Just as physicians monitor their hospitalized patients in real time to determine the best way to treat them, economists are leaning towards real-time tracking of economic conditions to make policy decisions – for example, Proxy for GDP in Olosco (2020) or Cavallo and Rigobon (2016).
In a new paper (Rigobon et al. 2002), we introduce a new half-real-time estimate of business opening and closing rates using the dataset behind the Google Places – Google Maps service. We see that the lifting of the COVID-19 ban in Canada coincides with a wave of re-entry of temporarily closed businesses, suggesting that government support may have helped hibernated businesses survive.
How policymakers can track business health in semi-real time to assess the underlying trade-offs between health and socio-economic restrictions during epidemics (on-going and others. Unfortunately, existing methods often rely on the low-frequency data proxies available with time lag – Such as tax disclosure, business registry or survey.Lack of timely information on business health, it is difficult to strike the right balance.Too little, too late‘Government support that causes continued losses to manufacturing companies, thus long-term productivity (Aghion et al. 2019) and employment (Sedlasek 2020), and’Too much, too wide‘Government support that enables non-productive‘ zombie ’farms to survive (e.g. Gourinchas et al. 2021, Cros et al. 2021).
The rapid nature of the COVID-19 epidemic has accelerated the search for timely measures of business mobility (Agresti et al. 2022). Crane et al. (2020) Google investigates the value of search, paycheck issues, and the use of phone-tracking data. Yelp (2020) uses its US platform to calculate temporary closures in the early stages of an epidemic. Kurman et al. (2021) A portion of the rebound of small business employment in the U.S. services sector has been identified from SafeGraph, Facebook and Google due to business reopening.
A new way to track business using Google Places
To assist policymakers in timely tracking of business dynamics at Rigoban et al. (2022) We introduce a new estimate of opening and closing rates using obsolete half-real-time data from Google Places. We track the presence and disappearance of ‘pins’ on Google Maps that represent unique businesses using a two-way algorithm (Figure 1).
Figure 1 Images of scraping algorithms to collect all downtown Ottawa restaurants (‘pins’) on Google Places by May 2021
Note:: The higher the business density along the main road (point), the more subtle the algorithm search grid (square).
To create an image of how business conditions are changing, we only need to collect identifiers, numbers and business conditions in each geographic area or sector. Since the Google Places API only provides up-to-date information to private businesses, the algorithm is repeatedly run to collect data for the same area and thus create a time series. Entries and exits are identified by unique business identifiers that appear and disappear from the dataset. Temporary closure and reopening are notified by changes in business conditions that are either effective or temporarily closed.
Application for food and retail business
Duprey (2022), we apply our method to a set of Canadian cities for the food service (‘cafe’, ‘bar’, ‘restaurant’, ‘nightclub’) and retail (‘store’) sectors, accurately by the most epidemic Affected. We have seen that the lifting of the COVID-19 restriction in the summer of 2021 has resulted in a large wave of business entries, largely driven by the reopening of temporarily closed businesses (Figure 2b). It suggests that government assistance has facilitated the survival of hibernated businesses and could contribute to a faster recovery after the lifting of restrictions. We further note that the reopening time of businesses coincides with the lifting of sanctions. In Vancouver, for example, the ban was lifted a month ago and is associated with an increase in new entries and prior reopening. Similarly, restrictions for nightclubs were lifted a month later, and the peak of opening rates has been delayed accordingly.
Figure 2 Removal of COVID-19 restrictions for retail in Ontario, Canada in 2021 and entry / exit rates
A) COVID-19 cases and restrictions
B) Retail sector entry / exit rate
Comments: Panel (a) shows the number of COVID-19 cases from the Canadian Public Health Agency for the province of Ontario. The vertical bars represent the three stages of the reopening of the economy and the shaded area lockdown and home stay. Panel (b) shows approximate month-end opening and closing rates for retail and food sectors from Google Places data for Toronto and Ottawa city centers.
At the time of the early 2021 restrictions, about 92% of businesses were in the retail sector (Figure 3a). Of the businesses that temporarily closed in the retail sector at the beginning of the April 2021 lockdown, 40% reopened and 30% closed permanently by the end of the summer of 2021 (Figure 3b). For the food sector, about 87% of businesses were open during the lockdown, and after the restrictions were lifted, about half of those that were temporarily closed have reopened. There were reopening events for most bars (62% of temporarily closed ones reopened), while most were for permanently closed cafes.
Figure 3 The share of business that was temporarily closed in the run-up to the resumption of the Canadian economy in June 2021
A) The share of business that is temporarily closed as soon as the economy reopens
b) Business status that was initially flagged as a temporary closure in April.
