1 INTRODUCTION
The fast unfold of a novel coronavirus (COVID-19) has prompted many nations to impose lockdowns and different social distancing measures to gradual the transmission of the virus. It isn’t clear, nonetheless, that these restrictions on the motion of people really alter individuals’s habits. It’s attainable that customers curtail their financial exercise in response to fears of the virus, even when this isn’t mandated by regulation. For instance, Goolsbee and Syverson (2021) use cell phone information to indicate that lockdown orders in the US accounted for less than a small quantity of the change in financial exercise induced by the COVID-19 pandemic.
On this research, we estimate the impact of China’s native pandemic insurance policies, specifically well being emergency declarations and lockdowns, on client habits. We give attention to the meals business, which on common accounts for one-fourth of client spending all over the world, with the share being greater in a growing nation setting like China (United States Division of Agriculture, 2021a). Particularly, we look at whether or not shoppers reply to emergency declarations, that are “info” occasions, otherwise than to “mandated” lockdowns. We manually accumulate the dates of emergency declarations and lockdowns in 51 main cities and make the most of the differential timing of those native insurance policies to quantify the impact on the wholesale costs of 36 key meals merchandise that represent about 65% of Chinese language shoppers’ meals expenditures.
Utilizing an event-study methodology that controls for seasonality in meals costs, we discover that day by day meals costs started climbing after emergency declarations, rising over 14% on common by the fourth day following a declaration, which is equal to 7.8 customary deviations of the worth change distribution.1 The consequences are most pronounced for non-perishable greens. This means that emergency declarations engendered expectations of additional governmental actions in response to the pandemic and led residents to “panic shopping for” of non-perishable meals merchandise, doubtlessly reflecting considerations about future shortages. The impression of emergency declarations, nonetheless, seems to have been short-lived, as shoppers roughly reverted to the established order in 10 days. Customers have been prone to have eased panic shopping for after they came upon that the majority meals objects remained out there.
We argue that offer chain disruptions weren’t the primary driver of our outcomes. First, the emergency declarations coincided with the beginning of the Lunar New 12 months (January 25, 2020), throughout which most wholesalers in China inventory stock for a minimum of 1 week.2 Second, if provide chains have been severely disrupted, we might have noticed a big enhance within the costs of perishable merchandise, which usually have extra weak provide chains and might be saved for a comparatively shorter interval than non-perishable merchandise (Ihle & Rubin, 2013).
One may argue that even when there have been no official well being emergency declarations, we might see the identical client habits as a consequence of fears of the virus, as noticed by Goolsbee and Syverson (2021) within the U.S. Though the negligible change of costs after the primary COVID-related loss of life in Wuhan helps the notion that value responses round emergency declarations are pushed largely by coverage, it’s attainable that residents didn’t be taught sufficient in regards to the virus earlier than official declarations. To deal with this chance, we repeat our evaluation round native lockdowns. Most lockdowns have been imposed in February 2020, when new case counts in our pattern cities had already peaked and commenced to drop. We discover that meals costs jumped instantly after lockdowns have been imposed and remained elevated till the top of the occasion window, 2 weeks after a lockdown. Non-perishable merchandise skilled a big enhance in costs, greater than 8% above their 2019 ranges, suggesting that there was a sustained enhance in demand for non-perishables. The consequences of lockdowns last more however the peak magnitude is 3.6 customary deviations of the worth change distribution, which is smaller than that following post-emergency declarations, suggesting that customers deliberate forward and hoarded non-perishables, as they anticipated to buy groceries much less regularly sooner or later.
Along with displaying sizable common results of lockdowns and emergency declarations, we additionally look at heterogeneity in demand shocks throughout merchandise, which has implications for client welfare. We discover that inside per week’s time following emergency declarations, 12 of the 19 pattern greens skilled value will increase of greater than 7 customary deviations of the worth change distribution. Costs of Chinese language cabbage and potatoes, the 2 most closely consumed greens (accounting for 40% of vegetable consumption), elevated by as much as 17.0 and 14.0 customary deviations, respectively, of the worth change distribution. The magnitudes of the results for many meat merchandise are usually a lot smaller. The heterogeneous results of lockdown orders are qualitatively comparable. As households spend a good portion of their meals budgets on greens, our outcomes counsel that giant COVID-19 policy-driven demand shocks to vegetable costs led to important client welfare losses, particularly amongst low-income households that spend greater parts of their meals budgets on greens (Tian & Yu, 2015).
Moreover, we research whether or not there was any “anticipation impact” by analyzing cities going through early and late lockdowns individually. In cities experiencing early lockdowns (on or earlier than February 6, 2020), meals costs rose instantly following lockdowns and continued to rise throughout the post-event window. In distinction, in cities going through late lockdowns (after February 6, 2020), costs barely modified after lockdown orders have been issued. Curiously, we discover that costs in cities with late lockdowns responded considerably to lockdowns imposed in different cities in the identical province, suggesting that residents rationally anticipated that their cities can be locked down following different cities in the identical province, as policymaking is usually constant inside a province.
Due to the dearth of large-scale retail value knowledge in China, we use wholesale value modifications to check client responses to COVID-19 insurance policies. Throughout our pattern interval, the typical modifications in retail and wholesale costs are comparable, primarily based on knowledge from eight cities.3 Chinese language meals markets are characterised by numerous small-scale rivals at every commerce degree (e.g., wholesale and retail) and few of them have important market energy (Zhong et al., 2020).4 Therefore, value actions in Chinese language meals markets mirror modifications in provide and demand situations. In distinction, the meals retailing sector in developed nations is often dominated by a number of giant grocery store chains. Going through substantial managerial prices of various costs throughout institutions, grocery store chains are likely to set costs uniformly and are much less nimble in responding to giant demand shocks throughout areas (DellaVigna & Gentzkow, 2019; Hobbs, 2020). This distinction in market construction doubtlessly explains a considerable fraction of the distinction between our outcomes and people within the research of retail costs by Gagnon and López-Salido (2020), who discover that Hurricane Katrina led to a mere 1.6% enhance in retail meals costs in New Orleans.5
Our research contributes to a rising literature that examines client habits in response to the COVID-19 pandemic. The prevailing literature is dominated by research on shoppers within the U.S. (e.g., Alexander & Karger, 2020; Baker et al., 2020; Chetty et al., 2020; Coibion et al., 2020; Goolsbee & Syverson, 2021) and different nations (e.g., Andersen et al., 2020 for Denmark; Chen et al., 2020 for China; and Chronopoulos et al., 2020 for the UK). These authors discover that combination client spending on groceries elevated considerably throughout the COVID-19 disaster in contrast with spending on different items. Our setting allows us to distinguish comparatively transient panic-buying habits (in response to emergency declarations) from a sustained enhance in client demand (throughout lockdowns). The bigger response in China than within the U.S. to lockdown insurance policies (Goolsbee & Syverson, 2021) highlights the significance of institutional variations. Lockdowns have been applied in most Chinese language cities after the peaking of recent instances, which afforded residents sufficient time to react to the virus itself and permit us to disentangle the impact of lockdowns.6 Within the U.S., nonetheless, the orders have been imposed in response to the arrival of the coronavirus and the orders have been usually much less strict than these in China.
By highlighting policy-driven demand shocks, as recommended by the differential results on costs of non-perishable and perishable meals merchandise, our research is said to a nascent literature on the impression of COVID-19 on product costs. Cabral and Xu (2020) discover some proof of value gouging for facial masks and sanitizers in early 2020 utilizing knowledge from Amazon. Wang et al. (2020) use survey knowledge and present that customers’ willingness to pay for meals elevated throughout the pandemic. Our research enhances Cakir et al. (2021) and Mahajan and Tomar (2021), who research potential disruptions of meals provide chains attributable to the COVID-19 pandemic, with the previous specializing in India and the latter on the U.S. and China.
