (end of January to end of March), we assume the lockdown results in a 30% loss across economic activities and thus a 30% decrease in labour demand/supply. The lockdown duration in China was about 2.5 months (from late January to early April 2020). In considering a modelling time frame of 1 year, the impact of lockdown on annual labour supply should be reduced by 2.5/12. As a result, the lockdown would cause a loss in annual labour supply by about 30%*2.5/12=6.2%. This estimation is largely consistent with the unemployment data released by China.
For the no-control case, no lockdown occurs, and therefore, the infection rate and mortality rate would be much higher. Medical scholars estimated an effective reproduction number of 3.54 (i.e. 1 patient on average causes 3.54 new cases; see Hao et al., 2020) and an asymptotic case fatality risk of 1.4% (Wu et al., 2020). While these estimations are useful for projecting the disease transmission dynamics, they are not suitable for the no-control case. With no control measures, everyone would eventually be infected (with the exception of a very small portion of the population who are somehow immune to the virus), so we assume a case infection rate of 95% (with 5% of the population being naturally immune to the virus). Adopting the same mortality rate per infected case of 5.5% as presented earlier, the mortality rate for the total population is 95%*5.5%=5.2%. To calculate the labour supply shock for the no-control case, we use the same elasticity values used in the prevention and control case.
Shocks related to tourism demand
Tourism demand is affected by COVID-19 mainly through two channels. First, visitors voluntarily refrain from travelling due to a concern of catching the virus during their trips. The other is the travel restrictions imposed by governments to reduce community transmission of the virus in the absence of a vaccine. Given the international spread of the outbreak, the Chinese government banned international visitors from most countries and imposed a compulsory 14-day hotel quarantine upon arrival, which meant that the inbound tourism market in China is almost ended. Consequently, no international tourism data at the national level are available for 2020. At the provincial level, Statistic Bureau Shanghai (2020) reveals that inbound arrivals to Shanghai (the second largest city in China) in March 2020 decreased by 94%. Using this as a benchmark for the country level and considering that combating COVID worldwide may improve by the end of 2020, we impose a 90% decrease in inbound tourists’ willingness to travel for the protection and control case.
Domestic tourism is a quite different story, given the relatively softer travel restrictions and the better control of COVID transmission within China. According to the Ministry of Culture and Tourism (2020), domestic tourism demand decreased by 77% in the first half of 2020. For the second half of the year, the estimate for domestic tourism demand can be indicated by the 8-day national holiday statistics (Xinhuanet, 2020): 637 million visitors (about 79.0% of total domestic visitors in 2019) and tourism revenue of 466.56 billion yuan (69.9% of revenue in 2019). These numbers indicate that there would be a 30% reduction in the second half of 2020 for both visitor numbers and revenue. For the whole year, we take an average of the two estimates as the annual reduction: (0.77+0.30)/2=51.4%. Thus, we impose a 51.5% reduction on domestic tourists’ willingness to travel. This projection is consistent with the projection of a 52% decrease in tourism revenue by the China Tourism Academy (2020).
If COVID-19 was not under control in China, it is reasonable to project that the impact of COVID-19 on domestic tourism would be very similar to that on inbound tourism. Based on this reasoning, we assume a 90% reduction in willingness to travel for both inbound and domestic tourism in the no-control scenario.
Health service demand shock
Testing potential patients and treating those with COVID-19 has incurred large costs. The cost of the additional health services required is projected to be about 80.5 billion yuan (Tang et al., 2020). The percentage of total health consumption of these additional costs is calculated. Based on the 2019 GDP of 99.0865 trillion yuan and the ratio of health consumption to GDP of 6.6%, the estimated health consumption in 2019 would be 99.09*6.6%=6.5 trillion yuan. So, the additional health consumption caused by COVID-19 is 0.0805/6.5=0.0124=1.24%. This is the shock we impose for the prevention/control case.
