RDP 9209: Financial Liberalisation and Consumption Behaviour 6. Time-Varying Parameter Estimates

To further explore the value of λ over time, the instrumental variables equation (15) is estimated as a rolling regression. The initial sample period is chosen to reproduce the value of λ for the 1960s reported in Table 1. Its value is then allowed to vary by adding an observation, while keeping the total number of data points in each regression unchanged (ie dropping the earliest observation from the previous regression).^[9] Only the countries in which financial liberalisation was thought to be important over the sample period are considered here, including the United Kingdom, which was excluded for technical reasons from the SURE procedure. The results, shown in Chart 1, permit further interpretation of the earlier findings.

What is particularly striking is the apparent correlation of changes in λ with factors other than financial liberalisation which influenced liquidity constraints prior to the 1980s. For all five countries, the excess sensitivity parameter declines in the late 1960s and/or the early 1970s (though much less so in the case of Australia). This corresponds with the easing of monetary policy at the time, which saw liquidity in the form of money balances expand rapidly. In the case of the United Kingdom, there was also a credit explosion in the wake of the introduction of Competition and Credit Control in 1971. The ready availability of money balances reduced liquidity constraints independently of the degree of financial regulation. However, the first oil shock in 1973 and 1974, and the firming of monetary policy at the same time, appears to have reversed these developments. The well-known transmission of a world-wide downturn in activity, in the presence of regulated capital markets, saw a marked rise in the dependence of consumption on current income – a phenomenon probably associated with increased precautionary saving behaviour.

These common patterns in response to shocks, it is worth noting, are consistent with the rationale given for the importance attached to the pooled (SURE) results presented in Section 5.

During the second half of the 1970s and/or throughout the 1980s, financial liberalisation became much more widespread in all of the countries considered here (Table 5). Moreover, following the inflation and income shocks of the 1970s, financial innovations to avoid existing regulations had in any case become more widespread. In the case of the United States, the Volker disinflation from 1979 to 1981 was associated with some rise in the excess sensitivity parameter. But following major deregulatory moves in the early 1980s, the sensitivity of consumption to current income appears to have moved into a phase of a sustained decline, despite major changes to the stance of monetary policy, the stock market crash and other shocks. In the cases of Japan, Australia and Canada there are also sustained declines from either the late 1970s or early 1980s, in spite of major nominal and real shocks during the 1980s.

Only the case of the United Kingdom presents something of a puzzle. From the mid 1970s to the mid 1980s, there is a sustained decline in the excess sensitivity parameter, in much the same way as for the other countries in this group. However, from about 1987 there is a marked reversal of this trend, a phenomenon which probably explains why the UK was rejected for pooling in Section 5. The reasons for this pattern are unclear, but one possibility is that financial institutions themselves may impose liquidity constraints. Throughout the second half of the 1980s, there was a remarkable build-up of debt in relation to net worth within the UK household sector.^[10] Asymmetric information problems in these circumstances, and in the absence of official regulations, may lead to equilibrium credit rationing by financial institutions.^[11] Alternatively, after a period of excessive borrowing following financial liberalisation, a debt overhang may have generated more conservative attitudes on the part of UK households.

This possibility underlines a more general qualification to the finding that excess sensitivity of consumption to income declined in the late 1970s and throughout the 1980s in a number of countries that liberalised their financial markets. Financial liberalisation in the presence of pre-existing excess demand for credit by households could be associated with once-for-all portfolio re-adjustments and period-specific apparent declines in the excess sensitivity parameter which might later be reversed. While this must be considered a very real possibility, it is nevertheless the case that countries in this group other than the United Kingdom all show evidence of increased consumption smoothing, in spite of widely differing experiences with regard to household indebtedness and shocks to real income experienced in the 1980s. Thus, for example, Australian households were relatively conservative in their borrowing through the 1980s (Callen 1991), and still show strong evidence of greater consumption smoothing in the face of major adverse movements in the terms of trade in the middle of the 1980s. It is possible, therefore, that the UK household sectors' excessive use of credit markets may itself be a period specific and a relatively unique phenomenon.

Footnotes

Note that this does not necessarily reproduce the Table 1 1980s λ estimates as the last observation of λ in the graph. This is because the length of the “1960s” sample period differs (depending on data availability) for each country. [9]

UK experience over this period is summarised by Franklin et al (1989). [10]

The credit rationing proposition resulting from asymmetric information is most clearly exposited in Stiglitz and Weiss (1981). Bernanke and Gertler (1989) present a model in which cyclical variations in the level of economic activity are amplified via the effects of agency costs on the price of external funding. The greater the level of corporate net worth the lower are agency costs. But net worth generally varies procyclically, aggravating deadweight agency costs, reducing investment and magnifying the extent of the downturn in activity. This effect reverses itself for an upturn in activity. Similar models have been presented by Greenwald and Stiglitz (1989), Greenwald and Stiglitz (1990) and Williamson (1987). [11]