Measuring business dynamics in real time

Thibaut Duprey, Artur Kotlicki, Daniel Rigobon and Philip Schnattinger

Just as doctors monitor in real time the vital signs of their hospitalised patients to determine the best course of treatment, economists are turning towards a real-time tracking of economic conditions to inform policy decisions (for example, through proxy for GDP and inflation). In a recent paper, we introduce a new quasi-real time estimation of business opening and closure rates using data from Google Places – the dataset behind the Google Maps service. We find that the lifting of COVID-19 restrictions in Canada coincides with a wave of re-entry of temporarily closed businesses, suggesting that government support may have facilitated the survival of hibernating businesses.

How can policymakers track business health in quasi-real time, to assess the inherent trade-off between health and socioeconomic restrictions during the pandemic, as well as help calibrate government support to avoid closures of viable businesses? Unfortunately, existing methodologies often rely on low frequency data proxies available with a time lag – such as tax disclosures, business registry, or surveys. In the absence of timely data on business health, it is hard to strike the right balance between too-little-too-late government support, that causes a persistent loss of productive firms (hurting long-run productivity and employment), and too-much-too-broad government support, that enables the survival of non-productive ‘zombie’ firms (for example, see Gourinchas et al and Cros et al).

The fast-paced nature of the COVID-19 pandemic accelerated the search for timely measures of business dynamism. Crane et al investigate the value of using Google searches, paycheck issuance, and phone-tracking data. Yelp uses its US platform to compute temporary closures during the early phase of the pandemic. Kurmann et al find that part of the rebound in small business employment for the US service sector is due to business reopenings, identified from SafeGraph, Facebook and Google.

A new method to track businesses using Google Places

To aid policymakers in timely tracking of business dynamics, our paper introduces a new estimation of opening and closure rates using non-traditional quasi-real-time data from Google Places. We track the appearance and disappearance of ‘pins’ on Google Maps that represent unique businesses using a bisection algorithm (Figure 1).

Figure 1: Illustration of the scraping algorithm to collect all downtown Ottawa restaurants (the ‘pins’) as of May 2021 in Google Places. The higher the density of businesses along the main streets (dots), the finer the algorithm search grid needs to be (squares).

To form a picture of how business conditions are changing, we need only to collect the identifiers, number, and status of businesses in each geographic area or sector. Since the Google Places API returns only the most recent information on individual business establishments, the algorithm is repeatedly run to collect data for the same area and thus to build a time series. Entries and exits are identified by unique business identifiers that appear and disappear from the dataset. Temporary closures and reopenings are informed by changes in the business status that is either operational or temporarily closed.

Application to the food and retail businesses

Our methodology is applied in Duprey et al to a set of Canadian cities for the food service (‘cafe’, ‘bar’, ‘restaurant’, ‘night club’) and retail (‘store’) sectors, precisely those most impacted by the pandemic. We find that the lifting of COVID-19 restrictions in the summer of 2021 (Figure 2a) led to a large wave of business entries, which were largely driven by the reopening of temporarily closed businesses (Figure 2b). This suggests that government support may have facilitated the survival of hibernating businesses and contributed to a faster recovery once restrictions were lifted. We further observe that the timing of the reopening of businesses largely coincides with the timing of the lifting of the restrictions. For instance, restrictions got lifted one month earlier in Vancouver and this is associated with an earlier rise of new entries and reopenings. Similarly, restrictions got lifted one month later for night clubs and the peak of opening rate is delayed accordingly.

Figure 2: The lifting of COVID-19 restrictions and entry/exit rates for retail businesses in 2021 in Ontario, Canada. Panel (a) displays the COVID-19 case count from the Public Health Agency of Canada for the Province of Ontario. The vertical bars are the three phases of the reopening of the economy and the shaded area represents the lockdown and stay-at-home order. Panel (b) displays the end of month opening and closure rates for the retail and food sectors estimated from Google Places data for the city centres of Toronto and Ottawa.

During the early 2021 restrictions, about 92% of the businesses in the retail sector were operational (Figure 3a). Among those businesses in the retail sector that were temporarily closed at the start of the April 2021 lockdown, 40% reopened and 30% permanently closed by the end of the summer of 2021 (Figure 3b). For the food sector, about 87% of businesses were open during the lockdown, and upon lifting of the restrictions, about half of those temporarily closed reopened. Most reopenings took place for bars (62% of those temporarily closed reopened), while most permanent closures took place for cafes.

Figure 3: Share of temporarily closed businesses around the June 2021 reopening of the Canadian economy. End of month estimates for April to September 2021 using Google Places data for the city centre of Toronto and Ottawa. On panel (a), the vertical bars are the three phases of the reopening of the economy and the shaded area represents the lockdown and stay-at-home order. On panel (b), we track only the subset of businesses that were identified as temporarily closed at the beginning of the lockdown in April 2021 and assess the recovery rate of those businesses.

The latest Omicron wave at the end of 2021 was not accompanied with business restrictions as severe as the wave of early 2021. Consequently, this wave is associated with a closure rate only slightly higher than the entry rate by the end of December 2021, with a reversal by the end of January (Figure 2b).


Moving forward, quasi-real-time business opening and closure rates could be used as an input for indices that track the overall health of the business sector. For example, Statistics Canada constructs a Real-Time Local Business Conditions Index by combining openings and closures with real-time traffic data around businesses to proxy both for the extensive and intensive margin of business activity. Alternatively, the information on individual business opening and closure could also be combined with real-time job vacancies to investigate the impact on labour dynamics. Eventually, high-frequency business health data could also help document the effect of natural disasters that are localized in space and time.

More broadly, the data collected at a micro level could provide a finer understanding of small business dynamics. For instance the flexibility of the data collection process could allow for the investigation of the rise of online retailers operating from the owners’ residence, or the relative dynamism of downtown businesses compared to commercial areas outside the city centres.

There are a few limitations of our method to note. First, the Google Places data is continuously updated but cannot be collected back in time, limiting the ability to benchmark results to pre-pandemic levels. Second, business closures are harder to assess because a business no longer exists to confirm the timing of its closure (see our survey for business openings). Third, the quality of estimation is dependent on the quality of the data, which is controlled by Google. Nonetheless, entry and exit estimates seem to correlate with data from Statistics Canada, despite differences in definitions. Eventually, as the digitalization of the economies continue, the coverage and reliability of real-time measures of business openings and closures 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.

Thibaut Duprey works at Bank of Canada, Artur Kotlicki works in the Bank’s Prudential Framework Division, Daniel Rigobon works at Princeton University and Philip Schnattinger works in the Bank’s Structural Economics Division.

If you want to get in touch, please email us at or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.