The shock for the no-control case is calculated based on the assumption that if the Chinese government did nothing to control the pandemic, the situation for the whole country would be similar to that of Hubei Province, which was hit hardest by COVID-19. Since Hubei Province accounts for about 4% of China’s GDP and population, we use this share to scale up the cost of the 80.5 billion yuan estimated for the prevention and control case: 80.5/4%=2012.5 billion=1.013 trillion yuan. Using 2019 health consumption as the base, we calculate the health consumption shock as 1.013/6.5=15.6%.
Shock on household consumption
Household consumption is greatly affected by COVID-19. There are several reasons for this. One reason is that the disease reduces consumer confidence: with a looming economic recession, households increase savings and reduce consumption. Another reason is that with travel restrictions, households are less likely to engage in outdoor activities. There is also a significant impact from loss of jobs and thus having less income to support consumption. According to PWC (2020), the total retail sales of consumer goods in the first quarter of 2020 reduced by 19%. Consequently, we impose a −19% shock on the household’s willingness to consume for the no-control case.
Thanks to the Chinese government’s adequate response to COVID, the confidence of consumers has gradually recovered. The government also tried to stimulate consumption directly. For example, the government issued 19 billion RMB coupons (China Central Government, 2020). As a result, consumption in China has recovered gradually and steadily. According to Zhang (2020), retail sales decreased by −20.5% in January and February compared with the same periods in 2019; however, in the months from March to July, the gap reduced to −15.8%, −7.5%, −2.8%, −1.8% and −1.1%, respectively. In August, retail sales were 0.5% higher than for the same period in 2019, making a −8.7% average monthly change from January to August. With an expected increase in retail sales for the rest of the year, we assume the reduction in consumption will continue to narrow by 2%, so we impose a −6.7% shock on household consumption.
Investment shock
According to Tan and Cheng (2020), total fixed asset investment declined by 16.1% in the first quarter of 2020, so we use this as the investment shock for the no-control case. In reality, the Chinese government put considerable effort into increasing investment. According to the Ministry of Finance of China (2020), government investment in 2020 amounted to 13,400 trillion yuan. Given the 2019 total government investment was 9,240 trillion yuan, government investment increased by 31,600/92,400=30%. This government investment strategy has paid off. The fixed assets investment from January to August only decreased by 0.3% compared with the same period in 2019. Assuming continuing improvement of investment for the rest of 2020, we impose a shock of −0.2% on total investment.
Productivity shock
COVID-19 can affect productivity or efficiency in different ways. Unpredictable hospitalization and sick leave disrupt the stability of the labour force. Rescheduling shifts and failing to fill production roles could lead to higher costs and lower productivity. The instability of supply of inputs and the uncertainty of demand may also hinder the operation of firms. In some cases, the stopping and restarting costs and the cost of training of new workers due to high turnover can also reduce productivity. Travel restrictions and social distancing measures may also add to the costs of production.
There is no empirical base for determining the size of shocks to productivity. Generally speaking, in the prevention and control case, due to the much shorter duration of the COVID pandemic and much lower infection and mortality rates, the disruption to production is shorter and lighter, so we impose a 1% decrease in productivity for this case. In the no-control case, however, there is a longer period of disruption with higher infection and mortality rates, so there would be more disruption to production and thus a greater impact on productivity. As a result, we impose a 3% decrease in productivity for the no-control case.
Shocks for COVID-19 simulation.
Simulation results
The above shocks are implemented using GEMPACK version 12 to simulate the two scenarios. Since a 1-year time frame is considered, the short-run simulation closure is used. Specifically, capital is assumed to be unchanged and immobile between sectors while labour input can move across sectors, leading to equalization of wages in the economy. The exchange rate of the Chinese currency is pegged to a basket of selected currencies selected by the Chinese government, so a fixed exchange rate regime is adopted for the simulation. To check the robustness of the simulation results, sensitivity tests were performed and showed that the results are only marginally sensitive to the change in parameter values. We group the results into two categories: macroeconomic effects and sectoral effects. All results listed in this article are expressed as percentage change after the shocks.
Macroeconomic effects
The overall economic effect on an economy can be indicated by real GDP or real gross national product (GNP), which is shown in Figure 1. Impact of COVID-19 on GDP, GNP and employment. Note: GNP: gross national product.
