Non-technical summary
Decision-making in monetary policy is based on a large amount of information arriving at different frequencies. Given that many important economic variables are released with considerable time lags at low frequencies, policymakers often face the problem of assessing the current state of the economy with incomplete information.
In New Zealand, the key labour market indicators, such as the unemployment rate, employment growth and labour force participation rate, are all published at a quarterly frequency and are published with some delay.
In this paper, we use the so-called MIDAS (mixed-data sampling) approach to incorporate mixed-frequency data to “nowcast” the current state of the labour market, based on monthly indicators, taking into account publication lags.
By taking into account the persistence and lags of the indicators, MIDAS builds a complicated dynamic relationship between the indicators and the labour market variables we investigate. The main purpose of the current study is to demonstrate the power of MIDAS in a prototypical model and its potential in building more comprehensive forecast models of labour market variables.
We show that better nowcasts of the current state of the labour market are obtained by using monthly data on dwelling consents, motor vehicle registrations, international migration and business confidence data, compared to first order autoregressive and time-averaging benchmarks.
The improvement is more dramatic in the case of forecast combinations. We also show that most of the improvement in forecast accuracy is obtained from the data available in the first month of the quarter. These results suggest that taking care of the mixed-frequency data with MIDAS improves our assessment of the current state of labour markets in New Zealand.