In the past 12 months clients have increasingly been asking us for real time data. This is understandable. Given the speed with which the COVID-19 pandemic and its economic repercussions unfolded, having up-to-the-minute information about how your region or sector is performing means you are better able to respond. Here, I look at the ways that real time data is used in economic analysis and the trade-offs that need to be considered when using real time data.

Technology has been the key enabler of real time data. Developments in sensor technology, telecommunications and IT networks, and social media platforms have all combined to create a world where data about almost every facet of our lives can be collected and transmitted as quickly as it takes us to click ‘Accept all cookies’.

Data collected from sensors and other infrastructure embedded throughout an electricity grid and channelled back to analysts via a telecommunications network enables electricity distributors to identify maintenance issues prior to breakdowns and better understand consumer demand.

In healthcare, wearable sensors and other healthcare equipment such as glucometers, connected scales, and blood pressure monitors keep tabs on a patient’s vitals and essential body functions enabling clinicians to respond to any changes in what can be literally a matter of life-or-death situation.

Social media platforms are creating tools to be able to process the huge volume of data created quickly and efficiently to better target advertising or be able to act on fake news or social media bullying as quickly as possible.
In finance, understanding and acting on information in real-time is vital on the stock exchange trading floor. Banks also use real time data to monitor fraudulent credit card transactions and detect warning signals in extremely early stages of where clients may go into default.

In contrast, ‘official’ datasets such as the Linked Employee Employer Dataset (LEED) produced by Stats NZ are increasingly seen as out of touch. Infometrics uses LEED for our historical employment estimates. LEED has a robust methodology that has remained consistent over time, it offers a 20 year time series, and very detailed industry and regional employment estimates. LEED is also reliable because it is based on a full census of businesses rather than a sample survey.

However, due to the time that it takes Stats NZ to collect and verify LEED, our latest employment estimates are for the year to March 2020. We will have to wait until January 2022 for LEED to show us the initial effects of COVID-19 on New Zealand employment.

Economists have been using real time data to get a more timely understanding of economic activity. For example, in our Quarterly Economic Monitor Infometrics uses traffic flows as an indicator of economic activity, electronic card transactions as an indicator of consumer spending, health enrolments as an indicator of population change, and Jobseeker Support Recipients as an indicator of unemployment. Stats NZ have even come to the party, publishing ‘experimental’ monthly employment data. Infometrics uses these datasets because they are available at the territorial authority level and are available with a time lag of just a few weeks.

Infometrics also uses a process called ‘nowcasting’ to provisionally estimate economic activity (Gross Domestic Product) for territorial authorities in our Quarterly Economic Monitor. Nowcasting in economics is the prediction of the present. The methodologies used to nowcast are very similar to traditional econometric forecasting methodologies. The drawback of nowcasting, of course, is that it produces an estimate of the present rather than a real time measure and is only as good as the relevance of the econometric methodology.

Real time datasets have their drawbacks. They are usually collected for purposes other than economic analysis. Their collection methodology may change over time or may not be transparent, and the data might not be adjusted for seasonal variations.

Infometrics examines real time datasets thoroughly before we include them in our dashboards. Our highest priority is ensuring that the data accurately reflect the economic phenomena we expect them to. Trust is important. Often, we will rely on several datasets that measure similar things. If we see consistent results across different datasets, that gives us confidence that we are interpreting the data correctly.

Above all, we always consider how the data is collected and the reason for which it is collected because this can influence the results. Jobseeker Support recipients is a good example. The data is collected to determine who is eligible for benefit payments not for labour market analysis. The results are therefore influenced by changes in benefit eligibility rules.

With data, there is usually a trade-off between quality and timeliness. We use LEED because it offers us high quality data. We use real time data for its timeliness, but our interpretation is always with a grain of salt.