The GDP results show a large impact of COVID-19 in both scenarios: a 4.18% decrease in the prevention and control case and a 10.95% decrease in the no-control case. Considering the comparative static nature of the model and the 6.1% Chinese GDP growth rate in 2019, the modelling results suggest that Chinese GDP growth after COVID-19 should be about 2% in the prevention and control case and −5% in the no-control case. The GNP figures are smaller than the GDP figures because the impact on international trade (exports and imports) is not included in the GNP. Nevertheless, the GNP result also shows the large difference between the two scenarios, so the results of both the real GDP and real GNP strongly endorse the prevention and control approach. This message is reinforced by the employment results: −6.56% in the prevention and control case in contrast to −15.02% in the no-control case. These results clearly suggest that the lax approach by many governments in the world is not tenable even from the perspective of economic growth.
The negative results on the GDP deflator indicate a slight decrease in price level in both scenarios. The price change results from both the supply side and demand side. While the decrease in demand in the wake of COVID-19 tends to depress prices in the economy, the supply-side factors tend to result in cost-push inflation. For example, the substantially reduced labour supply will increase labour costs while the decreased productivity due to production disruption will implicitly increase production costs. The factors on both sides may have a different influence on prices of different commodities but, overall, the GDP deflator suggests that the demand side dominates. Adding the percentage change in GDP deflator to the percentage change in real GDP, we can obtain the percentage change in nominal GDP, which indicates an even larger negative impact.
Figure 2 shows the impact of COVID-19 on household consumption, investment demand and tourism demand. The changes in demand result from the changes in tastes and sentiments during the pandemic. In the simulation, the taste and sentiment changes are implemented through shocks to the CGE model. Compared with the demand shocks imposed, the changes shown in Figure 2 are smaller. For example, there is a shock of −51.4% in domestic tourism demand for the active prevention and control case and −90% for the no-control case, but Figure 2 shows a decrease of 41.8% and 85.79% for domestic tourism in two scenarios, respectively. This is because, by excluding the price effect, Figure 2 shows the changes in demand in real terms. As demand decreases, the price decreases, so the cuts in spending indicated by the shocks are greater than the decrease in the amount of goods purchased. Impact of COVID-19 on demand.
Based on the effects in real terms shown in Figure 2, COVID-19 has significant impacts, and therefore, taking an active prevention and control approach can make a big difference. While consumption and investment amounts to over a 10% decrease in the no-control case, they reduce to about a third in the prevention and control case. The active approach also halves the negative impact on domestic tourism. Inbound tourism reduces by a similar amount in both scenarios. This largely results from the international travel restrictions in response to the uncontrolled COVID-19 cases in other countries, so this result does not diminish the effectiveness of the active approach by Chinese government.
The impact of COVID-19 on international trade is shown in Figure 3. Under both scenarios, the pattern of change is the same: the balance of trade deteriorates due to export volumes decreasing and import volumes increasing, so the balance of trade deteriorates. The terms of trade improve in both scenarios. This pattern can be explained by the increase in the price of exports. As stated earlier, the supply-side factors (e.g. the decrease in labour supply and productivity) tend to push up the prices, and the demand-side factors (e.g. the decrease in consumption, investment and tourism) tend to push the prices down. For export-oriented goods, the demand-side factors have little impact because the preference of foreigners is assumed not to change. As a result, the supply-side factors have a major impact and lead to an overall higher export price. This higher price deters foreign demand and thus export volume decreases. On the other hand, the higher domestic price means imports are relatively cheaper and thus the import volume increases. Impact of COVID-19 on international trade.
The terms of trade are the ratio of export price to import price. The import price is determined by the world market, which is held constant, so the increase in export price necessitates an increase in the terms of trade. Under the no-control scenario, the reductions in both labour supply and productivity are much larger, so the export price increases much more and the terms of trade are much higher. The larger export price hike also means a much larger reduction in export volume, a larger increase in import volume and a more significant deterioration in the trade balance.
