A GIS-Based Spatiotemporal Analysis of the Relationship between the Outbreak of COVID-19, Delta Variant and Construction in Sydney and Melbourne

Abstract

The outbreak of the Delta Variant of COVID-19 presents a natural experiment without modern precedent. As authorities scrambled to control the spread of the disease in Australia’s largest cities, construction workers were allowed to keep working on site without the benefit of mandatory vaccination, unlike their peers in healthcare, defense, education or aviation. Using publicly available COVID-19 surveillance data, we analyzed the geographic spread of the Delta Variant and its relationship with construction in both cities. The period of this study covers the identification of the first case of community transmission to the achievement of 90% full vaccination in the eligible population. We show how the risk profile of construction workers varies according to socio-economic status such that Machinery Operators and Drivers were most at risk, followed by Laborers, owing to where they tend to live in each city. Moreover, these highly mobile workers may unknowingly serve as vectors for the spread of infectious disease to the most vulnerable communities in an urban setting. Remarkably, we also found that the risk profile of construction businesses can also be described similarly in terms of annual income. Sole traders and small businesses were mostly located in vulnerable areas, which presents threats to business continuity that public policy must address. We observed that the first eight weeks of an outbreak are critical; after this time, vulnerable workers and most construction businesses will see steep rises in their exposure to the risk of infection until the disease is brought under control. Accordingly, we recommend short, sharp pauses of all construction works on site to control the spread of future pandemic outbreaks once cases of community transmission are detected. Fiscal policy must support workers and small business owners, so they are not forced to choose between their health and earning a living during these periods. The government and trade unions must commit to mandatory vaccination for construction workers to safeguard their communities. Health authorities must continuously engage with particularly vulnerable workers as immunity wanes and vaccine boosters become necessary. Digital disinformation must be tirelessly countered by consistent expert medical advice at all levels of the industry.

1. Introduction

The construction industry has driven Australia’s economic recovery from the pandemic [1]. Public policies exempting construction sites from lockdowns have allowed workers to carry on while professionals work from home [2]. Problematically, these policies reflect the priorities of high-status workers while abandoning others to the brutality of heavy industry. Our research shows how the risk of COVID-19 infection is unevenly distributed across the sector by socio-economic status (“SES”) [3] and how those risks change as outbreaks unfold. This research takes advantage of a rare natural experiment [3] where international, state and municipal (i.e., Local Government Area) borders were closed under strict quarantine laws [4]. Using publicly available COVID-19 surveillance data from the health departments of New South Wales (NSW Health) and Victoria (Health Victoria), along with census data from the Australian Bureau of Statistics (“ABS”), we analyzed the relationship between the spread of COVID-19, the Delta Variant and construction in each city. The period of our study covers the detection of the first case of community transmission [5,6] and the attainment of 90% full vaccination in the local population [7,8]. This research contributes to the emerging body of knowledge on how a modern pandemic impacts the industry by addressing the following questions:

I.

What are the spatiotemporal relationships between the spread of COVID-19 and construction workers?

II.

What are the spatiotemporal relationships between the spread of COVID-19 and construction businesses?

III.

What are the spatial relationships between the spread of COVID-19 and construction work?

IV.

How are those relationships differentiated by SES factors?

2. Background

Construction Workers have been called upon to play a special role in Australia’s economic recovery from the pandemic [1]. At the outset, a range of fiscal policies were deployed to stimulate activity throughout the sector, including fast-tracking public works and direct payments to homeowners for renovations. Throughout the crisis. works on site were allowed to continue despite little knowledge of the disease and its long-term effects. Unlike workers in healthcare, defense, education and aviation, construction workers were initially exempt from mandatory vaccination for reasons that appeared to be more ideologically driven than based on medical facts.

The Delta Variant of Sars-CoV-2 was first detected in the affluent suburbs of Sydney’s east due to critical failures in the state’s international quarantine system [5] before spreading to the working-class suburbs of the west. This prompted closures of state and Local Government Area (“LGA”) borders to stem the spread of disease. Amid growing concerns, construction workers were a factor behind the spread of the virus among vulnerable communities. The New South Wales Government was forced to suspend construction works for several months. During this period, construction workers were urged to get vaccinated, and those from LGAs with especially high infection rates were required to provide evidence of vaccination in order to return to work in other parts of the city.