2 BACKGROUND
As of July 1, 2021, authorities in 221 nations and territories reported over 184 million COVID-19 instances and practically 4 million deaths (World COVID-19 Tracker, 2021). In early December 2019, the virus emerged within the metropolis of Wuhan, which is the capital of Hubei Province in China. On January 11, 2020, the Wuhan Municipal Well being Fee introduced the primary COVID-related loss of life and 41 infections, a report broadcast by Xinhua Information, China’s official information company. Responding to the quickly spreading virus, on January 20, 2020 Wuhan established a command middle for epidemic prevention and management, which issued an order to lock town down on January 23, 2020. The order suspended all public transportation within the metropolis, together with native buses, long-distance buses, subways, and ferries. Metropolis residents have been informed to not depart Wuhan, as all departing flights and trains have been briefly canceled. Inside 1 day, one other 14 cities in Hubei have been locked down. By January 26, 2020, all cities in Hubei Province, the place practically 60 million individuals reside, have been locked down.
The virus unfold shortly to a number of different Chinese language cities. The primary instances exterior of Wuhan have been confirmed on January 20, 2020 in Beijing and Shenzhen, and by January 23, 2020, 29 of China’s 31 provinces had reported infections.7 Outdoors of Hubei, many jurisdictions responded initially by declaring Degree 1 public well being emergencies, indicating that the best degree of mobilization is required and permitting provincial emergency administration places of work to reply and entry nationwide assist. Such declarations convey a way of urgency and seriousness relating to the scenario. In contrast to information experiences, which give factual details about the standing of the pandemic, emergency declarations are forward-looking and engender expectations of additional governmental actions in response to the pandemic.8 On January 23, Guangdong, Hunan, and Zhejiang provinces declared emergencies. In the future later, 13 further provinces, together with Hubei, elevated their emergency responses. By January 25, 2020, all 30 provinces, excepting the Tibet Autonomous Area, had made emergency declarations (Ma, 2020; see On-line Appendix Determine B1).
Following provincial emergency declarations, most municipal and native governments ordered residents to remain residence and away from their workplaces. The primary batch of cities exterior of Hubei, together with Sanmenxia, a metropolis in our pattern, introduced their lockdowns on February 1, 2020. By the top of February the governments had confined about 508 million individuals to their properties (Lin, 2020). Below such stay-at-home orders, residents may transfer into or out of their communities solely with legitimate IDs, whereas guests weren’t allowed to enter. Keep-at-home orders have been usually accompanied by faculty closures, gathering restrictions, and non-essential enterprise closures. In some areas, stricter lockdowns have been imposed and residents have been positioned beneath a type of home arrest or residence quarantine, with just one particular person per family allowed to exit as soon as day-after-day or each different day to purchase meals and important provides. We refer to those measures as “lockdowns” on this research. Determine 1 illustrates the timeline for the main occasions talked about above, together with the primary COVID-related loss of life in Wuhan, emergency declarations, and lockdowns.
Emergency declarations, lockdowns, and meals costs. On this determine, the vertical strains denote the timing of the primary COVID-related loss of life in China (Wuhan), the lockdown of Wuhan and different cities, and the declarations of well being emergencies. The grey line represents the typical wholesale meals value index in 2017 and 2019 (prior years) and the yellow line represents the typical wholesale meals value index in 2020. We plot value indexes by days relative to the Lunar New 12 months (Day 0). For every day, we common every product’s costs throughout cities earlier than averaging throughout merchandise. The common value on January 1 (Day −24) is assigned a worth of 100
Worry of ever-tightening restrictions prompted many residents to refill on meals and different important objects. For instance, on January 23, 2020 a number of Wuhan residents informed Reuters that residents have been speeding to stockpile provides, leaving grocery store cabinets empty of merchandise corresponding to meat, greens, and on the spot noodles (Reuters, 2020). Caught at residence and fearful that lockdowns may final weeks and even months, households throughout China additionally hoarded meals merchandise, doubtlessly contributing to a surge in meals costs, with month-to-month meals value inflation of 4.4% and 4.3% in January and February 2020, respectively (Nationwide Bureau of Statistics of China, 2020). We observe that meals provides have been unlikely to be considerably affected throughout the week following the emergency declarations. As reported by Li (2020), Shao et al. (2020b), and Xinhuanet (2020), wholesalers usually inventory sufficient stock for a minimum of 1 week’s value of gross sales across the Lunar New 12 months vacation, the beginning of which coincided with the well being emergency bulletins.
In contrast to within the U.S., the place grocery store chains rationed sure objects and set procuring hours for the aged to maintain provides out there, rationing was not noticed in China. This distinction is probably going attributable to the distinct construction of Chinese language meals markets. The dominance of small meals retailers and distributors in China means that any rationing rule have to be imposed by the federal government. It’s, nonetheless, difficult to determine rationing standards given the substantial heterogeneity in value modifications throughout meals merchandise that we observe, as proven in Sections 4 and 6. As well as, imposing rationing in a market that includes numerous aggressive gamers can be pricey, if not inconceivable. On February 1, 2020, the central authorities as a substitute issued a warning towards value gouging for private hygiene merchandise and important meals objects (State Administration for Market Regulation, 2020). The warning states that “retailers’ revenue margins of offered items could not exceed the margins on or earlier than January 19, 2020.” Though the wording appears restrictive, the assertion explicitly permits sellers to go by means of price will increase to shoppers.
The next lockdowns, nonetheless, threatened to have an effect on the meals provide chain adversely as many localities closed off roads to passing site visitors, slowing truck shipments and elevating freight prices. Some truck drivers have been pressured into quarantine as that they had traveled by means of uncovered areas. One of the important points with the motion restrictions was the query of tips on how to ship greens and fruits throughout cities (Shao et al., 2020a). China includes a excessive charge of meals self-sufficiency, nevertheless it depends closely on inter-province commerce due to uneven and seasonal manufacturing throughout areas. To make sure that meals objects and different items have been capable of go by means of roadblocks, the central authorities had established a “inexperienced channel” system by January 30, 2020 (earlier than all lockdowns apart from these in Hubei cities have been imposed), which allowed truck drivers to journey between cities with particular passes. Some native governments, corresponding to town of Chongqing, additionally employed truck drivers to maintain provides flowing (Wernau, 2020). Within the meantime, farms and slaughterhouses have been instructed to not shut down (Ministry of Agricultural and Rural Affairs of China, 2020a). As well as, Beijing rolled out different measures to assist producers restore operations, together with slashing rents and social insurance coverage funds and rising lending to small producers. Whereas provide chain challenges remained, these authorities insurance policies largely ensured a easily functioning provide chain throughout China.
This dialogue underscores three key factors that information our empirical evaluation. First, the emergency declarations and lockdown insurance policies have been designed to scale back contact between people and scale back the danger of virus transmission. These insurance policies would possibly, nonetheless, have generated sizable externalities within the meals market by means of value will increase, doubtlessly inflicting welfare losses for all residents. Second, if lockdowns throughout cities diverse in strictness, these imposing much less strict lockdowns might need skilled much less client hoarding, enabling them to keep up extra steady meals costs. Third, as cities imposed lockdowns sequentially, residents in cities not but locked down might need anticipated an eventual lockdown and begun to hoard meals merchandise earlier than official lockdown orders have been issued.
3 DATA
3.1 Meals costs
We compile a metropolis–product panel of wholesale meals costs for main Chinese language cities for the interval operating from January 1, 2020 by means of February 29, 2020 and its corresponding lunar calendar interval in 2019, operating from January 12, 2019 by means of March 12, 2019.9 Every day costs are obtained from the Chinese language Ministry of Commerce. The info file info on product identify, date, value, and places of wholesale markets. We combination the info to town–product degree by averaging costs throughout markets in the identical metropolis.
The preliminary dataset covers 289 merchandise and 103 cities. In our evaluation, we give attention to 36 main merchandise which can be monitored by China’s Ministry of Agricultural and Rural Affairs due to their significance in day by day meals consumption (Ministry of Agricultural and Rural Affairs of China, 2019, 2020b). These merchandise cowl 5 essential classes: greens, fruits, meat, eggs, and seafood. Utilizing per capita consumption knowledge from the China Well being and Diet Survey, we estimate that collectively they represent a minimum of 65% of shoppers’ meals expenditures in China.10
We then prohibit our pattern to 88 cities that reported costs for each 2020 and 2019, as product costs throughout the identical lunar calendar interval in 2019 are used because the comparability group. We exclude cities that reported costs for a minimum of one product on fewer than 36 days (60% of the pattern size) in both yr and cities that reported a minimum of one value lower than as soon as per week in both yr. We additionally require that every metropolis–product pair exists in each years.11 For a similar motive, we additionally drop two cities that reported the costs of not more than 5 merchandise per day all through our pattern interval. After these steps, our last knowledge set covers 51 cities (from 23 provinces) which can be residence to just about 25% of China’s inhabitants. The pattern contains 110,758 product × metropolis × day observations.