Sectoral effect
It is widely accepted that an epidemic/pandemic will affect the tourism sectors more severely than other sectors. The simulation results show the severity of the impact of COVID-19 on each sector. To compare the tourism and non-tourism sectors, the results for the tourism sectors are organized towards the left and the non-tourism sectors are towards the right. We examine the sectoral effect in terms of output, basic price and investment.
The percentage change in sectoral outputs is shown in Figure 4. It is apparent that the 13 tourism sectors on the left (from the food and drink sector to the other business sector) suffer a much greater loss in output than the 16 non-tourism sectors on the right (from the agriculture sector to the health sector). Under the no-control scenario, the loss in output is generally less than 10% for the non-tourism sectors, but is 8% to 54% for the tourism sectors. The biggest losses occur in the restaurant, accommodation, trade service, rail transport, air transport and entertainment sectors, with outputs decreasing by 19% to 54%. The output changes in the non-tourism sectors vary mildly. While most experience output losses, the finance and insurance sector and the real estate sector obtain marginal output gains. Interestingly, although the other transport sector is not a tourism sector, it experiences a more than 8% loss in output, which is comparable to that for some tourism sectors such as the road transport, food and drink, and wood and print. This can be explained by the close links between the other transport sector and the tourism sectors: a large part of the other transport sector is to provide supporting services to road, air and rail transports. Impact of COVID-19 on sectoral output.
The prevention and control measures improve the performance of the tourism sectors greatly. The size of the output loss reduces from 54% to 20% for restaurant, from 28% to 14% for trade service and from 26% to 11% for rail transport. As the prevention and control measures implemented by the Chinese government take the form of a general approach (i.e. not tourism-oriented), this result shows that the tourism sectors are well integrated into the economy. The impact of these measures on the non-tourism sectors is also significant: the output losses are halved as under the prevention and control scenario. However, the prevention and control measures have little impact on the education, real estate, and finance and insurance sectors.
While output change is an important indicator of sectoral performance, it does not reveal the full picture of the stress faced by sectors. The rising costs and reduced demand due to COVID-19 squeeze industries from both sides. The reduced demand pushes down the commodity prices, so the producers are unable to pass on the increased production costs to consumers. However, if a sector can manage to partially pass on the increased costs to downstream industries or consumers, its financial performance will improve. Consequently, the price of outputs for each sector is another important indicator. We have different prices in the model due to different tax rates for different agents. For simplicity, we use the basic price (the price before tax) of each local product. The changes in basic prices are shown in Figure 5. Impact of COVID-19 on basic prices.
With a few exceptions, after the COVID-19 shock, the prices of tourism goods generally decrease while the prices of non-tourism goods increase. This pattern may result from two factors. One is that the negative tourism demand shocks are much larger than the other demand shocks. The other is that the tourism sectors are generally downstream, so they are unable to pass on their costs to other industries.
Focusing on the tourism goods in the no-control case, the restaurant, air transport and trade service sectors show a decrease in price by 43%, 32% and 19%, respectively. If we also consider the output effect shown in Figure 4, we find these sectors are under substantial pressure. For example, the restaurant sector experiences a reduction of output by 54% and a decrease in price by 43%, so the revenue losses in this sector will be 97%. Similarly, the revenue losses for accommodation, air transport, rail transport and trade services are in the range of 50–60%. The prevention and control measures ease the conditions for these sectors considerably. The two exceptions are the wood and print sector and the ICT sector, which experience a slight increase in output price in both scenarios. This may be due to tourism demand being only a small component of these sectors.
The positive price changes for most non-tourism sectors mitigate the impact of the output loss for these sectors. For the agriculture, coal, construction, education and health sectors, the price increase is higher than the output loss, so these sectors would avoid a revenue loss. The biggest winners appear to be the real estate sector and the finance and insurance sector. The prices of their products increase by 20–30% while the output increases slightly. The only anomaly is the other transport sector, which experiences a 30% decrease in price and an 8% decrease in output, so its revenue loss would amount to 38%. As explained earlier, this anomaly is largely due to its close link with the tourism sectors, so its performance is similar to that of tourism sectors. The prevention and control measures moderate the price change for all tourism and non-tourism sectors, which is consistent with the impact of sectoral output levels.
