RDP 2003-01: Business Surveys and Economic Activity 5. Results

The business survey indicators that are examined are confidence, business conditions (sales/output), profitability, capital expenditure, inventories, exports, employment, input costs and selling prices. Correlation coefficients and results from the autoregressive model and Granger causality tests are discussed below.

5.1 Confidence

Of all the concepts that business surveys attempt to measure, ‘confidence’ is perhaps the least well defined. At a theoretical level, confidence does not play a major role in the analysis of economic behaviour, and is perhaps akin to ‘animal spirits’ in the business cycle literature. Three of the six surveys cited above have a forward-looking question relating to confidence, though the period that is being referred to varies from the next quarter (NAB) to the next year (Yellow Pages). On the whole, the responses to the confidence question are considerably more volatile than for responses relating to more measurable concepts, such as sales and output, and may reflect the sensitivity of confidence indicators to news items.

The correlation coefficients listed in Table 3a suggest that confidence, as measured in both the NAB and ACCI-Westpac surveys, is fairly closely correlated with year-ended growth in private final demand and the broader-based measure of industry sales – correlations with year-ended growth in GDP are smaller but still positive. Correlation coefficients over the shorter common sample period (Table 3b) are lower than over the full sample periods, suggesting that the confidence net balance statistic for the longer-running NAB and ACCI-Westpac surveys has picked up the very large swings in output (such as that associated with the early 1990s recession), though they have not coincided as well with the smaller cyclical swings that have occurred over the past decade. Correlations for the confidence indicator in the Yellow Pages survey are low, though this could relate to the focus of the Yellow Pages survey on small enterprises for most of the period (the survey was expanded to include medium-sized enterprises in November 2000). This survey also asks for a view on business conditions 12 months ahead, rather than the shorter horizons in the other surveys.

Tables 3c and 3d detail the results based on the simple autoregressive model. This tests whether survey measures of confidence explain, at a statistically significant level, private final demand in the period in which the observation was taken, or private final demand in the period to which the survey question specifically refers. For both periods, all survey measures of confidence are significant. The Granger causality tests suggest that there is some information in the business confidence indicators over and above lags of private final demand for the ACCI-Westpac and NAB surveys (Table 3e). Nevertheless, the results also suggest that firms, in formulating an opinion, may be basing that view at least partly on an extrapolation of past trends, such as lags of private final demand (this, of course, may be a reasonable thing to do).

5.2 Business Conditions

The term ‘business conditions’, as used in this study, refers to the survey questions relating to the more tangible concepts of output and sales, as well as the less tangible concepts such as business and trading conditions.^[9] These aspects of the business surveys receive considerable attention in economic commentaries as, in theory, they most closely approximate aggregate economic measures of demand, such as GDP and industry sales.

Survey respondents are asked to report in both ‘actual’ and ‘expected’ terms. On the whole, the actual series have slightly higher correlation coefficients than the expected series (Tables 4a and 4b). The relative similarity of the correlation results reflects the fact that actual and expected conditions are themselves quite highly correlated (in the case of the NAB and Dun & Bradstreet surveys, for example, the correlation coefficients between actual and expected business conditions are 0.96 and 0.86 respectively). At a practical level, it means that it does not make a lot of difference whether the actual or expected series is used for analytical purposes.

The correlation coefficients suggest that the strongest relationship is between the survey indicators and annual growth in a broad-based measure of industry sales. The correlations with industry value added, which is approximated by industry sales less intermediate consumption, are weaker. A strong positive correlation also exists between the survey indicators and annual growth in private final demand, and to a lesser extent, GDP. The correlations are much higher for year-ended growth than for comparisons over shorter periods, though this probably reflects the higher volatility in the more frequent series. Again, correlations are not as strong over the shorter, common sample period, which does not include the early 1990s recession.

A couple of points are worth noting where comparable data are available at the industry level. Firstly, the observation that the correlation with the broad-based measure of industry sales (such as the manufacturing, retail and wholesale grouping) tends to be higher than with the industry-specific measure of sales, suggests that respondents, at times, might be replying according to the economic conditions of the overall economy rather than to conditions in their own industry. In contrast, the negative correlation coefficients for the mining industry indicate that the output of this sector is not strongly correlated with the domestic economic cycle.

