RDP 2003-01: Business Surveys and Economic Activity 6. Turning Points

One of the difficulties in using tools such as correlation analysis, the autoregressive model and Granger causality tests, is that they do not necessarily tell us anything about the accuracy of business surveys in predicting turning points in the economic cycle. This is particularly important for some of the more aggregated economic activity indicators such as those relating to business conditions (including sales/output), employment and selling prices. A simple test of the ability of business surveys to pick turning points is to examine visually whether a turning point in an economic variable is coincident with, or has been led by, a sharp movement in a business survey indicator. Where a significant movement in a business survey indicator has routinely led a turning point, the information is particularly useful. Even in cases where the business survey indicator is a coincident indicator, the information is beneficial because of the timeliness of some of the business survey indicators. (As noted earlier, this is particularly the case for GDP, which can lag the surveys by up to two months.) Nevertheless, any assessments are highly judgmental because of the quarterly volatility in both the survey and statistical measures.

6.1 Business Conditions

The various business survey indicators of business conditions are plotted against quarterly growth in GDP in Figure 2.^[13] The positioning of the axes is intentional, as a long-run average reading of zero for the business surveys has tended to equate with trend quarterly growth in GDP of just under 1 per cent.

The ACCI-Westpac is the only surviving business survey from the late 1985/86 slowdown period. In terms of tracking overall activity, the ACCI-Westpac survey measure was, at best, a coincident indicator as the economy slowed and probably lagged the recovery by a quarter. This is somewhat surprising as the slowdown in output of the manufacturing sector (to which the survey applies) began a quarter earlier than for the overall economy and recovered a quarter earlier as well.

The three surveys reporting around the time of the early 1990s recession were coincident indicators of the slowdown. In a couple of instances (Dun & Bradstreet and ACCI-Westpac), sentiment was turning down by the second half of 1989, but this was from well above long-run average levels. The Dun & Bradstreet ‘expected output’ series (not shown on Figure 2) showed a clear one quarter lead of the comparable actual series through this period, though this leading relationship appears to have diminished over time. The ACCI-Westpac survey also provided some advance warning of the depth of the recession that was to come, though this probably reflected the focus of this survey on the manufacturing sector – manufacturing output fell by just under 4 per cent in the year to June 1990, whereas GDP rose by nearly 2 per cent over the same period. Most surveys picked up the improvement in sentiment through 1991, though the indices were still well below long-run average levels suggesting that growth in sales/output was still weak.

All surveys recorded a strong rebound in conditions from well below long-run average levels to well above long-run average levels over the three years from early 1991 to early 1994. However, the surveys (with the possible exception of NAB) appear to have provided a false signal in 1995 when they turned negative but growth in output remained around trend.

Another possibly false signal was around the time of the Asian financial crisis when there was a marked deterioration in business sentiment over what, in the event, turned out to be a period of above-trend growth. The fall in sentiment was particularly noticeable for those firms exposed to the external sector, such as some parts of the manufacturing sector. Sharp falls in business conditions to below long-run average levels were recorded in the ACCI-Westpac and AIG-PWC surveys of manufacturers. In the event, however, growth in manufacturing output slowed only modestly to around 3½ per cent (this compares with growth in GDP (excluding manufacturing) of around 4½ per cent).

Over the course of 2000, the survey indices of business conditions declined, suggesting some slowing in the rate of growth of the economy, albeit back towards rates of growth experienced in the mid 1990s. The fall, however, was from high levels, possibly reflecting a simple reversal of the ebullient expectations built up over the course of 1999, and it wasn't until the December quarter that most of the indices fell to below long-run average levels. The exceptions were the actual sales index in the Yellow Pages survey of small business, which fell below its long-run average in the June quarter 2000, and the AIG-PWC survey of manufacturing which dipped briefly below its long-run average level in March quarter 2000 and again in September quarter 2000. The recovery in sentiment appears to have lagged growth in output. In fact, much of the rebound in business conditions coincided with the release of stronger-than-expected GDP data for March 2001 quarter in June.

6.2 Employment

The survey measures of expected employment in the next quarter are plotted against the quarterly percentage change in the ABS measure of employment in Figure 3. The survey measures are plotted with respect to the quarter in which the observation was made, rather than the quarter to which the observation refers. The employment indicators have probably done better than the business conditions indicators in picking turning points. As noted earlier, this may reflect firms knowing more about their own prospective hiring intentions, since they are the ones taking the decisions, than they do about future sales. It may also reflect the fact that changes in employment tend to lag changes in output. Accordingly, changes in employment are easier to predict.

The employment shakeout in the early 1990s recession was well telegraphed by most of the surveys, with negative net balances being recorded by most of the surveys in the second half of 1989. The return to consistently positive growth was also matched by a strong recovery in hiring intentions. Most of the surveys appear to have anticipated the acceleration in employment growth that occurred in late 1997 to early 1998 and again towards the end of 1999, with net balances for most of the surveys moving higher in the middle of 1997 and 1999 respectively. In contrast, the weakness in the job market in late 2000 seems to have come as a surprise. Employment was very weak in December quarter 2000. However, this only coincided with weakness in business expectations for March quarter 2001.

6.3 Selling Prices

The survey measures for selling prices also appear to be picking up some price pressures in advance of the official statistics. Figure 4 shows the various business survey measures of selling prices plotted against weighted median inflation (excluding the effect of tax changes around the time of the introduction of the GST in September quarter 2000). The two periods of a noticeable acceleration in price pressures in the inflation-targeting period – early 1995 and in the second half of 2000 – have been picked up in virtually all of the surveys. Strength in output growth over these periods would have been a factor for survey respondents, but it may also reflect survey respondents picking up some of the movement in the exchange rate, which eventually flows through to consumer prices – large depreciations were experienced in 1992/93 and over the second half of 1997 and 1998.

Footnote

A further note of caution is warranted as the ABS series are based on the latest available data rather than the data that were available at the time. [13]