In Melbourne, the appearance of the Delta Variant was traced back to an international quarantine hotel in the city’s Central Business District (“CBD”) [6] Like Sydney, the virus quickly spread to Melbourne’s working-class suburbs despite a swift return to lockdown. Construction work was also suspended following findings of widespread non-compliance with COVID-19 Safety measures, and a deadline by which all workers were required to provide evidence of their first and second vaccination as a condition for being able to return to work was imposed, which triggered protests by some workers over several days [9].In both cities, safety measures set by health authorities included [10,11]:

• requirements to register a COVID Safety Plan with the relevant state health authority for each site;
• providing every employee with a digital copy of the COVID Safety Plan;
• installing and maintaining posters with official health information on site;
• digital check-ins using the state government’s QR Code system as a condition of entry on site;
• provision and mandatory use of PPE;
• maintaining one person per square meter;
• provision of hand sanitizer and exclusive washing facilities for workers; and
• appointment of a dedicated COVID Safety Marshall to monitor compliance.

These were substantively similar to the safety measures imposed on all businesses allowed to operate during lockdown. Nevertheless, when these measures failed to deliver the desired changes in behavior on construction sites and it became necessary to stop construction so as to slow the spread of disease, the decision was consistently framed by political commentators in terms of costs to “the economy” rather than lives saved in an emergency [12,13].

Striking a balance between health and economic policy has been a source of political tension throughout the pandemic. Pressure to skew regulations in favor of commercial interests has yielded dramatic reversals of policy [14], which has drawn criticism from experts concerned with not losing the hard-fought ground in controlling the spread of COVID-19 [15]. Overcoming the rhetoric of corporate predation is made all-the-more difficult by the way in which the powerful actively represent ordinary workers as expendable. For example, “dying for the sake of the economy” was an idea espoused by politicians and commentators in the United States [16]. The British Prime Minister infamously said, “let the bodies pile high!” [17]. No wonder peak industry bodies in this country have been forced to raise concerns about the toll COVID-19 is taking on a sector plagued by mental distress and ill-health [18].

With that in mind, our aim is to refocus the discourse on workers and their communities in order to show how the epidemiology relates to industry in social terms. Our objective is to quantitatively analyze the vast amounts of COVID-19 surveillance data published by health authorities in this country in light of existing data on the construction industry.

3. Theoretical Background and Literature Review

3.1. Theoretical Background—Social Epidemiology

To achieve the aims and objectives of this research project, approaches from the field of Social Epidemiology have been employed to study the available data. Social epidemiology is a branch of preventative medicine that considers the patterning of a disease among entire populations as determined by a range of factors including the environment, economy and politics.

A key theory in this field is that of Fundamental Cause, which argues that life expectancy is primarily driven by socioeconomic status (“SES”) such that one’s position in the prevailing social order is the fundamental determinant of health outcomes [3]. Moreover, health outcomes are distributed along a gradient favouring the wealthy, who tend to apply increasing amounts of economic and political resources towards protecting their well-being and interests. If unchecked, this can lead to pernicious inequities in healthcare systems. Social epidemiologists identify three key factors which determine SES and drive mortality in the long term: education, income and employment. They argue that, irrespective of the particularities of the risk in question, low-status individuals and their communities are more likely to bear the burden of disease than others [3].

Another theory upon which this research has relied is Rose’s Paradigm, which argues that small changes in risk can have profound impacts when distributed throughout a given population (See: Figure 1). Therefore, the aim of social epidemiology is to provide recommendations for SES interventions, which are likely to improve rather than impair health outcomes at the population level [3].

There are, of course, several challenges associated with adopting these approaches, which the design of this research has accommodated.

Firstly, the relevant health outcome, the “risk”, must be clearly defined so that an analysis of upstream and downstream policies is possible (see Figure 2) [3]. With that in mind, this research concerns the risk of COVID-19 infection, the (upstream) decision to exempt construction sites from lockdowns and the (downstream) need to suspend construction to impose mandatory vaccination when non-compliance with safety regulations was found to be widespread on sites.

Secondly, SES can be determined by a range of other factors such as gender, ethnicity, religion and sexuality, depending on the context [3]. Accordingly, this research has been designed to consider SES among construction workers by way of vocation so as to show the patterning of disease burden within the industry. To that end, data distinguishing professionals, managers, tradespersons, laborers and so forth have been obtained from the ABS and subjected to analysis.

Finally, reverse causation and confounding factors are well-established counterarguments to causal inference in social epidemiology [3]. However, this natural experiment presents an opportunity to overcome these issues insofar as the Sydney and Melbourne cases were rare occasions where all but essential workers were required to work from home, local, international, state and LGA borders were closed and only one variant of COVID-19 was in circulation at the time.

3.2. Theoretical Background—Social Epidemiology and the Construction Industry

The construction industry is comprised of many different types of laborers, tradespersons and professionals living and working in a wide range of environments and social conditions. Understanding which types of workers are more vulnerable to disease and to what degree will allow for nuanced responses to a variety of health crises affecting the sector. At present, what little research there is treats the sector as a homogeneous population, with little appreciation of how bodies can be marked by SES within a heavy industry setting in this country.