Most of the survey measures of actual business conditions are significant at the 10 per cent level in the autoregressive model of GDP, though fewer surveys meet this hurdle when the survey measures of expected business conditions are employed (Tables 4c and 4d). The Granger causality tests indicate that only in the case of the ACCI-Westpac and Dun & Bradstreet surveys do business conditions Granger-cause GDP while the reverse causality does not apply (Table 4e). For the other surveys, Granger causality is not established in either direction.

5.3 Profitability

Respondents to business surveys are also asked about profitability. Tables 5a and 5b detail correlation coefficients between the survey measure of profitability and two measures of profitability produced by the ABS. The first measure is the national accounts measure of corporate gross operating surplus (GOS). The second measure is company profits before income tax, net interest and depreciation (with a breakdown also provided by industry).^[10] While the latter measure is used to compile the national accounts measure of corporate GOS, there are important differences between the two series. Unlike the company profits measure, the national accounts series for corporate GOS includes profits of companies employing less than 20 people, as well as profits of the agriculture and community service industries. More importantly, the national accounts measure of corporate GOS makes an adjustment for changes in profits attributed to a revaluation of inventories (known as the inventory valuation adjustment) and excludes other valuation adjustments such as those relating to foreign exchange gains or losses and revaluation of assets.

The most notable feature of the results is that, on the whole, the business survey measures of profitability are more closely aligned with company profits than with the national accounts measure of corporate GOS over the common sample period. Thus respondents appear to be responding to surveys in terms of profitability that includes valuation gains and losses. Quarterly correlations are also reasonably positive, notwithstanding the greater volatility in these data compared with year-ended calculations. In terms of the company profits measure, there is not much difference between actual profitability in the current quarter and expected profitability for the next quarter. On the whole, the deterioration in correlation coefficients in the post 1990s recession period is much smaller than for other survey questions. There is a reasonably strong correlation between profits by relevant industry and the manufacturing-based surveys, and the manufacturing and retail components of the NAB survey.

Consistent with the generally positive correlations of the survey-based measures of profitability, the Dun & Bradstreet and NAB survey measures of actual profitability are significant at the 1 per cent level in the autoregressive model of corporate GOS (Table 5c). For expected profitability, the ACCI-Westpac and Dun & Bradstreet survey measures are significant at the 1 per cent level (Table 5d). The Granger causality tests suggest that a number of the business surveys (ACCI, ACCI-Westpac, Dun & Bradstreet and NAB) are forward-looking with survey results Granger-causing company profits (Table 5e).

5.4 Capital Expenditure

Business capital expenditure is a relatively volatile component of the expenditure side of the national accounts. Therefore, business surveys are potentially quite useful in forecasting actual outcomes. Some business surveys examine total business capital expenditure, whereas others divide it into capital expenditure on machinery and equipment (around 50 per cent of business investment as recorded in the national accounts), and capital expenditure on buildings and structures (around 25 per cent of total business investment).^[11] The surveys provide actual and expected readings for capital expenditure, with the exception of the ACCI-Westpac and NAB surveys which only provide expected data.

Tables 6a and 6b detail correlation coefficients for the various business survey measures of capital expenditure with the national accounts measure of business investment (where applicable, business investment is split into machinery and equipment investment, and buildings and structures investment), as well as the ABS capital expenditure (Capex) survey measure of business investment (which also provides a breakdown by industry). The machinery and equipment component of the latter survey feeds into the national accounts measure of business investment in machinery and equipment. The national accounts measure of buildings and structures investment is derived from surveys other than Capex (such as the ABS Building Activity Survey and the Engineering Construction Survey).

The results for capital expenditure are mixed, though a number of points stand out. Firstly, survey measures of capital expenditure are more closely aligned with the national accounts measures of investment than the Capex measure over the common sample period, though they don't appear to offer much assistance in forecasting quarterly movements in investment. Secondly, the correlation coefficients for the expectations components of the business surveys are reasonably high, perhaps reflecting the long gestation periods normally associated with capital expenditure plans. More importantly, however, those surveys that look at investment plans over the year ahead (ACCI-Westpac and NAB) appear to be providing some additional information as the survey responses are significant in the autoregressive model (Table 6d) and are forward-looking (Table 6e).