Comments: Estimates of the end of the month from April to September 2021 using Google Places data for Toronto and Ottawa city centers. In panel (a), the vertical bars represent the three stages of the reopening of the economy and the shaded area lockdown and home stay. In panel (b), we only track subsets of businesses that were identified as temporarily closed at the beginning of the lockdown in April 2021 and evaluate the recovery rate of those businesses.
The latest Omicron wave in late 2021 was not accompanied by the same severe trade restrictions as the early 2021 wave. As a result, this wave is associated with a slightly higher closing rate than the entry rate at the end of December 2021. A reversal at the end of January (Fig. 2b).
Conclusion
Going forward, half-real-time business opening and closing rates can be used as an input for indicators that track the overall health of the business sector. For example, Statistics Canada (2021) creates a real-time local business condition indicator by combining open and closed with real-time traffic data around businesses to proxy for both broad and intensive margins of business activity. Information on opening and closing individual businesses to investigate the impact on labor mobility can also be combined with real-time job vacancies. Finally, high-frequency business health data can also help document the effects of natural disasters localized in space and time.
More broadly, data collected at a micro level can provide a subtle idea of small business dynamics. For example, the flexibility of the data collection process may allow the emergence of online retailers operating from the owner’s residence or investigating the relative dynamics of business in the city compared to commercial areas outside the city center.
There are several limitations to our approach to note. First, Google Places data is constantly updated but cannot be collected in a timely manner, limiting the ability to benchmark results at the pre-epidemic level. Second, it is difficult to evaluate the closure of a business because a business no longer exists to confirm the closing time (for opening a business, see Duprey et al. 2022 survey). Third, the quality of the estimates depends on the quality of the data, which is controlled by Google. Nevertheless, estimates of entry and exit statistics, despite differences in definition, appear to be related to data from Canada (2020) (Duprey et al. 2022).
Finally, as the digitization of the economy continues, the coverage and reliability of real-time systems for opening and closing businesses will continue to improve. As such, it will become increasingly important to policymakers and researchers. Thus, data providers like Google Places, SafeGraph, and others may want to consider compiling (and possibly monetizing) business health statistics themselves.
Authors Note: The opinions expressed in this column are those of the authors, and do not represent the Bank of Canada or the Bank of England and its committees.
References
Agion, P, A Bergeaud, T Boppart, PJ Klenow and H Li (2019), “Growing Missing from Creative Destruction”, American Economic Review 109 (8): 2795-2822.
Agresti, S, F Calvino, C Criscuolo, F Manaresi and R Verlhac (2022), “Tracking Business Dynamics During the COVID-19 Epidemic: New Cross-Country Evidence and Visualization Tool”, VoxEU.org, 17 January.
Cavallo A and R Rigobon (2016), “Billion Value Project: Using Online Value for Measurement and Research”, Journal of Economic Perspectives 30 (2): 151-178.
Cont, R, A Kotlicki and R Xu (2021), “Modeling COVID-19 Infection: Risk Assessment and Targeted Mitigation Policy”, Royal Society Open Science 8, 201535.
Crane, LD, RA Decker, A Flaaen, A Hamins-Puertolas, and C Kurz (2020), “COVID-19 Exit Business During Epidemic: Non-Traditional Measures in Historical Context”, Board of Governors of the Federal Reserve System.
Cros, M, A Epaulard and P Martin (2021), “Schumpeter will catch Covid-19? Evidence from France ”, VoxEU.org, 4 March.
Duprey, T, DE Rigobon, P Schnattinger, A Kotlicki, S Baharian, TR Hurd (2022), “Business closed and (re) opened in real time using Google Places”, Bank of Canada Staff Working Paper 2022-1.
Gourinchas, PO, S Kalemli-Ozcan, V Penciakova and N Sander (2021), “COVID-19 and small and medium-sized enterprises: a 2021 ‘time bomb’?”, AEA Papers and Procedures 111: 282-286.
Kurmann, A, E Lale and L Ta (2021), “The Impact of COVID-19 on Small Business Mobility and Employment: Real-Time Estimates with Homebase Data”, CIRANO.
Rigobon, DE, T Duprey, A Kotlicki, P Schnattinger, S Baharian, TR Hurd (2022), Journal of Risk and Financial Management 15 (4): 183.
Sedlacek, P (2020), “Lost Generation of Farms and Overall Labor Market Dynamics”, Journal of Monetary Economics 111: 16-31.
Statistics Canada (2020), “Experimental Estimates for Business Opening and Closing for Canada, Provinces and Territories, Census Metropolitan Areas, Adjusted to Seasons”.
Statistics Canada (2021), “Real-Time Local Business Conditions Index: Concepts, Data Sources, and Methods”.
Woloszko, N (2020), Real-time activity tracking with Google Trends, OECD Publishing.
Yelp (2020), Local economic impact report.