Case-study: Impact of 2010-2011 Queensland floods on businesses
Share this post
Let’s apply DiD technique to calculate the impact of 2010-2011 Queensland floods (Australia) on businesses.
A series of floods occurred in Queensland from November 2010 to January 2011, triggered by various storm cells such as Cyclones Tasha, Anthoy, and Yashi (Queensland Reconstruction Authority 2011).
2010-2011 Queensland flood
The extensive flooding led to the designation of over 78% of the state as a disaster zone by March 11, 2011, covering an area larger than the combined size of France and Germany.
Want to keep
Deakin University online course,
Natural Disaster Recovery and Management
Despite the evacuation of thousands, the disaster resulted in 37 fatalities (Queensland Government 2011). Additionally, it caused significant physical damage, including the impairment of 9,170 kilometres of Queensland’s state road network, damage to 29% of the Queensland rail network, power outages affecting over 478,000 homes and businesses, harm to 54 coal mines, closure of 11 ports, 279 national parks, and 411 schools (Queensland Reconstruction Authority 2011; World Bank 2011).
The findings presented in this step are drawn from a study conducted by Hannah Nguyen, a researcher from Deakin University.
The dataset employed in this research is the Business Longitudinal Analysis Data Environment (BLADE), provided by the Australia Bureau of Statistics. BLADE serves as a comprehensive economic data tool offering insights into the performance of the Australian economy and businesses over time. The coverage periods of these datasets can vary, ranging from 2001 to 2022, depending on the specific data item. For this study, data from 2006 to 2015 are utilised, encompassing five years before and after the occurrence of the flood.
Various datasets within BLADE are utilised in this study, including the Indicative Data Items, the Business Activity Statement, the Pay as You Go, the Business Income Tax, the Birthdate, and the Business Locations Data. It is important to note that the number of observations may vary across datasets due to information being collected from diverse administrative sources.
Treatment and control group
Based on the Business Locations Data, a business is assigned in the treatment group if its business address falls within the borders of one of the four affected Local Government Areas (LGAs) – Brisbane, Ipswich, Lockyer Valley, and Somerset – in the Brisbane River catchment area during July to September 2010, which is the quarter preceding the Queensland floods.
The research team identified the control group to be the Swan River catchment area (incorporating Perth), the Yarra River catchment area (incorporating Melbourne), the Parramatta River catchment area (incorporating Sydney), and the Torrens River catchment area (incorporating Adelaide). These control groups serve as a reference for the trajectory that would have unfolded for businesses residing in the Brisbane River catchment area Local Government Areas (LGAs) had the floods not taken place. This comparison allows for the calculation of any deviations in businesses’ indicators –whether losses or gains –that can be attributed to the occurrence of the floods.
We identify 7,368 businesses (22% of the sample) in the treatment group and 25,765 businesses (78% of the sample) in the control group that we use for DiD calculation below.
The tables and figures required for conducting the DiD calculation in this case study can be found in the case study document.
Overview of business indicators
Table 1 display statistics (mean value) of businesses’ indicators including sales, good and service tax (GST payable), intermediate inputs, non-current assets, full time equivalents (FTE) and wage expenses by year and treatment status for the period 2006-2015. Please note FTE is number of full- time equivalent employees, and all dollar figures are measured in 2011 Australian dollars (AUD).
As the value of these indicators are continuous and can fluctuate significantly from year to year, we often use log point of these variables to minimise outliners (observations that are largely different from the majority of other observations). FTE is one exceptional variable as its value is count instead of real number, so we could use FTE without the need to transform into log (but the results are not sensitive to using log and level of FTE). Table 2 displays the value of the mentioned variables in log points.
Visual analysis and parallel trend assessment
Figures from 1 to 6 display indicators from Table 2. In each panel, the left-hand side figure plots original data (meaning data from Table 2) with dotted curve representing mean value of control group and solid curve is for the treatment group.
To assess the degree to which the parallel trend assumption holds true, the figures on the right-hand side depict a downward shift of the dotted curve. As observed in these figures, the parallel trend assumption is met for a span of three years preceding the floods (2008-2010) for all indicators, with the exception of intermediate input expenses and non-current assets, which tend to align for a period of two years prior to the floods.