The built environment itself also plays a significant role in the spatial patterning of disease disadvantage [19], and, in response, social epidemiologists have been early adopters of Geographic Information Systems (GIS), Artificial Intelligence and Big Data [3]. Ironically, these tools are commonly available to construction industry practitioners, and Australia is a highly digitized society with large, publicly available datasets on many aspects of the human condition.

Social epidemiology offers conceptual tools and techniques for understanding the pandemic and its impact on the construction industry beyond business-as-usual projects and supply chain management. We do not propose merely mimicking the work of academics from that discipline; rather, we call on practitioners to recognize that to speak of the construction industry is to also speak of workers and their communities. Industry stakeholders have already begun talking openly about depression and suicide, which is shockingly common within the sector [18]. Further collaboration with data engineers, life sciences experts and industry stakeholders could transform the sector by reframing the political discourse in ways that actually benefit workers.

3.3. Theoretical Background—Systematic Literature Review

3.3.1. Systematic Search

A systematic literature review (“SLR”) builds a solid foundation for all research projects. It establishes what studies have been conducted so far and what gaps in the research remain [20]. SLRs also provide scholars with an opportunity to record how their research unfolded in a transparent manner. A systematic search of the literature was carried out using SCOPUS and Web of Science (WoS) [21]. SCOPUS is a well-established database covering scientific research across a wide range of disciplines, and WoS is particularly useful for its in-depth coverage of construction technology as a discipline in its own right. Cross-checking both databases reduces the risk of missing relevant studies given that most articles published on WoS are also available on SCOPUS [21]. The following query string was used on SCOPUS:

(TITLE ((construction OR mining) AND (industry OR site OR workplace OR worker* OR employee*)) AND TITLE ((covid* OR sars-cov-2 OR pandemic))) AND (LIMIT-TO (AFFILCOUNTRY, “United Kingdom”) OR LIMIT-TO (AFFILCOUNTRY, “United States”) OR LIMIT-TO (AFFILCOUNTRY, “China”) OR LIMIT-TO (AFFILCOUNTRY, “Australia”) OR LIMIT-TO (AFFILCOUNTRY, “Italy”) OR LIMIT-TO (AFFILCOUNTRY, “Singapore”) OR LIMIT-TO (AFFILCOUNTRY, “Japan”) OR LIMIT-TO (AFFILCOUNTRY, “Canada”) OR LIMIT-TO (AFFILCOUNTRY, “Chile”) OR LIMIT-TO (AFFILCOUNTRY, “South Korea”) OR LIMIT-TO (AFFILCOUNTRY, “Sweden”) OR LIMIT-TO (AFFILCOUNTRY, “Germany”) OR LIMIT-TO (AFFILCOUNTRY, “New Zealand”) OR LIMIT-TO (AFFILCOUNTRY, “Portugal”) OR LIMIT-TO (AFFILCOUNTRY, “Switzerland”)) AND (LIMIT-TO (DOCTYPE, “ar”)) AND (LIMIT-TO (LANGUAGE, “English”))

This returned 39 articles which were eligible for further analysis. A similar search was carried out on WoS, which identified 42 articles suitable for inclusion in the SLR for this study. Figure 3 illustrates the method by which articles were progressively filtered using the PRISMA standard.

The results were filtered to include only studies carried out in industrialized countries with comparable practices and workplace conditions to those of Australia, notwithstanding the strikingly different political approaches of countries such as the United States and the United Kingdom to the Pandemic. References to mining were included to cast as wide of a net as possible, given that mines are an important part of the building industry supply chain, and many within the sector work on remote mining operations on a Fly-In-Fly-Out (FIFO) Basis [23].

A total of 42 articles were included in the SLR for this research, and we subjected these articles to a content review. In the first instance, abstracts were reviewed and screened to eliminate papers focused on commercial, legal or technical topics not directly relevant to the direction of our research (n = 18). The remainder (n = 24) were subjected to a full content review, of which most (n = 17) related to mitigating the effects of public health regulations on the delivery of construction and infrastructure projects underway at the outset of the pandemic. The measures discussed tended to include non-pharmaceutical, hygienic interventions on site and working remotely to the extent that mobile computing permits. Remarkably, the literature is silent on vaccination, whether mandatory or not. Only a handful of particularly useful studies related to the health and wellbeing of workers (n = 7). These are reported in the following section of this paper.

