A.2 Auckland firm performance in COVID-19 lockdown Te whakatutukitanga o ngā pakihi o Tāmaki Makaurau i te rāhui KOWHEORI-19
A.2 Auckland firm performance in COVID-19 lockdown | Te whakatutukitanga o ngā pakihi o Tāmaki Makaurau i te rāhui KOWHEORI-19
Our analysis of firm performance during lockdowns in section 6.2.3 uses data from the IDI, specifically anonymised tax data for firms from the Longitudinal Business Database.
We focus on quantifying lost sales revenue for Auckland businesses and the uneven distribution of these losses. The scale of revenue losses suggests other impacts, including financial insecurity, job loss and stress on business owners and employees.
We estimate the sales revenue impact using comprehensive Goods and Services Tax (GST) data and using Christchurch firms as a counterfactual for Auckland firm performance in the absence of lockdowns.
There is no perfect counterfactual as we cannot observe what would have happened to Auckland in the absence of a lockdown, in particular what would have happened to COVID-19 cases and the impact on businesses.
The average Christchurch firm is not like the average Auckland firm. We address this concern by matching like firms, weighting, and controlling directly for firm characteristics.
Despite substantial geographic distance between the two locations, Christchurch firms may have been affected by the Auckland lockdowns, for example, through supply chains. We show that government support, in the form of the COVID-19 Wage Subsidy Scheme (WSS), is limited for Christchurch firms during these periods compared with during national lockdowns, mitigating this concern. We cannot rule out that Christchurch firms in the tradables sector benefited from the Auckland lockdowns due to, for example, increased demand from Auckland consumers or reduced competition from Auckland firms. Given the closer proximity of other cities to Auckland, we maintain that this outcome is unlikely.
The entire South Island of New Zealand was free of COVID-19 during the period of Auckland-specific lockdowns. As such, our counterfactual represents the behaviour of firms and consumers in the absence of the risk of contracting COVID-19. This is unlikely to be the appropriate alternative outcome for Auckland, given that COVID-19 was present in the community. We interpret the estimated effects, therefore, as a bound on the potential negative impact of Auckland firms' sales. We mean this in the sense that – in the absence of a lockdown, the (unlikely) best case scenario is that the COVID-19 outbreak did not spread further. A scenario where COVID-19 spreads is assumed to be worse for the local economy and, therefore, our estimated effects are an upper bound for how much better off Auckland firms would be without regional lockdowns.
A.2.1 Data and methods | Ngā raraunga me ngā tikanga
We use the Longitudinal Business Database and related user-generated tables for most variables. Plant-level data in the Fabling-Maré labour tables63 identify active employing firm locations, and the Business Register is used to identify working-proprietor-only locations. Monthly GST sales and purchases come from the method developed by Fabling64 that apportions two- and six-monthly returns to a monthly frequency.65
GST data are additionally used to track firm exit (that is, the permanent absence of GST activity). Restricting the analysis to firm (GST) sales and purchases allows us to include working-proprietor-only firms in the analysis, which are otherwise restricted to annual frequency of observation. We exclude the government and not-for-profit sectors, for whom sales revenue is unlikely to be a good measure of firm performance or continuity.
Employer WSS data are drawn from the IDI and linked to firms using confidentialised Inland Revenue numbers. The WSS was paid to firms during lockdowns to enable them to retain workers in cases where the impact of COVID-19 and associated events was substantial.67 We use these data to confirm that we have correctly identified Auckland firms, to allay concerns that Christchurch firms were adversely affected by Auckland-specific lockdowns, and to establish industry groupings that were more and less affected by the pandemic. Because of the WSS, we do not consider firm employment outcomes, which may present an unduly rosy picture for recipients.
A.2.2 Lockdown identification | Te tautuhi i ngā wā rāhui
New Zealand operated a four-stage alert level system over the period where lockdowns occurred. Levels 3 and 4 of New Zealand's Alert Level System can be understood as 'soft' and 'hard' lockdowns respectively, because they required people to stay at home, closed schools and businesses, and involved heavy restrictions on public gatherings.67
We follow this categorisation treating all Alert level 3 or 4 periods as lockdowns, ignoring distinctions between the two levels for ease of analysis. Using the timeline reported by the Department of the Prime Minister and Cabinet,68 we summarise the periods of nationwide and Auckland-specific lockdowns. For our analytical work, we simplify this classification to the monthly level of our data by treating September 2021 as a NZ-wide lockdown month, despite most of the month being Auckland-specific.
A.2.3 Matching firms | Ngā pakihi e whakataurite ana
To establish a counterfactual for Auckland firm outcomes, we apply a combined matching and regression approach. Matching follows a similar methodology to a 2014 study of the impact of the Canterbury earthquakes on firm performance.69
Firms are located in the Auckland or Christchurch functional urban area if the majority of total labour input is located in one of those areas.70 Thus multi-location firms are included in the analysis when secondary locations are small relative to the primary location.
While the quintile-based process assures an even distribution of Auckland firms across cells (within industry-labour type), cells can have a relatively small number of Christchurch firms. We drop cells where there are fewer than one Christchurch firm for every 20 Auckland firms. Assigning a weight of one to each Auckland firm, Christchurch firms are weighted by the ratio of Auckland to Christchurch firms in the relevant cell. The weighted count of Christchurch firms, therefore, equals the (un)weighted count of Auckland firms in each cell. These weights are used throughout the analysis.
Since GST is mandatory above a minimal filing threshold, and our data-based exclusions are minor, we consider the resulting sample of 96,636 private-for-profit Auckland firms – 51,708 working-proprietor-only and 44,928 with employees – to be highly representative of the population. Results represent the average experience of Auckland firms relative to similar Christchurch firms. We track sales, purchases and firm survival from April 2019 (a year prior to the base month) through to March 2024 (the latest complete March year) – a total of 60 months. By tracking results prior to the base month, we can assess the quality of the matching by checking for parallel trends violations that might invalidate the comparison.
63 Richard Fabling and David C. Maré, Addressing the absence of hours information in linked employer-employee data. Working Paper 15–17, (Wellington: Motu Economic and Public Policy Research, 2015).
64 Richard Fabling, Still medalling: Productivity gets a bronze (data source). Working Papers 24–06, (Wellington: Motu Economic and Public Policy Research, 2024).
65 These data are updated annually by Richard Fabling, and available to all authorised users of the Longitudinal Business Database via secure datalab access. We used the 202410 instance, which were the latest available at the time of analysis. Sales and purchases are GST-exclusive and deflated by the consumers price index.
66 For an evaluation of the WSS and discussion of the related data see Dean Hyslop, David C. Maré, and Shannon Minehan, COVID-19 Wage Subsidy: Outcome evaluation, Working Paper 23–03, (Wellington: Motu Economic and Public Policy Research, 2023).
67 NZ Royal Commission of Inquiry into COVID-19 Lessons Learned: Phase One, Main Report (2024).
68 Department of the Prime Minister and Cabinet, Timeline of Aotearoa New Zealand's significant events and key All-of-Government activities. Department of the Prime Minister and Cabinet, 1 September 2023, https://www.dpmc.govt.nz/sites/default/files/2023-10/pr-timeline-significant-events-activities.pdf
69 Richard Fabling, Arthur Grimes and Levente Timar, 'Natural selection: Firm performance following the Canterbury earthquakes'. Working Paper 14–08, (Wellington: Motu Economic and Public Policy Research, 2014).
70 We rely on plant-level employment to identify the relative size of firm locations because sales/output is only available at the firm-level.