Forecasting religious change: a Bayesian model predicting proportional Christian change in New Zealand

Abstract

For over a century, sociologists of religion have been describing declining trends in religious affiliation across most industrialized countries, a trend that Max Weber characterized as “the disenchantment of the world.” Although secularization is a matter of ongoing debate, there is relatively little predictive modeling. Using New Zealand census data from 1966–2006, we develop a Bayesian predictive model to forecast the proportion of the population identifying as Christian in the future. A surprisingly simple linear model revealed that Christian affiliation in New Zealand has been steadily declining at a constant rate of around 0.90% per year since 1966. The model explained 97.4% of the variance in proportional change in the New Zealand Christian population over the past half-century. The model also offers testable predictions. We estimate that 46.1% of the population will identify as Christian in New Zealand in 2015, with a further decline to 41.7% by 2020. The 95% credible interval for the 2015 estimate is between 42.5% and 49.6%. The 95% credible interval for the 2020 estimate is between 37.7% and 45.8%. We consider both the strengths of this Bayesian crystal ball and its potential limitations.

Keywords:

• Christianity
• religious change
• Bayesian models
• forecasting
• New Zealand

Acknowledgements

The Canadian Social Sciences and Humanities Research Council (SSHRC): ‘The Evolution of Religion and Morality’ supported this study.

Notes

1. An annotated version of the Mplus syntax and data that we used to construct and test this model are available online at: http://www.psych.auckland.ac.nz/uoa/chris-sibley-supplementary-files. The model can be easily adapted to forecast predicted values for other years.