On every day of the pattern interval, all the 36 merchandise are included in our pattern. The composition of merchandise, nonetheless, varies throughout cities. For every metropolis, we depend the variety of merchandise which can be ever reported throughout our pattern interval (i.e., the utmost variety of merchandise). We estimate that every metropolis on common experiences costs for 21 merchandise. Every product is roofed by 32 cities on common, with no particular geographic distribution throughout the nation.
In On-line Appendix Determine B2, we plot, for every of the 51 cities, the day by day variety of merchandise with out there costs for 2020. The plots present that, for many cities, the variety of merchandise out there remained constant all through the pattern interval. The sample for 2019 is nearly similar. Nonetheless, a number of cities, together with Changde, Dalian, and Hangzhou, didn’t file costs for days across the Lunar New 12 months Day (January 25, 2020). This doubtless displays the truth that some market displays don’t report back to work throughout a very powerful vacation season in China.12 These omissions don’t seem like geography-specific and are unlikely to bias our outcomes. We additionally test the composition of merchandise in every metropolis and discover that it’s constant all through our pattern interval.
A number of cities, notably Jiaxing and Qingdao, skilled fluctuating product numbers over time. This sample may mirror stockouts, which can end in underestimating the results of emergency declarations or lockdowns, primarily as a result of, if the merchandise had nonetheless been out there, counterfactually their costs would have been elevated, resulting in bigger coefficients than when the costs have been lacking.13 In robustness analyses reported in On-line Appendix B1, we discover that our outcomes stay largely unchanged when utilizing a balanced panel for every product–metropolis pair (subsequently excluding metropolis–product pairs that embrace potential stockouts).
3.2 Provincial and metropolis authorities actions
We discover each emergency declarations and lockdown insurance policies on this research. Given the decentralized coverage responses of provinces and cities, there is no such thing as a single useful resource documenting such governmental actions. Our set of emergency declarations are collected manually from every province’s well being fee. As soon as a province declares a state of emergency, it applies to all of the cities inside the province.
For 44 cities we accumulate info on lockdowns from metropolis authorities web sites.14 For the 4 cities that didn’t challenge their very own lockdown orders, we use lockdown orders introduced by their respective provinces, as cities inherit provincial orders as soon as the latter are issued. Three cities, Anyang (Henan Province), Urumqi (Xinjiang Province), and Xiangxi (Hunan Province), didn’t challenge lockdown orders, nor did their provinces. We subsequently exclude them in our evaluation round lockdown dates. Emergency declarations and lockdown dates are listed in Appendix Desk A1.
We additionally accumulate info on lockdown strictness in every metropolis primarily based on the texts of lockdown orders. Of the 48 cities that imposed lockdowns, 22 restricted guests to residential communities and required residents to enter or exit with legitimate IDs. The remaining 26 cities enforced de facto home arrests and allowed just one particular person per family to go away their properties to buy day-after-day or each different day. We classify the previous as much less strict lockdowns and the latter as strict lockdowns. On-line Appendix Determine B3 plots the spatial distribution of lockdown strictness.
To evaluate whether or not we will research the results of emergency declarations and lockdown orders individually, we plot the cumulative variety of cities that adopted every coverage on a given day. In Determine 2, we observe that emergency declarations usually have been issued nicely earlier than lockdown bulletins, with a median distinction of 12.3 days. For the three cities in Hubei, although—Wuhan, Jingzhou, and Huanggang—lockdowns have been imposed both earlier than or on the identical days as emergency declarations. We thus exclude these cities when learning the results of emergency declarations.
Rely of emergency declarations and lockdowns. On this determine, we plot the cumulative variety of cities that issued emergency declarations and imposed lockdowns between January 1, 2020 and February 29, 2020
4 DESCRIPTIVE STATISTICS
On this part, we current abstract statistics for costs for every of the 36 key merchandise. For every product, we common its costs throughout cities each day. In Desk 1, we report every product’s imply value, median value, customary deviation, and its proportion change in value between January 1 and February 29, 2020. Based mostly on statistics proven in column (4), from January 1 to February 29, 2020, costs of the 36 merchandise elevated by 14.2% on common throughout the 51 cities, indicating important meals value inflation throughout the COVID-19 disaster. Within the final two columns of Desk 1, we additionally report whether or not a product is categorized as perishable or non-perishable and the share of a family’s meals funds spent on that product.15 Meat budgets account for the biggest share (30.8%), adopted by budgets for greens (17.5%), seafood (7.8%), and fruits (5.0%). Based mostly on per capita consumption, nonetheless, greens represent round 28% of Chinese language residents’ diets, as they devour greens at greater than twice the speed of meat and fruits (China Statistical Yearbook, 2018).
Abstract statistics
| Wholesale costs (yuan) | Perishability | % in client basket | |||||
|---|---|---|---|---|---|---|---|
| Imply | Median | Std. Dev. | Change (%) | ||||
| (1) | (2) | (3) | (4) | (5) | (6) | ||
| Greens | Bok choy | 3.56 | 3.56 | 0.57 | −5.13 | P | 0.69 |
| Cabbage | 2.35 | 2.32 | 0.37 | 29.59 | NP | 0.11 | |
| Carrot | 2.47 | 2.58 | 0.34 | 41.96 | NP | 0.26 | |
| Cauliflower | 3.49 | 3.54 | 0.40 | −5.41 | NP | 0.36 | |
| Celery | 3.36 | 3.38 | 0.52 | 19.95 | NP | 0.41 | |
| Celtuce | 2.89 | 2.82 | 0.32 | 19.28 | NP | 0.28 | |
| Chinese language cabbage | 1.99 | 2.18 | 0.45 | 80.80 | NP | 1.69 | |
| Cucumber | 5.95 | 5.97 | 1.01 | 18.48 | P | 1.06 | |
| Eggplant | 6.22 | 6.09 | 0.70 | 30.89 | P | 2.53 | |
| Garlic chives | 6.12 | 6.32 | 1.14 | −23.25 | P | 0.66 | |
| Garlic shoots | 10.01 | 10.28 | 1.10 | 45.45 | NP | 0.17 | |
| Beans, snap | 10.60 | 10.43 | 0.87 | 17.49 | P | 2.68 | |
| Onion, inexperienced | 2.94 | 2.82 | 0.56 | 28.74 | NP | 0.09 | |
| Onion | 3.27 | 3.31 | 0.33 | 17.35 | NP | 0.12 | |
| Peppers, bell | 6.62 | 6.45 | 1.05 | 75.96 | NP | 1.77 | |
| Potato | 3.08 | 3.28 | 0.38 | 47.56 | NP | 2.09 | |
| Spinach | 4.47 | 4.36 | 0.87 | −10.95 | P | 0.48 | |
| Tomato | 6.81 | 6.69 | 0.79 | 12.94 | P | 1.77 | |
| Turnip | 1.65 | 1.55 | 0.29 | 15.81 | NP | 0.29 | |
| Fruits | Apple (Fuji) | 6.89 | 6.89 | 0.16 | 2.03 | NP | 3.07 |
| Banana | 5.45 | 5.46 | 0.29 | 14.70 | P | 0.46 | |
| Grape (Kyoho) | 8.65 | 8.55 | 0.59 | 5.00 | NP | 0.51 | |
| Pineapple | 5.37 | 5.32 | 0.58 | −10.89 | P | 0.03 | |
| Watermelon | 5.10 | 5.13 | 0.38 | 22.42 | NP | 0.87 | |
| Meat | Beef | 73.31 | 73.34 | 0.89 | 3.52 | P | 5.10 |
| Rooster | 18.17 | 18.14 | 0.67 | −4.49 | P | 3.08 | |
| Mutton | 68.73 | 68.45 | 0.97 | 4.35 | P | 1.15 | |
| Pork | 48.47 | 48.76 | 1.62 | 12.37 | P | 21.50 | |
| Eggs | Eggs | 7.88 | 7.47 | 0.79 | −20.35 | NP | 4.01 |
| Seafood | Bighead carp (reside) | 13.07 | 12.77 | 1.16 | 18.90 | P | 0.71 |
| Widespread carp (reside) | 12.10 | 12.16 | 0.43 | 6.98 | P | 1.06 | |
| Crucian carp (reside) | 15.88 | 16.29 | 0.91 | 11.56 | P | 2.01 | |
| Grass carp (reside) | 12.65 | 12.68 | 0.32 | 5.32 | P | 1.54 | |
| Largehead hairtail | 39.86 | 39.57 | 1.86 | 1.48 | P | 1.75 | |
| Silver carp (reside) | 6.98 | 7.02 | 0.47 | −1.66 | P | 0.58 | |
| Yellow croaker | 43.59 | 43.75 | 1.33 | 0.51 | P | 0.14 | |
- Be aware: On this desk, we report abstract statistics for wholesale meals costs between January 1 and February 29, 2020. Our value knowledge cowl 51 cities and 36 key merchandise. Costs are quoted in yuan per kilogram. Change (%) is the proportion change in costs from January 1 to February 29, 2020. P and NP point out perishable and non-perishable merchandise, respectively.