3.3.2. Content Review

An important distinction must be made between the pandemic, a geo-political event, and the actual effects of the COVID-19 disease on human beings and their communities. Only one article describing the condition of a construction worker in India who presented with severe abdominal pain and thrombolytic complications was identified by the search. Apart from the work of a small number of American academics looking at the Pandemic’s impact on certain minorities within the building industry [24,25], little has been done to fill this gap in the research internationally.

Prizadeh and Lingard [4], in a longitudinal online multi-wave survey of Australian construction industry professionals (n = 412) required to work from home under public health orders, identified a consistent, gradual decline in mental health among employees working from home, which is mitigated by satisfaction with the work–life balance over time. Taking into account professional tasks such as attending meetings online and coordinating works on site, alongside factors such as diet, physical activity, mental well-being and sleep, the authors noted the detrimental impact of isolation combined with anxiety and constant media coverage of the pandemic. While some professionals expressed a preference for working remotely, the study also found that there was no significant difference in the mental health outcomes of those participants who worked in office settings during the period of the study, suggesting an ambiguous relationship between mental health and work location. A nuanced view, the authors note, must take into account the personal circumstances of each individual, such as marital status, family composition, caring responsibilities and so on.

In another Australian study, Oo and Lim [5] carried out an online survey of female construction industry professionals (n = 102) in relation to their perceptions of the pandemic’s impact on their careers over 6 months. They noted the significant changes to working arrangements under public health orders requiring employees to work from home whenever possible, in addition to general childcare responsibilities and home-schooling. The study found that, during the period, most participants remained employed (96%), with around a third reporting increased hours (27.6–34.7%). Interestingly, some participants reported working fewer hours (13.3%), citing improved work efficiency and fewer distractions. Notwithstanding the profound changes to working location and responsibilities imposed by the public health orders, participants reported only modest negative impacts and were confident in their ability to remain employed in the long term.

Asare et al. [23] conducted a longitudinal study on the mental health of workers engaged on a fly-in-fly-out (FIFO) basis by remote mining operations in Western Australia. They found that, among the participants, levels of mental distress varied such that older, experienced workers were more resilient to the hardship of social distancing and quarantine. In the outline of their research, the authors noted the difficulty in obtaining data relating to educational attainment or other SES indicators due to participants’ reluctance to voluntarily provide such information.

Afkhamiaghda and Elwakil [26] modeled the spread of COVID-19 among construction workers based on demographic and commercial data such as population density, localized investment in construction and worker distribution across the country. In particular, the authors determined that the age of the workforce placed it at a special risk of infection.

Baker et al. [24] analyzed the burden of the disease on workers based on national employment data, taking into account the frequency of reported exposure to COVID-19 in a workplace setting in that country, and found that construction and extraction industry workers represented 8.9% (n = 491,990) of the total population of workers exposed to COVID-19 more than once per month.

Bui et al., 2020 [27] considered the disproportionate burden of the disease on immigrant and ethnic communities in Utah, where it was found that construction sites represented 15% of workplace outbreaks in that state. The authors noted systemic factors that unduly disadvantaged construction industry workers—in particular, the lack of sick leave entitlements.

Similarly, Pasco et al. [28], found that unrestricted construction activities in Texas increased hospitalizations by over 4000% among construction workers and by almost 400% among residents. Wearing PPE, sanitization and monitoring onsite reduced transmission by 50%.

Bou Hatoum et al. [29] studied complaints (n = 1382) among construction workers registered with the Occupational Health and Safety Authority (OSHA) and found that most complaints were from workers involved in non-residential construction in California, Oregon and Indiana and related to personnel who had tested positive for COVID-19 yet reported to work, the failure to comply with mask mandates and practice social distancing as well as the failure by employers to provide PPE and sanitization.

3.3.3. Literature Review—Filling the Gaps

Our research aims to fill gaps in the literature by providing a large-scale study of COVID-19′s impact on construction in two major metropolitan areas over several months using high-quality, geocoded census, planning and COVID-19 surveillance data. Our study is unique, as it incorporates disaggregated data which differentiate construction workers and businesses according to SES factors. We identify what kinds of construction workers and businesses are most at risk and how that risk changed over time. We also show the relationship between the intensity of construction work and where the disease was recorded. To the best of our knowledge, no studies to date have been conducted using a similar method which focuses on the construction industry as a heterogenous, stratified population of vocations, business and activities.

4. Materials and Methods

4.1. Study Area

This study includes Australia’s two largest cities, Sydney and Melbourne (see Figure 4). These case studies have been selected because they were the only capital cities to suffer from major outbreaks of the Delta Variant of COVID-19 under Australia’s strict policy of international and state border closures and quarantine. Both cities are located on Australia’s East coast and serve as this country’s major economic engines and international gateways. Together, they represent 40% of the total population and contribute approximately AUD 1.2 trillion to the GDP (EUR 780 billion) [30].

