With the growing availability of data and improved modelling techniques the frontier of policy advice continues to move out. To illustrate the potential of new data and techniques, Treasury staff presented a small selection of the empirical work taking place in the organisation at the recent annual New Zealand Association of Economists (NZAE) conference. Over the course of nine presentations staff covered a range of topics including measuring poverty, wealth inequality, firms’ dynamic capabilities, and housing affordability.
Several presentations made use of the Treasury’s microsimulation model of the New Zealand personal tax and transfer system – known as the Tax and Welfare Analysis (TAWA) model.
Modelling and measuring poverty
A main purpose of the TAWA model is to project future rates of child poverty. This involves modelling different poverty measures, including those based on a level of income that is fixed over time, and those based on a poverty threshold that moves as incomes grow. Measures may also vary depending on whether they are based on income or on material deprivation, and whether they account for housing costs or not.
Yet as Meghan Stephens showed in her presentation, different measures of poverty can suggest different trends. There is a group of people who consistently show up in all the main poverty measures – but in general the overlap between measures is small. This points to the importance of understanding the strengths and weaknesses of different modelling approaches and using the measure best suited to the question you are aiming to answer.
This conclusion was reinforced by Yvonne Wang in her paper on the sensitivity of child poverty projections to key forecast variables. Many of Yvonne’s results were intuitive: for example, we can expect child poverty rates to decrease when main benefits increase in generosity. Yet some results illustrated how complex modelling poverty can be. For instance, a uniform increase in the growth rate of wages may increase measured poverty based on a moving threshold. This reflects how the increase in wages increases median income (which the poverty threshold is based on) not just the incomes of the poorest.
And, of course, any transfers designed to reduce poverty can influence the incentives to work that people face. To illustrate, Michael Eglinton presented effective marginal tax rate profiles and budget constraints for a range of scenario families. These scenarios were generated by the Treasury’s Income Explorer Tool, which is a tool built in the programming language R and which the team plans to share with other researchers.
Better understanding our data
These modelling improvements go hand-in-hand with the improved access to and use of data, particularly linked administrative and survey data. For instance, as Cory Davis showed in his presentation, the Treasury has improved the way the TAWA model estimates take-up of the Accommodation Supplement by drawing on administrative data on the receipt of welfare programmes contained in Stats NZ’s Integrated Data Infrastructure (IDI).
Further, COVID-19 has sparked interest in collecting more timely economic data, and in 2020 Stats NZ organised an experimental release of weekly data on the number of jobs filled in New Zealand. In his presentation, Robert Templeton discussed how these new, more timely data can provide a better understanding of trends in the number of filled jobs in New Zealand. He highlighted challenges too – particularly when it comes to making seasonal adjustments – which reinforces the need to use new data with care.
Tod Wright discussed similar challenges when comparing Household Economic Survey (HES) data to that drawn from IR3 tax returns. IR3 data is potentially an important source of data on self-employment income, overseas income, and interest and trust income. We know the HES significantly undercounts some income components, but Tod’s work with the IR3 data has shown that these data have limitations too.
This reinforces the importance of the work by Ben Ching and Oscar Parkyn, who presented a new method for estimating the distribution of individual wealth in New Zealand. They explored an experimental method that infers the wealth distribution from the distribution of capital income by combining Inland Revenue administrative data with macroeconomic data on the aggregate household balance sheet. Their preliminary results suggested there is more wealth at the top of the distribution than estimated by HES, which is consistent with international evidence that suggests that household surveys are subject to limitations in measuring the very top of the wealth distribution.
New data and techniques are providing insights in other areas too. In his paper, Tim Ng presented his work with data on business practices and attitudes regarding innovation and change to estimate measures of dynamic capabilities. This work, which won the conference’s 2021 David Teece Prize in Industrial Organisation and Firm Behaviour, looks at how entrepreneurial businesses can succeed in innovating in competitive, volatile and uncertain markets. The research is part of a larger project in which the measures will be tested for how well they explain variation in business performance, and whether dynamic capabilities play a role in helping business adapt and respond to emerging opportunities, as well as adverse shocks to their business environment.
Chris Parker presented a theoretical model of uncompetitive urban land markets that may help to explain why housing in New Zealand is so overpriced. Chris argued that the assumptions underlying standard economic models used to analyse housing markets do not apply in New Zealand. If we build a model on more realistic assumptions, we can see barriers to fringe land market competition contribute to upward pressure on house prices. Chris will soon publish a post on the Treasury blog that will explain the model in more detail and provide an overview of what underlies our house prices.
Treasury thanks the NZAE for organising its annual conference in economics – and the opportunity to share some of our work, as well as learn from others’ research.