Demographic Projections: User and Producer Experiences of Adopting a Stochastic Approach

Abstract

Statistics New Zealand is one of the few national statistical agencies to have applied a stochastic (probabilistic) approach to official demographic projections. This article discusses the experience and benefits of adopting this new approach, including the perspective of a key user of projections, the New Zealand Treasury. Our experience is that the change is less difficult to make than might be expected. Uncertainty in the different projection inputs (components) can be modelled simply or with more complexity, and progressively applied to different projection types. This means that not all the different demographic projections an agency produces need to adopt a stochastic approach simultaneously. At the same time, users of the projections are keen to better understand the relative certainty and uncertainty of projected outcomes, given the important uses of projections.

Introduction

“…In this connection, it is also worth to report the user and producer experiences of Dunstan and Ball (2016) of Statistics New Zealand after they had implemented a probabilistic approach in 2012; see also Dunstan (2019). They stress that the change from a deterministic to a probabilistic forecast was less difficult to make than one might expect.…”

“…Collectively, those different scenarios can give some impression of the degree of uncertainty, but not in any quantified way. The probability that an outcome will be within a certain range is unknown (Dunstan and Ball 2016). We do not know if chances are 30, 60, or 90% that Norway in 2060 will have between 5.8 and 7.8 million inhabitants.…”

“…Indeed, the statistical agencies of some countries have started to publish their forecasts in the form of probability distributions, following common practice in, for example, meteorology and economics. A key use of probabilistic demographic forecasts is in modelling the long-term fiscal impact of an ageing population by policy agencies (Tuljapurkar 1992; Lee and Tuljapurkar 2000;Alho et al 2008;Dunstan and Ball 2016).…”

Probabilistic demographic forecasts

“…Collectively, those different projections can give some impression of the degree of uncertainty, but not in any quantified way. The probability that an outcome will be above or below a given scenario is unknown (Dunstan and Ball 2016).…”

Subnational population forecasts: Do users want to know about uncertainty?

“…Unlike other disciplines, notably meteorology (e.g., Joslyn and Savelli 2010; Morss, Demuth, and Lazo 2008;World Meteorological Organisation 2008), there has been relatively little engagement with end users of population forecasts to ascertain their views on measures of uncertainty and how they interpret them (but see Dunstan and Ball 2016;UNECE 2018). Answers to these questions should prove informative in guiding the development of approaches to communicating population forecast uncertainty and the extent to which effort needs to be put into persuading users of the value of uncertainty measures.…”