Sydney, with a population of approximately 5.3 million, is the state capital of New South Wales and is comprised of 34 LGAs. Melbourne, with a population of approximately 5 million, is the State Capital of Victoria and is also divided into 34 LGAs.

LGAs are subdivisions of State and Territory governments which have been granted administrative authority to carry out municipal functions such as the maintenance of roads and public parks in addition to the power to collect rates, grant approvals for construction works and impose fines for breaches of local government regulations. LGAs also serve as statistical units for the national census and are the basis upon which COVID-19 surveillance data have been geographically collected nationwide. Additionally, LGAs also served as internal boundaries for restricting movement during the Delta outbreak. For example, residents of Sydney’s most severely infected LGAs were required to stay within those areas under Public Health Orders.

4.2. Study Period

Our study covers the detection of the first case of community transmission of the Delta Variant in Sydney and Melbourne to the attainment of 90% full vaccination in the local population. Which can be traced back to critical failures in the international quarantine systems of both cities in May of 2021. It also spans the Australian winter, when the spread of the virus was facilitated by seasonal changes in behavior among the general public: such as working and gathering in poorly ventilated indoor settings. It is concluded by a dramatic shift in policy from contact tracing each case of infection, conscientious mask wearing and generally looking out for one another to “assuming responsibility for one’s own safety” and “learning to live with the virus”, which was brought about by a change in government in New South Wales [31].

4.3. Data

De-identified COVID-19 surveillance data from this research have been obtained from datasets published by NSW Health and Health Victoria. Demographic and employment data have been obtained from the ABS. Business data, including the number of construction businesses, size and turnover, have been also obtained from the ABS. Planning data which set out the number and value of development approvals and construction certificates in LGAs have been obtained from the NSW Department of Planning and the Victorian Building Authority.

4.4. Pearson’s Correlation Coefficient

Pearson’s Correlation Coefficient [32,33] was used to test the strength of the relationship between the variables in this research. The number of infections in an LGA was compared to:

i.

the number of construction workers;

ii.

the number of construction businesses;

iii.

the number of construction projects; and

iv.

the total value of construction work underway.

In studies (i.) and (ii.), the variables were disaggregated by SES such that construction workers were separated by vocation (i.e., professionals, managers, laborers, etc.) and construction businesses were distinguished by size (i.e., number of employees and turnover). Additionally, we tested changes in the relationship between infections and construction workers over the period of study to better understand the evolving risk profile of persons of differing SESs within the sector. By carrying out tests (iii.) and (iv.), we sought to better understand the relationship between construction activity and the spread of the disease across Sydney and Melbourne.

4.5. Data Visualization—GIS Analytics

GIS Analytics have become an indispensable part of social epidemiology [3]. For this study, Datawrapper, QGIS and ArcGIS were used to produce mono-variate and bivariate choropleth maps using the data obtained. Both techniques use color ramping to show how the geocoded data overlap and diverge in a way that is simple and informative. Graduated symbology was also used to impactfully illustrate the relationships between datasets.

5. Results

5.1. Sydney

Our analysis of the data showed that, in Sydney, infections were only weakly correlated with Construction Workers in general (ρ = 0.4). However, further analysis showed the relationship to be variable based on SES factors. Of the various types of construction workers, Machinery Operators, Drivers and Laborers (ρ = 0.6) were most at risk. Technicians and Clerical and Administrative Staff were moderately at risk (ρ = 0.5). Professionals and Managers, as expected, were very weakly correlated with infections (ρ = 0.07 and ρ = 0.2, respectively). In Figure 5, the relationship between the most exposed construction workers, Machinery Operators and Drivers, and infections across Sydney is illustrated using bivariate mapping. As can be seen, the incidence of the Delta Variant of COVID-19 is strikingly similar to the distribution of workers known to be exempt from public health orders who travel from site to site in the conduct of their trade.

The results from our investigations also showed how the correlation between construction workers and infections changed over time. In the first month of the outbreak, all workers’ exposure to risk dropped sharply, particularly among Machinery Operators, Drivers and Laborers. This is probably because the Delta Variant first entered the local population via a negligent quarantine worker living in the affluent eastern suburbs of Sydney. However, as the outbreak continued, we observed a steep rise in all workers, except for Professionals and Managers, over four months from June to October. By October, 6 months into the outbreak, the risk profile for vulnerable construction workers peaked and started to fall as the total vaccination rate approached 90% for the population. Throughout the outbreak, it was also observed that the risk profile of Professionals remained below its original levels. In Figure 6, the change in correlation between infections and construction workers’ SES is plotted over time. As can be seen, low-status workers were exposed to a mounting risk as the outbreak persisted.