We observe substantial variation in value modifications throughout merchandise, with greens being probably the most closely affected meals merchandise. Solely 4 of the 19 greens skilled value declines. The worth of Chinese language cabbage, probably the most frequent, storable, and reasonably priced inexperienced greens in northern China, elevated by 81%. Different greens that skilled the best value inflation embrace bell peppers, potatoes, and carrots. Normally, the costs of non-perishable greens, corresponding to Chinese language cabbage, bell peppers, potatoes, and carrots, elevated to a considerably higher extent, with eggplant being the exception. In distinction, perishable greens, corresponding to bok choy, cauliflower, spinach, and tomatoes, skilled value declines or modest will increase in costs. It’s value noting that in China the classification of greens is detailed and particular. Due to this fact, there’s little variation inside every meals product, corresponding to cabbage.
Meat costs remained comparatively steady, with beef experiencing solely a 3.5% enhance throughout cities. This doubtlessly displays a mixture of few disruptions of the meat provide and a modest enhance in client demand as Chinese language shoppers want contemporary meat merchandise and use deep freezers much less extensively. This contrasts with the scenario within the U.S., the place beef costs for shoppers rose greater than 20% from March to June 2020 on account of “the mix of decrease provides from processors and better client demand,” in response to Mead et al. (2020). Few slaughterhouses or meat processing crops in China have been shut down, whereas a few of their counterparts within the U.S. have been pressured to shut in response to COVID-19 outbreaks (Almeida & Hirtzer, 2020). We observe that an oversupply of eggs that began earlier than COVID-19 hit doubtless contributed to the big decline in egg costs throughout our pattern interval (Craymer, 2020).
Seasonality may have contributed to the worth will increase in China, because the virus outbreak occurred shortly earlier than the Chinese language Lunar New 12 months (January 25, 2020), which usually sees an increase in meals costs. To account for this challenge, we look at whether or not there have been differential modifications in costs across the Lunar New 12 months between 2020 and prior years. The common costs throughout our pattern interval and the matched interval in 2017 and 2019 are proven in Determine 1.16 We see that value modifications in 2020 largely tracked these in prior years earlier than Wuhan was locked down (with some slight variations in early January), after which costs continued to rise and reached an area peak 6 days later. After a quick fall, costs started rising once more, which coincided with the primary lockdowns imposed exterior of Hubei Province. The common post-lockdown 2020 costs have been considerably greater than prior-year costs, though the typical costs don’t mirror heterogeneity in value modifications, an element we deal with intimately in Part 6. Such value will increase occurred at a time of the yr once we would usually observe steady and even reducing costs, as was the case in 2017 and 2019, suggesting a big enhance in client demand throughout our pattern interval (whereas provide chains throughout China have been functioning comparatively easily).
5 METHODOLOGY
during which is the day by day value for product p in metropolis c on day d relative to the Lunar New 12 months in yr y. The dummy variable equals 1 if the worth is from 2020 and 0 whether it is from 2019, which controls for systematic variations in meals costs between 2019 and 2020. is an indicator variable that equals 1 whether it is s days earlier than (after) the emergency declaration in metropolis c and 0 in any other case. The parameters are event-study coefficients that hint out deviations from the frequent developments that metropolis–product pairs expertise within the days main as much as and following the occasion. Product fastened results, , management for time-invariant product traits corresponding to perishability. Metropolis-day fastened results, , are included to regulate for city-specific seasonal developments and regional shocks. Time-invariant metropolis traits corresponding to metropolis measurement are captured by as nicely. As a result of the indicator variables for are extremely correlated with day fastened results, are excluded and solely the interplay phrases are stored. The reference interval within the mannequin is . Normal errors are clustered on the metropolis degree. In different specs, we exchange city-day fastened results with city-specific time developments and use wild cluster bootstrapped customary errors. The outcomes are offered in On-line Appendix B3.
Our occasion window for emergency declarations ends 10 days following a declaration as a result of lockdowns on common passed off 12.3 days after emergency bulletins. To review the results of lockdowns, we use the identical specification besides that the occasion window runs from 10 days earlier than to fifteen days after a lockdown. We select a post-lockdown window of 15 days as a result of a number of cities started to reopen 15 days after locking down. In Subsection 6.2.1, we prolong the post-lockdown window to 29 days after lockdown orders and look at whether or not meals costs revert to before-lockdown ranges.
6 EMPIRICAL RESULTS
On this part, we report our essential set of outcomes relating to the results of emergency declarations and lockdowns on meals costs, together with cross-sectional heterogeneity primarily based on the perishability of merchandise and lockdown strictness and timing.
6.1 Emergency declarations
In Determine 3a, we plot parameter estimates for the phrases representing interplay between 12 months 2020 and ED(s), as described in Equation (1), with a corresponding 90 p.c confidence interval. Detailed outcomes are reported in On-line Appendix Desk B1. On days previous a declaration of emergency, costs are basically no totally different from costs throughout the identical interval in 2019. None of these coefficients is statistically important, supporting the parallel developments assumption. After the emergency declaration, costs begin to climb, rising by over 14% on Day t = 4 relative to the 2019 degree, earlier than reverting by Day t = 10. Throughout many of the post-event days, the coefficient estimates are statistically totally different from 0 on the 5% degree. For instance, the estimated impact for Day t = 4 is equal to a 7.8 customary deviation of day by day value modifications prior to now winter.18 These outcomes counsel a considerable demand shock instantly following official bulletins of emergencies. As wholesalers in China usually inventory stock for a minimum of 1 week’s value of gross sales throughout the Lunar New 12 months season, it’s unlikely that any supply-chain disruption would happen within the instant aftermath of an emergency declaration, which coincided with the beginning of the Lunar New 12 months (January 25, 2020).
Meals costs round emergency declarations. On this determine, we summarize the proof derived from our event-study regressions (Equation 1) of the results of emergency declarations on wholesale meals costs for 51 cities and 36 key merchandise in our pattern. All regressions embrace city-day and product fastened results. Normal errors are clustered on the metropolis degree. In Panels (a) and (b), we present the estimated coefficients on “days relative to the emergency declaration” together with the 90% confidence intervals. Panel (a) plots the coefficients for the total pattern, whereas Panel (b) exhibits the estimated coefficients for perishable (blue line) and non-perishable (orange line) merchandise, respectively
We then discover whether or not there are any differential responses to the declarations with respect to merchandise with excessive and low levels of perishability. The pre-event 2020 costs of non-perishable merchandise are barely greater than 2019 costs, whereas not one of the coefficients is important. Following the declarations, costs enhance quickly, with the worth on Day t = 4 rising 20% greater than that in 2019. The worth surge is above 10% in 4 of the ten post-event days (p-value<0.01). We start to see reducing value progress after Day t = 4.
The pre-event meals costs of merchandise with a excessive diploma of perishability are near the 2019 costs, with no important coefficients or clear developments. In distinction to the findings for non-perishables, perishable product costs enhance barely 3 days following declarations, earlier than experiencing a big decline of as much as 6% after Day t = 7. It’s value noting that the distinction in value modifications between the 2 teams turns into statistically important beginning on Day t = 3. This distinction ranges from 8.4 to 16.1 proportion factors. These outcomes are per the proposition that residents took emergency declarations significantly and engaged in panic shopping for of non-perishable meals objects, doubtlessly as a consequence of considerations about future shortages.19 Therefore, the worth will increase seem like pushed primarily by elevated demand for non-perishable merchandise.