The prolonged impact of COVID-19 is indicated by its impact on sectoral investment, as shown in Figure 6. Since sectoral investment is determined by the profitability of the sector, the magnitude of change in investment also signals sectoral profitability. Impact of COVID-19 on investment.
Some tourism sectors experience a large investment reduction: restaurant by 86%, accommodation and rail transport by 57%, entertainment and trade service by 49% and air transport by 43%. With the sharp decrease in investment in these sectors, their capacity and profitability in the coming years will be negatively affected, so the recovery of these sectors would be slow. The prevention and control policy significantly improves the investment in these tourism sectors; however, the magnitude of the investment reduction is still large. For example, there is a 46% reduction for the restaurant sector and a more than 20% reduction for the trade service, rail transport, air transport, accommodation and entertainment sectors. Some tourism-oriented policies may be necessary to support the recovery of these tourism sectors.
For the non-tourism sectors, investment reduction under the no-control scenario is generally less than 20%. The exception is the other transport sector, with 37% reduction in investment. With a significant increase in investment, the finance and insurance sector and the real estate sector seem well positioned to strengthen their competitiveness. The prevention and control measures also moderate the reduction in investment to less than 10% for the non-tourism sectors, so the long-term recovery of these sectors seem less worrying.
Conclusions
Using a tourism CGE model with the latest IO database for the Chinese economy, we have simulated the impact of COVID-19 on the Chinese economy, with special reference to its tourism sectors. Due to lack of recent detailed national-level survey data on tourism expenditure, we used the survey data in Hainan province and Beijing to update the domestic tourism expenditure pattern. This limitation may be overcome if the detailed national survey data are available.
The simulation results show that without a pandemic control policy, Chinese real GDP would reduce by 11%, employment would decrease by 15% and both domestic and inbound tourism would decrease by about 88%. However, the prevention and control policy implemented by the Chinese government greatly mitigated the impact of COVID-19, as indicated by a 4.2% reduction in real GDP, a 6.6% increase in unemployment rate and a 41.8% reduction in domestic tourism demand.
The sectoral results show that tourism sectors are far more severely affected by COVID-19. Under the no-control scenario, the tourism sectors of restaurant, accommodation, entertainment, trade service, rail transport and air transport are in a very difficult situation, with output reductions ranging from 20% to 54%, revenue decreasing by 30% to 97% and investment decreasing by 40% to 86%. The non-tourism sectors experience much less impact, with the output reduction generally being less than 10% and investment decreasing by less than 20%. With an increase in the prices of most non-tourism goods, the revenue losses faced by the non-tourism sectors are even less than the degree of output decrease. The exceptional case is the other transport sector. Due to its close links with the tourism sectors, its output decreased by 8% and investment decreased by 37%. Although the prevention and control measures reduced the impact on all sectors significantly, the impact on the tourism sectors is still substantial.
The study has two policy implications. The first is that the prevention and control approach is the preferred response to an epidemic or pandemic. Currently, the justification for this approach is largely based on the argument of the value of lives. Without disease control measures in place, more people will die, and this loss of life cannot be offset by any economic gains. In spite of this argument and fearing economic loss, some countries have embraced a lax policy to encourage economic activities. However, even on the grounds of pure economic argument, the simulation results do not support a no-control policy. The Chinese experience, as well as the experience of Singapore, Korea, New Zealand and Australia, shows that a stricter policy capable of controlling the disease in a shorter period of time, consequently also minimizes detrimental economic impacts.
The second policy implication of this study is that during an epidemic/pandemic, the tourism sectors can benefit greatly from a general economic policy that cushions the economy, such as increasing government investment, stimulating consumption and tax cuts. However, the epidemic/pandemic will most certainly have a large, prolonged impact on the tourism sectors, so the recovery in these sectors will be slower than other sectors. As a result, a complementary tourism-oriented recovery policy is necessary to regain the confidence of the tourism industry and to enable a full and speedy recovery of the whole economy.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