Surprisingly, the data also revealed a weak relationship between infections and construction activity based on the number of construction certificates issued per LGA (ρ = 0.09) and the total value of Development Approvals per LGA (ρ = 0.04), which was contrary to expectations. However, the data revealed a strong relationship between infections and business counts per LGA (ρ = 0.8). Similar to construction workers, the relationship is variable based on SES factors such that Sole Traders bore the highest risk of infection (ρ = 0.8), while businesses with 1–19 employees (ρ = 0.7) and 20–199 employees (ρ = 0.6) were moderately at risk. Businesses with over 200 employees bore the least risk (ρ = 0.4).

These findings were echoed in an analysis of business size in terms of income. The data revealed a strong relationship between turnover and exposure to risk. Small companies with an annual turnover of less than AUD $200,000 had the highest exposure (ρ = 0.8), medium-sized companies with incomes ranging from AUD $200,000 to AUD $5,000,000 were also at a high risk (ρ = 0.7), medium-sized companies with a turnover between AUD $5,000,000 and AUD $10,000,000 sustained a moderate risk and companies with incomes over AUD $10,000,000 were at a low risk (ρ = 0.4).

In Figure 7, these relationships are illustrated spatially such that, similar to construction workers, the footprint of infections closely resembles the distribution of construction businesses.

The data for Sydney construction businesses also returned a similar pattern to that of construction workers’ correlation with infections over time. According to the data, at the start of the outbreak, construction businesses of all sizes experienced a sharp drop in risk, followed by a steep rise in risk over six months, peaking in October as full vaccination of the eligible population approached 90%. In the case of businesses, the only outlier was large companies with annual turnovers of AUD $10 million or more, whose risk profile remained relatively low and stable over the outbreak (see Figure 8).

To better understanding the geographic distribution of workers and businesses, their relationship and the pattern of disease across the city, the data again returned useful results. While the data show that where construction workers live is weakly correlated with the number of construction projects (i.e., construction certificates issued) in LGAs (ρ = 0.4), their population is strongly correlated with the total value of Development Applications in LGAs (ρ = 0.9). This relationship is consistently high across SES groupings. Similarly, where construction businesses of all sizes (i.e., income brackets) are located is also strongly correlated with the value of construction work in LGAs (on average, ρ = 0.7).

Based on the above, we estimate that, in Sydney, around 19,000 Machinery Operators, Drivers and Laborers are at a high risk (i.e., 12% of the labor force). Between 15,000 and 18,000 sole traders and small businesses are in LGAs most likely to be affected by future outbreaks (i.e., up to 20% of construction businesses).

5.2. Melbourne

The Melbourne Data also showed a weak correlation between infections and construction workers as a whole (ρ = 0.4). However, further investigation also revealed variable relationships in the data based on SES factors. Unlike Sydney, the data revealed that, in particular, the distribution of Machinery Operators and Drivers was even more strongly correlated with the spread of the Delta Variant in that city (ρ = 0.8). Laborers were also moderately at risk (ρ = 0.6). All other categories of workers were exposed to relatively low levels of risk. As expected, professionals (ρ = 0.01) and managers (ρ = 0.2) were very weakly connected to the spread of the Delta Variant, as were clerical staff and administrative workers (ρ = 0.3). Similarly, technicians and trades workers were also weakly correlated with infections (ρ = 0.4). In Figure 9, the relationship between infections, Machinery Operators, Drivers and Laborers is illustrated spatially. As can be seen, the footprint of the disease tends towards the Northwest and Southeast of the city in a similar pattern to that of construction workers.

Changes in the correlation between construction workers and the spread of COVID-19 were also distinctly different in Melbourne compared to Sydney. In the first two months, the risk profile of almost all classes of construction workers spiked, with Professionals and Managers most exposed.

Only Machinery Operators and Drivers experienced a sharp drop in exposure to risk. Again, these results probably account for how the Delta Variant gained a foothold in that city—that is, through failures in an international quarantine hotel in Melbourne’s CBD. Nevertheless, over the following four months, that position changed such that their exposure increased steeply above all other workers, peaking in October as the population approached 90% full vaccination. Laborers also suffered a mounting risk over the course of the outbreak, while all other construction workers experienced relatively low levels of exposure, as illustrated in Figure 10 below.

As was the case in Sydney, the spread of the Delta Variant in Melbourne appears to be moderately correlated with the distribution of construction businesses as a whole (ρ = 0.6). This can be differentiated by SES factors such that sole traders were the most at risk according to the data (ρ = 0.7). Businesses with 1–19 employees (ρ = 0.5) and 20–199 employees (ρ = 0.5) were moderately exposed in equal measure, while large companies with over 200 employees exhibited a very weak negative correlation with the spread of the Delta Variant in Melbourne (ρ = −0.1).