The demand shock related to emergency declarations seems to be quick lived, as shoppers realized that the majority meals objects have been nonetheless out there, aided by elevated stock in preparation for the Lunar New 12 months vacation. Our outcomes present little proof that offer chains have been considerably affected, in contrast to extreme supply-chain disruptions in a part of the U.S. throughout the COVID-19 disaster (Mead et al., 2020). If that had been the case, we might have noticed value will increase in each product teams, and extra so for perishable merchandise which can be much less prone to have been stocked abundantly by wholesalers.20
We have now so far mentioned the outcomes of regressions that weight every product equally. These outcomes, nonetheless, masks extensive heterogeneity in demand shocks throughout merchandise, which makes it troublesome to evaluate the implications of client habits for client welfare. To make clear this challenge, we estimate Equation (1) for every product, excluding product fastened results. In an untabulated evaluation, we discover that, inside per week’s time following emergency declarations, 12 of the 19 pattern greens skilled value will increase of greater than 7 customary deviations of the worth change distribution. Costs of Chinese language cabbage and potatoes, the 2 most closely consumed greens (accounting for 40% of vegetable consumption), elevated by as much as 50.9% and 18.7% (17.0 and 14.0 customary deviations), respectively. The magnitudes of the results for many meat merchandise are considerably smaller (lower than 1.9 customary deviations), with mutton being the exception. As households spend a good portion of their meals budgets on greens, our outcomes counsel that giant COVID-19 policy-driven demand shocks to vegetable costs led to important client welfare losses, particularly amongst low-income households that spend extra closely on greens (Tian & Yu, 2015).
We emphasize that anticipation results are unlikely to play a serious function when figuring out the results of emergency declarations. All emergency declarations in China have been issued inside a brief window, from January 23 to January 25, 2020. Due to this fact, even when some residents anticipated emergency declarations to be introduced of their jurisdictions following declarations in different areas, that they had little time to reply beforehand. Lockdowns have been, nonetheless, imposed sequentially over a span of 20 days, for which anticipation would possibly play a bigger function. We look at this attainable phenomenon in Subsection 6.2.3.
6.2 Lockdowns
One could argue that the results present in Subsection 6.1 are a results of residents responding to details about the coronavirus. That’s, even when there have been no official emergency declarations, we might have seen the identical habits in response to fears of the virus. As a placebo check, we observe that on January 11, 2020, about 2 weeks earlier than the emergency declarations, Wuhan introduced the primary loss of life attributable to COVID-19, which was broadcast throughout the nation by Xinhua Information. Presumably, residents would have reacted to the information and engaged in stockpiling if this piece of reports had influenced them. As proven in On-line Appendix Determine B4, nonetheless, meals costs didn’t expertise any noticeable change across the information occasion.
Though the negligible impact of Wuhan’s first loss of life helps the notion that the results on meals costs round emergency declarations are policy-driven, it’s attainable that earlier than the official declarations residents had not but discovered sufficient in regards to the severity of the outbreak. To deal with this chance, we repeat our evaluation round official lockdowns, all of which passed off in February 2020, besides in Wuhan, Huanggang, and Jingzhou in Hubei Province.21 If there stay value results round lockdowns, that is robust proof that client responses have been policy-driven. Presumably, shoppers would have discovered way more in regards to the virus by the point lockdowns have been applied, when new instances already peaked (On-line Appendix Determine B5a).
6.2.1 Predominant outcomes
As proven in Determine 4a, throughout the 10-day interval earlier than a lockdown, costs are flat relative to these in 2019. On the day of a lockdown, meals costs leap by 2.2% and proceed to extend till Day t = 5, when it sees a rise of 6.4% or 3.6 customary deviations of the worth change distribution. Curiously, value ranges stay elevated till the top of the occasion window, t = 15, with many of the coefficient estimates being statistically totally different from 0 on the 5% degree. On-line Appendix Desk B2 experiences the coefficient estimates.
Meals costs round lockdowns. On this determine, we summarize the proof derived from our event-study regressions (Equation 1) of the results of lockdowns on wholesale meals costs for 48 cities and 36 key merchandise in our pattern. All regressions embrace city-day and product fastened results. Normal errors are clustered on the metropolis degree. In Panels (a), (b), and (c), we present the estimated coefficients on “days relative to the lockdown order” together with the 90% confidence intervals. Panel (a) plots the coefficients for the total pattern, Panel (b) exhibits the estimated coefficients by excluding the three cities in Hubei, Wuhan, Jingzhou, and Huanggang, and Panel (c) plots the estimated coefficients for perishable (blue line) and non-perishable (orange line) merchandise, respectively
As a result of the three cities in Hubei Province (Wuhan, Huanggang, and Jingzhou) imposed lockdowns in late January 2020, at which period residents may not have identified a lot in regards to the virus, we repeat our evaluation by excluding these cities. The outcomes, reported in Panel (b), are similar to our essential outcomes and help the notion that the demand shock was largely policy-driven.
Determine 4c exhibits that non-perishable merchandise see a big enhance in costs, beginning instantly after lockdown orders are issued. On the finish of the occasion window, non-perishable costs are 9.6% greater than their 2019 ranges. Nonetheless, costs of perishable objects see little change, with a slight downward development after Day t = 5. The distinction between the 2 teams can also be important on the 5% degree on most days after Day t = 5. The estimated coefficients are proven in On-line Appendix Desk B3. Performing regressions for every product individually yields outcomes which can be qualitatively much like these obtained for emergency declarations, with greens being probably the most closely affected meals merchandise, though the magnitudes of the results are usually smaller.
Customers seem to have responded to emergency declarations, that are info and nonbinding occasions, otherwise than to mandated lockdowns. Emergency declarations have a bigger however extra momentary impact on meals costs whereas the impact of lockdowns is extra sustained.22 This implies that customers engaged in panic-buying habits when emergencies have been introduced, whereas lockdowns prompted residents to plan forward and retailer non-perishables for at-home consumption as they anticipated to buy groceries much less regularly. That is per the findings in Gupta et al. (2020) and Cronin and Evans (2020), who present that emergency declarations within the U.S. had bigger results on mobility and social distancing than stay-at-home orders did. It’s because emergency declarations, usually the earliest “info” actions taken by governments, doubtlessly concerned higher uncertainty than subsequently imposed mandates.
One could suspect that the price-gouging warning, which was introduced on February 1, 2020, may doubtlessly bias our estimates of the lockdown results downward. Nonetheless, the dominating function of small gamers in Chinese language meals markets (at each the wholesale and retail phases) makes it troublesome for sellers to interact in value gouging in such a aggressive market. As well as, the warning referred to cost margins quite than value ranges, indicating that, when retailers bid up wholesalers’ costs in response to stronger client demand, each wholesalers and retailers have been capable of enhance costs.
6.2.2 Lockdown strictness
Provided that lockdowns have been applied on the native degree, there could also be further variations in lockdown strictness throughout cities. The adversarial results on meals costs could possibly be extra extreme in cities beneath stricter restrictions as a result of their residents have been much less prone to depart their properties, implying a higher want for stockpiling meals for residence consumption. On this subsection, we use lockdown strictness, an indicator for cities that allowed just one particular person per family to exit to buy day-after-day or each different day, to check whether or not cities with stricter lockdowns skilled steeper meals value will increase. Particulars relating to how we code the indicator variable are offered in Subsection 3.2.23 We use the identical methodology as in Subsection 6.2.1. The regression outcomes are plotted in Determine 5. Non-perishable merchandise in cities with strict lockdowns see value hikes after the orders are issued, with statistically important coefficient estimates on most days. On the finish of the occasion window, non-perishable costs are 14.7% greater than their ranges in 2019. Costs of perishable merchandise, nonetheless, expertise little change. The distinction between the 2 teams can also be important on the 5% degree for 9 days throughout the post-event window. Alternatively, in cities with free lockdowns non-perishable merchandise expertise modest value modifications. The distinction between the 2 product teams can also be not important on the 10% degree.
