Data structures and types

There are various ways of organising data.

Below, we will explore common data structures, aiming to enhance quantitative analysis by understanding what information the data can convey and what it cannot.

Data structures

Because of differences in data availability, data is organised in different ways.

Cross-sectional data

Cross-sectional data involves gathering information about different entities or subjects during a specific event in a natural disaster. It represents observations taken at a single point in time or within a short time frame, ensuring each observation’s independence from others at different times.

An example of this is a survey carried out in Oct 2011, six months post the 2010/11 Queensland floods. The survey, conducted by Chamber of Commerce and Industry Queensland aimed to collect data on the operational status of businesses and the challenges they faced in their recovery from the flood.

Repeated cross-sectional

Repeated cross-sectional data involves the collection of data from multiple cross-sectional samples at distinct time intervals. It provides a way to examine changes in a population or phenomenon over time without following the same subjects longitudinally.

Each cross-sectional survey is conducted independently, capturing the characteristics of different individuals or entities at each time point. When examining the survey related to the 2011 Queensland floods, an example of repeated cross-sectional data could be seen in a subsequent survey conducted, for instance, in 2013. Despite the survey’s focus on similar outcomes, it incorporates a random sampling of businesses.

Time series data

Time series data involves observations on a variable or several variables over time. This type of data is collected at regular intervals, such as daily, monthly, or yearly, to track changes and trends over a specified period.

Time series data allows for the analysis of patterns, seasonality, and long-term trends. One example of this scenario could involve a study that tracks the monthly sales of a company over the last five years, aiming to observe the impact of Covid-19 on sales.

Panel data (longitudinal data)

Panel data combines elements of both cross-sectional and time series data by observing multiple entities over multiple time periods. Each entity (such as individuals, companies, or countries) is observed at different points in time, allowing for the analysis of both individual variation and time-related trends.

A case in point is a study that tracks the business performance of 100 small and medium enterprises (SMEs) in a cross-sectional manner over a 5-year period (time series), spanning from financial year 2009/2010 (pre-Queensland floods 2009/2010) to 2013/2014 (post-floods). In this study, each SME is observed at multiple time points.

Advantages of panel data

In quantitative analysis, panel data is preferred over other data structures. This type of data, enable the exploration of two dimensions of variation: the variation of outcomes across units (such as individuals, businesses, and states) and variation within a unit over time. These variations prove valuable for establishing associations between outcomes and events (as disasters for example) and facilitate the application of more quantitative techniques to reveal causal effects or stronger associations.

Panel data is especially advantageous in real-life settings where there is no control or manipulation, as seen in laboratory experiments.

Data types and sources

Quantitative analysis draws data from diverse sources, each serving distinct purposes with varying levels of quality and accuracy. It is important to carefully assess the quality and accuracy of the data before conducting the analysis.

Administrative data

Administrative data refers to information collected for routine administrative or operational purposes by organisations, institutions, or government agencies. These datasets are often generated as part of day-to-day activities and can include records of transactions, registrations, and interactions.

Administrative data is often of high quality and accuracy, though it is not collected for research purpose, so it may not contain information that is desired for studied outcomes.

Tax records, healthcare records, educational records, and employment records are examples of administrative data. In the context of disaster studies, these datasets often lack information on disaster status (affected or not affected individuals by disasters). However, they often provide details on locations (normally not at the residential level but at some geographic scale like county, village, or Statistical Area level 1, 2, 3 etc.

To identify affected regions, complementary datasets such as flood maps and bushfire maps can be utilised in conjunction with these administrative records.

Survey data

Survey data is collected by systematically gathering information from a sample of individuals or entities through structured questionnaires or interviews. Surveys are designed to capture specific information, and the data collected can provide insights into attitudes, behaviours, and characteristics of the target population.

Designed with the specific purpose of gathering study-related information, survey data has inherent advantages. However, a drawback lies in the subjectivity and self-report nature of the information, which can lead to inaccuracies or errors, particularly evident in the context of natural disasters – a situation prone to biased responses.

For example, in a survey asking people about distress and panic levels (scoring from 0 to 10 for example) during disaster strike, those with the highest level of distress or panic may not response to the survey as these people are likely to suffer from mental illness such as post-traumatic stress disorder (PTSD) which makes them less motivated to response to a survey.

Survey data is also costly and often contains smaller sample size compared to administrative data. For the Queensland flood scenario, in addition to the dedicated study survey, valuable insights can be obtained from broader national surveys, including those focused on health, consumer satisfaction, and social attitudes.

Official statistics

Official statistics are data collected and published by government agencies to inform the public and aid in decision-making. These datasets encompass various subjects, such as demographics, economics, and social indicators.

Comparable in quality and accuracy to administrative data, some datasets provide aggregated outcomes like unemployment rates, crop yields, and educational attainment at the provincial or state level. Conversely, Census data delve into individual-level information but for a limited portion (5%, 10%) of the population.

Other types of data

These are commonly used data types and sources frequently employed in the economic analysis of disaster effects. Additionally, there are various other data types, including experimental data gathered through controlled experiments to test hypotheses and establish causal relationships.

Big data involves large and intricate datasets that surpass the capabilities of traditional data processing methods, incorporating sources like social media data, sensor data, and extensive transaction records. Qualitative data offers non-numerical information that captures the richness and depth of experiences, opinions, and behaviours, often obtained through methods such as in-depth interviews.

Remote sensing data involves collecting information about the Earth’s surface from a distance, typically using satellite or aerial platforms.

Reflect and share

Check out this insightful Forbes article (2023) that underscores the significance of data in reducing the impacts of natural disasters. The piece explores how data is applied before, during, and after such events, and it also delves into the challenges associated with making the most of data.

Reflect on the article’s reference to innovative strategies for addressing data collection challenges during a natural disaster.

What particular solutions or technologies do you think could be effective in this scenario?

Share your thoughts in comments section.

References

_{White, Jeff (2023) ‘How data insights can save lives when natural disasters strike’ Forbes Technology Council, accessed on 24 November 2023}