Moreover, the correlation between business and the Delta Variant also varied according to income. Surprisingly however, very small construction businesses with incomes between AUD 0 and AUD 50,000 were only moderately at risk (ρ = 0.5), while small business with incomes between AUD $50,000 and AUD $200,000 (ρ = 0.7) and AUD $200,000 and AUD $2,000,000 (ρ = 0.6) bore the highest risk of infection. Medium-sized companies earning AUD $2,000,000 to AUD $5,000,000 (ρ = 0.4) and AUD $5,000,000 to AUD $10,000,000 (ρ = 0.4) were weakly correlated with disease, and, as expected, large companies earning over AUD $10,000,000 (ρ = 0.1) were very weakly related to the spread of infections in LGAs.

Remarkably, we observed a similar pattern when we analyzed changes in the relationship between business size and infections over time to that of construction workers. In the first two months, the risk profile of most construction businesses spiked and then saw steady increases over the duration of the pandemic. Companies with an income of AUD $10,000,000 or more saw the most dramatic increase; however, by the third month, their exposure remained at very low levels. Small companies with incomes between AUD $50,000 and AUD $200,000 and between AUD $200,000 and AUD $2,000,000, on the other hand, experienced a sudden drop in their exposure within the first month followed by a steady increase over the duration of the pandemic, plateauing as the rate of vaccination for the eligible population approached 90% in Melbourne (see Figure 11 and Figure 12).

Interestingly, unlike the Sydney data, the Melbourne data revealed that infections were strongly correlated with the number of Building Permits (the equivalent of Construction Certificates in Victoria) in LGAs (ρ = 0.7) and not with the value of work underway (ρ = −0.1). Figure 13 and Figure 14 below outline the distribution of construction projects underway in Melbourne and Sydney at the time compared to the footprint of the Delta Variant, which aligns with observations on the ground.

Anecdotally speaking, both cities had major infrastructure works underway during the outbreak of the Delta Variant. In Sydney, transport infrastructure projects worth AUD $29.5 billion were underway at the time [34]. In Melbourne, the city had undertaken the removal of dozens of at-grade railway crossings citywide to improve public safety, which was estimated to cost AUD $8.5 billion [35]. In addition to the construction of AUD $1.8 billion in new prisons and correctional facilities [36]. In both cities, major public works were troublingly close to, or within, LGAs most burdened by the disease. As mentioned earlier, we know that both governments took steps to address concerns about the spread of COVID-19 on construction sites, with the NSW government going as far as to dedicate an entire day to vaccinating as many construction workers as possible at its mass vaccination hub at Sydney’s Olympic site, dubbed “Super Sunday” [37]. This shortly preceded a peak in the risk of infection faced by construction workers on site. Victoria took the radical step of imposing a state-wide mandate for all construction workers to address the issue.

6. Discussion

This article presents findings from our study of the outbreak of the Delta Variant of COVID-19 in Australia’s two largest cities and addresses our proposed spatiotemporal analysis and relationships between the location of infections and construction.

Publicly available health, demographic and planning data were analyzed and mapped using GIS. The strength of the relationship between the various factors (construction workers, businesses, the value of construction work and construction certificates) was tested using Pearson’s Correlation Coefficient. We also studied how the strength of that relationship changed over time in each city and differentiated construction workers based on SES factors according to the rule of Fundamental Cause. Analogous tests were also carried out in relation to construction businesses based on their size (i.e., the number of employees) and annual turnover (reported income). The period of study covers the identification of the first case of community transmission and the attainment of 90% full vaccination of the eligible population.

To date, few studies have empirically considered the evolving relationship between the footprint of COVID-19 across major cities and their population of construction workers and businesses according to SES. Moreover, we are not aware of any studies that examine these events under a strict policy of border closures and quarantines similar to that pursued by the Australian government. In that regard, our research takes advantage of a large-scale natural experiment without modern precedent.

We found that, during the outbreak of the Delta Variant, machinery operators and drivers were the construction workers most at risk for infection in both Sydney and Melbourne, followed closely by laborers. While it is tempting to assume that their exposure can be attributed the nature of their work, we have no empirical data to support that finding at this time. The better view is that these workers are lower on the SES ladder and tend to live in LGAs that are more vulnerable to outbreaks of infectious disease. Exempting machinery operators, drivers and laborers from lockdowns without mandatory vaccination may create vectors by which the disease can enter those LGAs despite lockdowns, unduly burdening their communities.