Meals costs and lockdown strictness. On this determine, we summarize the proof derived from our event-study regressions (Equation 1) of the results of lockdown strictness on wholesale meals costs for 48 cities and 36 key merchandise in our pattern. All regressions embrace city-day and product fastened results. Normal errors are clustered on the metropolis degree. In Panels (a) and (b), we present the estimated coefficients on “days relative to the lockdown order” together with the 90% confidence intervals. Panel (a) (Panel b) plots the estimated coefficients for perishable (blue line) and non-perishable (orange line) merchandise for cities with strict (free) lockdowns
6.2.3 Timing of lockdowns
As proven in Desk A1, there’s a significant variation in lockdown timing, starting from January 23 to February 14, 2020, whereas emergency declarations all passed off between January 23 and 25, 2020. This allows us to discover whether or not food-price patterns differ between cities going through early and late lockdowns. Of the 48 cities that imposed lockdowns, we classify cities with early lockdowns as these experiencing lockdowns on or earlier than February 6, 2020 (the median lockdown time), whereas classifying the remaining cities as going through late lockdowns.24
In Desk 2 we report that meals costs leap after the lockdown orders in cities experiencing early lockdowns and proceed to rise throughout the occasion window. On Day t = 15, the typical value is 10.4% greater than the same-period 2019 value. Aside from an preliminary modest enhance of about 3% on Days t = 2 and t = 3, value modifications in cities going through late lockdowns usually are not totally different from 0 on the 10% degree. Extra importantly, throughout the 5-day window earlier than lockdowns, meals costs already started rising, suggesting that, as cities imposed lockdowns sequentially, residents in cities not but locked down anticipated an eventual lockdown and commenced to stockpile meals merchandise earlier than the official lockdown order was issued. This displays an anticipation impact. One may argue that customers’ worry of an infection in cities going through late lockdowns results in these outcomes. As Determine B5b within the On-line Appendix exhibits, nonetheless, in these cities the variety of new instances and case progress charges started to drop round Day t = −5 and continued a downward development since then. This implies that the anticipation of coming lockdowns, quite than worry of the unfold of virus, was extra prone to drive such client habits change in cities going through late lockdowns.
Timing of lockdowns and meals costs
| Dependent variable: ln (value) | ||||||
|---|---|---|---|---|---|---|
| Early lockdowns | Late lockdowns | Wuhan | ||||
| Coefficient | Std. Err. | Coefficient | Std. Err. | Coefficient | Std. Err. | |
| (1a) | (1b) | (2a) | (2b) | (3a) | (3b) | |
| 12 months 2020 | 0.103*** | 0.021 | 0.197*** | 0.024 | 0.080 | 0.079 |
| 12 months 2020 × Day = −10 | −0.021 | 0.020 | 0.015 | 0.026 | −0.054 | 0.112 |
| 12 months 2020 × Day = −9 | 0.032 | 0.029 | 0.000 | 0.027 | −0.048 | 0.113 |
| 12 months 2020 × Day = −8 | 0.044 | 0.033 | −0.029 | 0.025 | −0.068 | 0.112 |
| 12 months 2020 × Day = −7 | 0.038 | 0.027 | −0.037 | 0.025 | −0.045 | 0.112 |
| 12 months 2020 × Day = −6 | 0.070** | 0.028 | −0.021 | 0.034 | 0.013 | 0.113 |
| 12 months 2020 × Day = −5 | 0.020 | 0.024 | −0.039** | 0.019 | −0.005 | 0.113 |
| 12 months 2020 × Day = −4 | 0.028 | 0.024 | −0.050** | 0.023 | −0.019 | 0.112 |
| 12 months 2020 × Day = −3 | −0.007 | 0.025 | −0.033* | 0.019 | −0.033 | 0.112 |
| 12 months 2020 × Day = −2 | 0.004 | 0.019 | −0.002 | 0.016 | −0.012 | 0.112 |
| 12 months 2020 × Day = 0 | 0.016 | 0.012 | 0.027 | 0.018 | 0.061 | 0.112 |
| 12 months 2020 × Day = +1 | 0.024 | 0.023 | 0.034*** | 0.012 | 0.120 | 0.112 |
| 12 months 2020 × Day = +2 | 0.044** | 0.022 | 0.032** | 0.014 | 0.265** | 0.122 |
| 12 months 2020 × Day = +3 | 0.046* | 0.025 | 0.002 | 0.020 | 0.295** | 0.132 |
| 12 months 2020 × Day = +4 | 0.074*** | 0.020 | 0.011 | 0.014 | 0.167 | 0.135 |
| 12 months 2020 × Day = +5 | 0.091*** | 0.023 | 0.029 | 0.021 | 0.288** | 0.120 |
| 12 months 2020 × Day = +6 | 0.097*** | 0.026 | 0.006 | 0.020 | 0.222* | 0.133 |
| 12 months 2020 × Day = +7 | 0.092*** | 0.023 | 0.011 | 0.023 | 0.235* | 0.133 |
| 12 months 2020 × Day = +8 | 0.088*** | 0.026 | 0.012 | 0.024 | 0.275** | 0.112 |
| 12 months 2020 × Day = +9 | 0.073*** | 0.027 | 0.014 | 0.024 | 0.367*** | 0.115 |
| 12 months 2020 × Day = +10 | 0.076*** | 0.029 | 0.019 | 0.025 | 0.336*** | 0.115 |
| 12 months 2020 × Day = +11 | 0.055** | 0.025 | 0.013 | 0.026 | 0.241** | 0.116 |
| 12 months 2020 × Day = +12 | 0.083*** | 0.029 | −0.009 | 0.025 | 0.276** | 0.115 |
| 12 months 2020 × Day = +13 | 0.090*** | 0.029 | −0.004 | 0.029 | 0.357*** | 0.117 |
| 12 months 2020 × Day = +14 | 0.086*** | 0.029 | −0.017 | 0.027 | 0.454*** | 0.117 |
| 12 months 2020 × Day = +15 | 0.104*** | 0.032 | −0.018 | 0.025 | 0.459*** | 0.115 |
| Fixed | 0.922*** | 0.042 | 0.987*** | 0.044 | 0.783*** | 0.042 |
| Observations | 1104 | 27,116 | 18,892 | |||
| Adj. R-squared | 0.90 | 0.86 | 0.90 | |||
| Metropolis-day FEs | Sure | Sure | Sure | |||
| Product FEs | Sure | Sure | Sure | |||
- Be aware: On this desk, we report the proof derived from our event-study regressions (Equation 1) of the results of lockdowns on wholesale meals costs for cities experiencing early and late lockdowns in addition to Wuhan, the primary metropolis that was locked down (January 23, 2020). The dependent variable is the logarithm of the wholesale value for every key product. Early lockdowns have been imposed on or earlier than February 6, 2020 and the remaining cities are categorized as going through late lockdowns. In every column, we report coefficient estimates and their related customary errors. All regressions embrace city-day and product fastened results. Normal errors are clustered on the metropolis degree. ∗, ∗∗, and ∗∗∗ point out statistical significance on the 10%, 5%, and 1% ranges, respectively.
To develop a case research, we look at how meals costs in Wuhan responded to its lockdown, the primary in China. As proven in columns (5) and (6) of Desk 2, relative to cities in different provinces, meals costs elevated extra considerably. Costs jumped 6.1% on the primary day of the lockdown and continued to extend towards the top of the occasion window, when costs elevated practically 46% in contrast with 2019 ranges. This implies that the shocking nature of Wuhan’s lockdown contributed to residents’ stockpiling of meals objects extra aggressively, which lends help to the differential findings between cities going through early and late lockdowns. We additionally conduct subsample analyses primarily based on the perishability of merchandise (not tabulated). For cities going through early lockdowns, the outcomes are much like our essential outcomes, with important value will increase for non-perishable merchandise. In cities going through late lockdowns, nonetheless, there have been no important modifications in costs for both non-perishable merchandise or perishable merchandise.
To additional discover a attainable channel of the anticipation impact, we give attention to cities that confronted late lockdowns (after February 6, 2020) and look at whether or not meals costs in such cities responded to the primary lockdown in the identical province. For Beijing and Shanghai, two direct-control municipalities, we use the date of the primary lockdown in an adjoining province. As proven in Determine 6, costs elevated instantly after the lockdown of the personal province’s first metropolis, suggesting that an anticipation impact did exist and residents seem to have discovered regionally. That’s, residents rationally anticipated that their cities can be locked down when different cities in the identical province had been closed, as policymaking is usually constant inside a province.
