3.3 Historical decomposition
ECM (A) and (B) can both be used to describe the behaviour of consumption over the 1990:1 to 2002:1 period. Figures 1 and 2 above showed that both models in general perform well in explaining fluctuations in consumption spending over recent history. However, there is a two-year period from the beginning of 1998 where the models do not match the data as closely as earlier in the period. The graphs in Figure 4 below show the contribution of each independent variable in the models to per capita consumption growth. The graph highlights the key drivers of consumption in ECM (A) and (B) over time. One drawback of using the models to explain consumption over history is that the models may not include all relevant variables.
3.3.1 1990-1993
Per capita consumption growth was weak in the early 1990s with a number of periods of negative quarterly growth, before picking up in 1993. Both models fitted the weak consumption growth up to 1992 well, but not the turning point in mid-1992. Poor income and non-financial wealth growth appears to be key drivers of weakness in consumption growth in the early 1990s. Changes in wealth had a negative impact on consumption in both models until early 1993. Both models also saw negative contributions from changes in income until 1992, reflecting weak labour market conditions and welfare reforms during this time period. The error correction terms were contributing to the negative consumption growth over the early part of the 1990s, suggesting that consumption was above its long-run equilibrium level during that time.
Changes in the unemployment rate initially had a positive impact on consumption. However, both models suggested that the increasing unemployment rate of the early 1990s had a negative impact on consumption from 1991 to early 1993. The impact of mortgage equity withdrawal is minimal.
3.3.2 1994-1997
Per capita consumption grew solidly over much of the 1994-1997 period, notwithstanding two or three quarters of declines. Changes in non-financial wealth made a positive contribution to consumption in both models over most quarters until 1997, with some particularly large impacts in 1994. Changes in income growth, boosted by solid employment growth over the 1994-1997 period, made a positive contribution in every period up until the middle of 1997, except for two periods in 1995. Labour market performance, and therefore income growth, was weaker from mid-1997 through 1998, when the economy was affected by the ‘Asian crisis’ and two summer droughts.
Both ECMs suggest a substantial positive impact from migrant transfers over 1994-1996, reflecting a build up in migrant transfers over this period as net migration increased. The sharp turnaround in net migration and migrant transfers from around 1997 appears to have had a negative impact on consumption from 1997 right through until mid-1999. There were a number of periods of positive and negative contributions to consumption from changes in the unemployment rate between 1994 and 1998. The error correction term generally had a negative impact on consumption in both models over the 1994-1997 period, indicating that consumption was running in excess of its long-run equilibrium level.
3.3.3 1998-2002
Per capita consumption grew solidly through 1998 and 1999 before slowing in 2000 and picking up again during 2001. Both models performed poorly over 1998 and 1999, when the fitted values were expecting weak consumption growth but actual consumption growth remained strong. The non-financial wealth variable appeared to be a key driver of the poor performance of the models with a large negative wealth effect predicted over the period, due to falling house prices. Alternatively, the models may not have given a large enough role to income growth over this period.
Both models showed substantial negative wealth effects right through from 1998 until 2001. The models expected a solid impact on consumption from changes in income, with a particularly large impact from 2001. This observation appears consistent with the strength in the labour market in late 2000/2001, including solid growth in wages in 1999 and in employment from 1999 onwards.
The error correction term had a positive impact in both models, largely due to the low predicted consumption in 1998. The turnaround in net migration and consequent pick up in migrant transfers in 2000/2001 appears to have had a positive impact on consumption.