When outbreaks occur, the first eight weeks are critical to the health of the construction industry. During that period, all construction industry workers may experience dramatic changes in their risk profile depending on the circumstances of the outbreak. After this, workers on the lower end of the SES spectrum are likely to experience a steep rise in their exposure to risk over several months until the outbreak is brought under control. Workers higher on the SES scale will experience relatively low levels of exposure over the duration of an outbreak, notwithstanding any spikes which may occur initially due to when and where the virus first appears.

Remarkably, the correlation between construction businesses and infections mirrors the risk profile of workers such that, in the initial eight weeks of an outbreak, most businesses may see dramatic changes to their risk environment depending on the circumstances under which the virus is first detected in the community, after which time sole traders and small businesses are likely to see mounting risk as the outbreak continues. Large construction businesses are likely to enjoy relatively low exposure to the risk of infection, as they tend to be located in LGAs higher on the SES scale.

Based on our research, we estimate that, in Sydney, around 5.6% of the total construction workforce (over 8600 machinery operators and drivers) were at a very high risk, in addition to a further 13.8% of the workforce (21,373 laborers) at a high risk during the Delta Variant outbreak. Similarly, only 1.5% of construction businesses (1449 out of 95,040) experienced very low exposure.

These figures were echoed in Melbourne, where we estimate that around 5.6% of the building industry (7789 machinery operators and drivers) were at a very high risk, on top of an additional 13.2% of high-risk individuals (18,330 laborers). Similar to Sydney, only 1.5% of construction businesses in Melbourne (1328 out of 88,980) experienced low levels of exposure to the risk of infection based on their environment.

This research is based on the outbreak of the Delta Variant, which is no longer the dominant strain of COVID-19. The period of study also covers a public policy environment of border closures strict quarantines, which has given way to lax public health measures and “learning to live with the virus”, despite the appearance of new sub-variants such as Omicron BQ1 and XBB, which once again threaten healthcare systems worldwide. Research must keep pace with the pandemic as it unfolds. Moreover, outbreaks of other diseases such as Monkeypox, Polio and Marburg Virus must be accounted for as the risk environment intensifies.

The demographic data used for this study are based on the 2016 census, which is due to be updated by the publication of the 2021 census later this year. Additionally, the reporting regime for COVID-19 has changed to include both Polymerase Chain Reaction (PCR) tests and Rapid Antigen Tests (RAT), which may impact the reliability of the results given that it is up to individuals to self-report positive RAT test results.

Based on our findings, we recommend short, sharp pauses of all construction works on site to control the spread of future pandemic outbreaks as soon as cases of community transmission are detected. Fiscal policy must support workers and small business owners so they are not forced to choose between their health and earning a living during these periods. Governments and trade unions must commit to mandatory vaccination for construction workers in order to safeguard their communities. Additionally, health authorities must continuously engage with particularly vulnerable workers as general immunity wanes and vaccine boosters become necessary. Digital disinformation must be tirelessly countered by consistent expert medical advice at all levels of the industry.

This study is useful for public health authorities and industry advocates and can be applied to a range of diseases. Future studies could consider other periods of the pandemic, as defined by the dominant strain, the availability of vaccines and changes in public policy settings.

7. Conclusions

This study has taken advantage of a natural experiment without modern precedent. Over seven months in 2021, while international borders were closed and cases of COVID-19 were virtually nil, an outbreak of the Delta Variant occurred in Sydney and Melbourne, which triggered closures of Australia’s internal borders in state-led efforts to contain the spread of the disease. Despite this, construction workers were called upon to keep working, without the benefit of mandatory vaccination. This study analyzed publicly available COVID-19 surveillance data, census data and planning data to better understand how the spread of the Delta Variant related to construction in both cities. We found that machinery operators and drivers were at a very high risk of infection owing to their socio-economic status and where they tend to live in both cities. We also observed a similar pattern of disease disadvantage among construction businesses based on size and income. In both cities, the first eight weeks of an outbreak are critical. After that period, the disease is likely to make its way into vulnerable communities, steeply increasing the risk profile of construction workers, sole traders and most construction businesses. Based on our findings, we recommend:

• Snap lockdowns when new pandemic outbreaks occur;
• Fiscal policies that support workers, sole traders and small businesses in the event that it becomes necessary to stop works on site;
• A serious commitment to mandatory vaccination for all construction workers;
• Ongoing engagement with particularly vulnerable workers to prevent disinformation and issue-fatigue.

This research is useful for industry academics and advocates advising policymakers tasked with anticipating new waves of COVID-19 and future outbreaks of other communicable diseases. Despite wishful thinking, the pandemic is not behind us. In fact, we have seen more infections and deaths in the last six months than in the last two years in this country alone.