Meals costs round personal province’s first lockdown. On this determine, we summarize the proof derived from our event-study regressions (Equation 1) of the results of the primary lockdown in a province on wholesale meals costs in cities that have been locked down subsequently. All regressions embrace city-day and product fastened results. Normal errors are clustered on the metropolis degree. Our pattern contains solely cities that have been locked down after February 6, 2020. For Beijing and Shanghai, two direct management municipalities, we use the date of the primary lockdown in an adjoining province. We present the estimated coefficients on “days relative to the lockdown order” together with the 90% confidence intervals
7 CONCLUSIONS
On this research, we present that Chinese language shoppers responded strongly to authorities restrictions on private mobility, which have been among the many world’s hardest, throughout the COVID-19 disaster. Particularly, we discover that emergency declarations, which engendered expectations of additional governmental actions in response to the pandemic, raised common meals costs by as a lot as 7.8 customary deviations of the worth change distribution. The consequences of emergency declarations have been bigger however extra transient than the results of lockdown bulletins. This implies that customers engaged in panic shopping for of non-perishables throughout emergency declarations whereas beneath lockdowns there was a sustained enhance in demand for non-perishables. We additionally present that worry of the virus is unlikely to be the primary driver of our outcomes.
We emphasize that we observe substantial heterogeneity in value modifications throughout merchandise, with non-perishable greens, corresponding to Chinese language cabbage and potatoes, being probably the most closely affected product group. Provided that Chinese language households spend a good portion of their meals budgets on greens and that vegetable consumption is double that of meat and fruits, our outcomes suggest that demand shocks in response to COVID-related restrictions affected client welfare negatively, particularly amongst poorer individuals. As low-income house owners spend greater parts of their meals budgets on greens (Tian & Yu, 2015), they doubtless endured important welfare losses. In distinction, welfare amongst high-income shoppers, who eat extra meat and fruits, was prone to be much less affected.
The distinct construction of Chinese language meals markets leads us to imagine that market mechanisms, on this case greater meals costs, carry out extra successfully than rationing in managing briefly elevated demand. In contrast to in Western nations, the place a number of giant retail chains dominate meals markets and rationing methods might be set uniformly throughout institutions, the dominance of small meals retailers in China implies that any rationing rule have to be imposed by the federal government. It’s difficult, nonetheless, to determine efficient rationing standards given the substantial heterogeneity in value modifications throughout meals merchandise that we observe in our evaluation. As well as, imposing rationing in a market that options numerous aggressive gamers can be pricey, if not inconceivable.
The extra pronounced response to lockdown insurance policies in China than within the U.S. (Goolsbee & Syverson, 2021) highlights the significance of institutional variations. Most Chinese language cities have been locked down after new instances had already peaked, affording residents sufficient time to react to the virus itself and allow us to disentangle the impact of lockdowns. Within the U.S., nonetheless, the orders have been imposed in response to the arrival of COVID-19 and have been usually much less strict than these in China. Various lockdown strictness and timing throughout cities and areas might need diminished supply-chain challenges, contributing to comparatively clean provide chains throughout China throughout the lockdown interval.
ACKNOWLEDGMENTS
The authors have significantly benefited from feedback and ideas from the editor (Nicolas Ziebarth) and three nameless referees. Frank Asche gratefully acknowledges monetary help from the Norwegian Analysis Council grant #324685.
ENDNOTES
APPENDIX A
A1 | Extra knowledge sources
A1.1 | Shelf-life knowledge
Info on shelf life for many greens and fruits is collected from Kansas State College’s Extension Meals Security.25 This report lists the shelf-life vary for every product. We classify a product as non-perishable if its higher shelf life restrict exceeds 2 weeks. In line with this desk, bell peppers, cabbage, carrots, cauliflower, celery, celtuce, Chinese language cabbage, inexperienced onions, onions, potatoes, and turnips are categorized as non-perishable greens. Cucumbers, eggplant, snap beans, spinach, and tomatoes are categorized as perishable greens. Relating to fruits, apples, grapes, and watermelon are categorized as non-perishable.
We seek for the shelf lives of merchandise that aren’t included on the Extension Meals Security record utilizing the US Division of Agriculture’s FoodKeeper App.26 Bok choy, garlic chives, bananas, and pineapples are perishable, as they need to be consumed inside per week following the date of buy. Eggs, which might be saved as much as 5 weeks, are thought to be a non-perishable product. Garlic shoots usually are not lined in both database. We classify them as non-perishable primarily based on survey responses from 10 family and friends members. Contemporary meat is the main product type of meat merchandise in China. For fish, reside freshwater fish and chilled seafood are offered largely in wholesale and retail markets. We subsequently classify all meat and seafood as perishable merchandise.
A1.2 | Mobility knowledge
We get hold of the within-city mobility index from Baidu Migration, a journey map supplied by the biggest Chinese language search engine, Baidu. The Baidu mobility index information day by day within-city mobility for 364 main Chinese language cities, together with all 51 cities in our pattern. For every of those cities, we extract its day by day mobility index for the interval operating from January 1 by means of February 29, 2020 and the corresponding lunar interval in 2019.
Baidu Migration collects knowledge primarily based on greater than 120 billion day by day location requests from the Baidu Map app, which has 420 million month-to-month lively customers or 60% of smartphone customers in China, and different apps that use Baidu’s location companies (Questmobile, 2019). It captures the proportions of native populations who journey inside a metropolis quite than the entire variety of journeys made by residents. This makes the index constant and comparable throughout time and cities.
A1.3 | Different knowledge
Knowledge on inhabitants and gross home product (GDP) per capita in 2019 are collected from every metropolis’s Statistical Yearbook. We additionally accumulate latitude and longitude codes for every metropolis from the Nationwide Geomatics Heart of China and compute the great-circle distance between every metropolis and Wuhan by making use of the Haversine components.27 COVID-19 day by day case counts are collected from the Chinese language Heart for Illness Management and Prevention, which supplies day by day updates on new case counts and cumulative counts of instances in every metropolis.28 We use every metropolis’s cumulative depend of instances on the day earlier than its lockdown to regulate for native outbreak severity.
A2 | Validation of the lockdown strictness measure
Determine A2 exhibits the day by day common mobility index throughout cities throughout our pattern interval in 2020 and the corresponding interval in 2019. Following an analogous development in 2019, within-city mobility began to lower 4 days previous to the 2020 Lunar New 12 months, however extra shortly, reflecting the results of emergency declarations associated to COVID-19. In the course of the Lunar New 12 months week (Day 0 to Day 6), mobility usually remained at a low degree however was anticipated to return to a traditional degree afterward, as was the case for 2019. Nonetheless, in 2020 mobility continued to drop and reached the bottom degree 15 days after the Lunar New 12 months earlier than rising step by step. In the course of the interval between Day 0 and Day 35, the typical day by day mobility index fell from 4.49 in 2019 to 2.39 in 2020.
Prolonged estimation window for meals costs round lockdowns. On this determine, we summarize the proof derived from our event-study regressions (Equation 1) of the results of lockdowns on wholesale meals costs for 48 cities and 36 key merchandise in our pattern. The estimation window runs from 10 days earlier than a lockdown to 29 days after a lockdown. All regressions embrace city-day and product fastened results. Normal errors are clustered on the metropolis degree. We plot the estimates coefficients on “days relative to the lockdown order” together with the 90% confidence intervals
Evolution of within-city mobility indexes. On this determine, we plot the day by day common mobility index throughout the 51 pattern cities throughout our pattern interval (January 1 to February 29, 2020) and its corresponding lunar calendar interval in 2019 (January 12 to March 12, 2019)
It’s attainable that though native lockdown orders themselves exhibit various strictness, residents may not totally adjust to such orders. We assess the validity of our lockdown strictness measure by analyzing whether or not stricter orders are associated to a higher lower in mobility inside cities, proxied by reductions in Baidu’s within-city mobility indexes. We common every metropolis’s mobility index from Day 0 to Day 15 after lockdowns and calculate the discount relative to the index within the corresponding interval in 2019. We additionally use the typical mobility index within the first week of 2020 (January 1 to January 7, 2020) in its place base to compute mobility discount. All cities in our pattern skilled diminished mobility by both measure, with modal and common reductions of 65.5% (66.1%) and 59.8% (58.7%) utilizing 2019 (2020) as the bottom, respectively.