Analysis and calculation methodology
Guided by visual investigation from figures 1 to 6 that the parallel trend assumption holds for almost all studied outcomes, we proceed to calculate the impact of the Queensland floods on businesses’ operation using DiD with three years preceding the Queensland floods as reference to calculate the changes. In particular, we calculate the average of mean value for both the treatment and control groups spanning the years 2008-2010 (refer to Table 2 for the relevant data) to establish the pre-flood value. This period represents a time when neither the treatment nor control group has been influenced by the floods (similar to points Yt1 and Yc1 in our diagram in previous step). Next, we’ll subtract the value of each indicator for each year post disasters from 2011-2015 for the pre-floods’ value (similar to points Yt2 and Yc2 in the previous step). Table 3 reports the above-mentioned calculation for control group in Panel A and treatment group in Panel B.
DiD calculation and impact analysis
Finally, Table 4 illustrates the DiD calculation, achieved by deducting the treatment’s changes (pre and post flood) from the control’s changes for each outcome and each year.
Take the first cell in column (1) as an example. The figure -0.095 indicates that, compared with the control group, the treatment group witnessed 9.5% reduction in sales in year 2011, which is attributed to the impact of the floods. We can see that the impact of the Queensland floods on sales is temporary, resulting in a loss for the treatment group in the year 2011. Subsequently, sales rebounded swiftly, as indicated by the minimal difference between the treatment and control groups from 2012 to 2015. The DiD results for sales largely align with the observations from Figure 1.
Affected businesses witness a decline in GST payable in the years 2011 and 2012, potentially attributed to the extension granted for tax payment to businesses affected by the disaster. Furthermore, there is a decrease in intermediate input expenses ranging from 3.7% to 6%, with a more substantial reduction observed in the years 2011-2013. The reduction in non-current assets, even in the years immediately following the floods, implies that businesses can promptly replace damaged assets.
It’s notable that the reduction in FTE and wage expenses is significant and does not diminish over time. To assess the relative reduction in FTE in 2011, we divide the FTE reduction of 0.487 by the mean pre-flood FTE value of the treatment group (8.45), resulting in a 6% reduction. This reduction gradually increases to 14% by 2015. Correspondingly, the reduction in wage expenses is substantial, varying from 20% to 29%.
Overall, the findings indicate a consistent decrease in labour inputs, likely attributed to the adoption of technologically advanced assets to replace the damaged ones. This transition leads to minimal losses in sales, accompanied by suggestive evidence of cost savings in intermediate inputs. In essence, businesses seem to recover rapidly and operate more efficiently in the aftermath of the flood.
Even though the results show little impact of the floods on businesses for overall sample, the impact is quite heterogenous with respect to firm size. In particular, the impact on businesses’ operation (the studied outcomes as mentioned above) tends to be persistent at least after five years after the floods for smaller businesses while medium and large size businesses rebound quickly after the flood. This is striking given that small businesses receive intensive support programs in the aftermath of disasters and should be noted for further support/intervention programs.
Reflect and consider the calculations we discussed in this step. Did you find anything a bit tricky to understand? Feel free to share any questions or uncertainties in the comments!
Queensland Government (2011) ‘Queensland Government response to the Floods Commission of Inquiry Interim Report’ Queensland Government accessed on 16 January 2024
Queensland Reconstruction Authority (2011) ‘Monthly Report, March 2011’ Queensland Reconstruction Authority accessed on 16 January 2024
The World Bank (2011) ‘Queensland recovery and reconstruction in the aftermath of the 2010/2011 flood events and Cyclone Yasi’ The World Bank accessed on 16 January 2024
Share this post
Natural Disaster Recovery and Management
Reach your personal and professional goals
Unlock access to hundreds of expert online courses and degrees from top universities and educators to gain accredited qualifications and professional CV-building certificates.
Join over 18 million learners to launch, switch or build upon your career, all at your own pace, across a wide range of topic areas.