We estimate an extraordinary least squares mannequin. Inhabitants and GDP per capita are added to regulate for metropolis measurement and revenue ranges. To regulate for the native severity of outbreaks, we additionally embrace every metropolis’s cumulative variety of confirmed instances on the day instantly earlier than the lockdown order. As well as, we management for fastened results for 3 main areas in China: North, South, and Northwest. Normal errors are clustered on the area degree. As proven in column (1) of Desk A2, relative to cities with free lockdowns, cities with strict lockdowns see a 12 percentage-point greater discount of mobility over the 2019 degree. Provided that the typical discount in mobility is 59.8%, the impact of lockdown strictness on mobility modifications is economically significant. Equally, cities with strict lockdowns are 36.1 proportion factors extra prone to expertise a mobility discount over the modal discount of 65.5%, as proven in column (2). Each estimates are important on the 1% degree. To acquire the outcomes reported in columns (3) and (4), we use another measure of mobility discount that’s primarily based on mobility ranges within the first week of 2020 and acquire constant outcomes.29 Excessive-income cities, proxied by GDP per capita, are likely to expertise a higher discount in mobility. This implies that residents in high-income cities, who are sometimes extra extremely educated, usually tend to adjust to authorities orders. We emphasize that the cumulative depend of instances doesn’t have an effect on mobility discount, indicating that variations in mobility reductions throughout cities mirror variations in lockdown strictness quite than voluntary choices taken in response to the virus. Taken collectively, the outcomes reported on this subsection counsel that the stricter a lockdown is the extra doubtless residents are to stockpile non-perishable merchandise for residence consumption.
Metropolis-level occasion dates
| Metropolis | Province | Emergency declaration | Lockdown order |
|---|---|---|---|
| Anshan | Liaoning | 1/25/20 | 2/5/20 |
| Anyang | Henan | 1/25/20 | |
| Baotou | Inside Mongolia | 1/25/20 | 2/12/20 |
| Beijing | Beijing | 1/24/20 | 2/10/20 |
| Changde | Hunan | 1/23/20 | 2/2/20 |
| Changsha | Hunan | 1/23/20 | 2/14/20 |
| Changzhou | Jiangsu | 1/24/20 | 2/2/20 |
| Chaoyang | Liaoning | 1/25/20 | 2/6/20 |
| Chengdu | Sichuan | 1/24/20 | 2/7/20 |
| Dalian | Liaoning | 1/25/20 | 2/9/20 |
| Dezhou | Shandong | 1/24/20 | 2/4/20 |
| Fuzhou | Fujian | 1/24/20 | 2/4/20 |
| Guang’an | Sichuan | 1/24/20 | 2/7/20 |
| Guangzhou | Guangdong | 1/23/20 | 2/7/20 |
| Handan | Hebei | 1/24/20 | 2/7/20 |
| Hangzhou | Zhejiang | 1/23/20 | 2/4/20 |
| Hefei | Anhui | 1/24/20 | 2/5/20 |
| Hegang | Heilongjiang | 1/25/20 | 2/9/20 |
| Hohhot | Inside Mongolia | 1/25/20 | 2/12/20 |
| Huanggang | Hubei | 1/24/20 | 1/23/20 |
| Jiaxing | Zhejiang | 1/23/20 | 2/4/20 |
| Jinan | Shandong | 1/24/20 | 2/5/20 |
| Jincheng | Shanxi | 1/25/20 | 2/3/20 |
| Jingzhou | Hubei | 1/24/20 | 1/24/20 |
| Jiuquan | Gansu | 1/25/20 | 2/6/20 |
| Mianyang | Sichuan | 1/24/20 | 2/9/20 |
| Nanchang | Jiangxi | 1/24/20 | 2/6/20 |
| Nanjing | Jiangsu | 1/24/20 | 2/4/20 |
| Ningbo | Zhejiang | 1/23/20 | 2/4/20 |
| Ningde | Fujian | 1/24/20 | 2/5/20 |
| Puyang | Henan | 1/25/20 | 2/6/20 |
| Qingdao | Shandong | 1/24/20 | 2/5/20 |
| Qinhuangdao | Hebei | 1/24/20 | 2/8/20 |
| Sanmenxia | Henan | 1/25/20 | 2/1/20 |
| Shanghai | Shanghai | 1/24/20 | 2/10/20 |
| Shangqiu | Henan | 1/25/20 | 2/10/20 |
| Shaoxing | Zhejiang | 1/23/20 | 2/4/20 |
| Suzhou | Jiangsu | 1/24/20 | 2/6/20 |
| Taiyuan | Shanxi | 1/25/20 | 2/5/20 |
| Tangshan | Hebei | 1/24/20 | 2/6/20 |
| Tianjin | Tianjin | 1/24/20 | 2/6/20 |
| Urumqi | Xinjiang | 1/25/20 | |
| Wuhan | Hubei | 1/24/20 | 1/23/20 |
| Wuxi | Jiangsu | 1/24/20 | 2/9/20 |
| Xi’an | Shaanxi | 1/25/20 | 2/2/20 |
| Xiangxi | Hunan | 1/23/20 | |
| Xianyang | Shaanxi | 1/25/20 | 2/5/20 |
| Yangquan | Shanxi | 1/25/20 | 2/6/20 |
| Yangzhou | Jiangsu | 1/24/20 | 2/5/20 |
| Yuxi | Yunnan | 1/24/20 | 2/8/20 |
| Zhangjiakou | Hebei | 1/24/20 | 2/6/20 |
- Be aware: This desk exhibits an inventory of emergency declaration and lockdown dates for every of the 51 Chinese language cities in our pattern.
Lockdown strictness and mobility discount
| Dependent variable | Relative to 2019 | Relative to first week of 2020 | ||
|---|---|---|---|---|
| Mobility discount (%) | Above modal mobility discount | Mobility discount (%) | Above modal mobility discount | |
| (1) | (2) | (3) | (4) | |
| Strict lockdown | 12.192*** (4.122) | 0.361*** (0.127) | 12.445*** (3.686) | 0.360*** (0.114) |
| ln (inhabitants) | 2.217 (5.761) | −0.002 (0.226) | −0.259 (5.331) | −0.085 (0.150) |
| ln (GDP per capita) | 12.744*** (2.889) | 0.498** (0.218) | 14.743*** (2.910) | 0.365*** (0.136) |
| ln (#Covid-19 instances+1) | 0.500 (2.745) | 0.055 (0.092) | 1.335 (2.364) | 0.058 (0.058) |
| Fixed | −126.629 (97.654) | −5.561 (3.421) | −114.676 (91.626) | −2.851 (1.715) |
| Observations | 48 | 48 | 48 | 48 |
| Adj. R-squared | 0.55 | 0.22 | 0.56 | 0.22 |
| Area FEs | Sure | Sure | Sure | Sure |
- Be aware: On this desk, we present outcomes relating to whether or not lockdown strictness predicts reductions in within-city mobility. Mobility discount (%) is the proportion discount of the mobility index for every metropolis, whereas Above modal mobility discount is an indictor equal to 1 if the mobility discount is above the modal discount of 65.5% (66.1%) utilizing 2019 (2020) as the bottom and 0 in any other case. For columns (1) and (2), we common every metropolis’s mobility index from Day 0 to Day 15 after lockdowns and calculate its discount relative to the index within the corresponding interval in 2019. For columns (3) and (4), we use the typical mobility index within the first week of 2020 (January 1 to January 7, 2020) in its place base to compute mobility discount. Lockdown strictness equals 1 for cities that allowed just one particular person per family to exit procuring day-after-day or each different day and 0 in any other case. Figures for metropolis inhabitants and annual GDP per capita (yuan) are from 2018. #Covid-19 instances is every metropolis’s cumulative variety of confirmed instances on the day instantly earlier than the lockdown order. Area fastened results are managed for the three main areas: South, North, and Northwest. Normal errors are clustered on the area degree. ∗∗ and ∗∗∗ point out statistical significance on the 5% and 1% ranges, respectively.
REFERENCES