At the industry level, the AIG-PWC survey has a relatively high positive correlation with business investment by manufacturers, while the Dun & Bradstreet survey has a similar relationship with the industries that it covers, viz manufacturing, retail and wholesale. The NAB survey has also performed reasonably well in tracking expenditure plans of the recreation, mining and manufacturing industries. Outside of these industries, however, the correlations are low, and in a number of cases, even negative. The low correlations for the Yellow Pages survey probably reflect the coverage of the survey – small and medium-sized enterprises. Moreover, some capital items are more likely to be treated as intermediate consumption, rather than as business investment (such as computers).

5.5 Inventories

Four of the surveys ask businesses about the level of inventories. The Dun & Bradstreet and NAB surveys ask respondents about changes in total inventories (i.e., both raw materials and finished goods); the ACCI-Westpac survey seeks separate responses for stocks of raw materials and stocks of finished products; and the AIG-PWC survey asks manufacturers about stocks of finished products only.

Tables 7a and 7b detail correlation coefficients between the various survey measures and inventories as measured in the national accounts and in the ABS Quarterly Economic Activity Survey (QEAS).^[12] Both the national accounts and QEAS survey measures of inventories include stocks of final goods, work-in-progress and raw materials (and therefore should be tracked more closely by the Dun & Bradstreet and NAB surveys). The QEAS survey forms the basis of the national accounts estimate of the change in inventories (and also provides a breakdown by industry), though it excludes inventories of the transport and construction industries.

Over the full sample period, the correlation coefficients for both the quarterly and year-ended measures of inventories are not that dissimilar. This is also true over the shorter, common sample period, though it is noticeable that the correlations are less positive. On the whole, the Dun & Bradstreet survey is providing the best guide to inventories, with this survey measure significant in the autoregressive model (Tables 7c and 7d) and forward-looking as measured in the Granger causality test (Table 7e). This may reflect the fact that the industries covered by the Dun & Bradstreet survey – manufacturing, wholesale and retail – account for a large share of total inventories.

5.6 Exports

Four of the business surveys report on exports – the one small difference amongst the surveys being that the ACCI-Westpac reports on export deliveries, whereas the others report on export sales. Correlation coefficients between the survey measures and total export volumes are shown in Tables 8a and 8b. Export volumes are also split into non-rural export volumes, resource exports and manufactured exports.

Correlations over the common sample period are strongest between the survey measures and year-ended growth in manufactured exports. In contrast to many of the other survey questions, however, the performance of the surveys following the early 1990s recession has improved. Results from the autoregressive model (Tables 8c and 8d) and Granger causality tests (Table 8e) were generally inconclusive.

5.7 Employment

All surveys have a question on employment. Two of the surveys (ACCI-Westpac and Dun & Bradstreet) refer specifically to employee numbers, while the remainder refer to ‘employment’ (with respondents possibly referring to either numbers or hours worked). Tables 9a and 9b list correlation coefficients between the survey measures of employment and the ABS measure of the number of employed and full-time equivalent employment (which takes into account hours worked). Correlation coefficients for employment by industry are also listed.

The correlation coefficients are generally quite high, with the coefficients being similar for the number of employed and full-time equivalents. In contrast to other economic variables measured in surveys, the correlations for quarterly growth are stronger than for year-ended growth. Employment expectations for the most part also tend to be just as good as the actual outcomes. This is important, as the actual survey measures, which are compiled on a quarterly basis, do not have any advantage, in terms of timeliness, over the monthly labour force survey data. High correlation coefficients for expected quarterly growth in the next quarter are evident in most of the surveys, with the possible exception of the AIG-PWC and Yellow Pages surveys, suggesting that most of the surveys are useful in predicting employment growth in the next quarter. There is a modest deterioration in performance over the common sample period. Specific industry data appear to be of limited use.

Consistent with the high positive correlation coefficients, the business survey measures are significant in the autoregressive model (Tables 9c and 9d), with the exception of the Yellow Pages survey. Tests of Granger causality for employment were also quite positive (Table 9e), with the ACCI, ACCI-Westpac and Dun & Bradstreet surveys being forward-looking. The strong numbers overall probably reflect the fact that employment is a variable that the survey respondents have a strong influence over. Moreover, growth in employment generally lags growth in output.

5.8 Input Costs

A number of input costs are covered in the various business surveys. Most of the surveys cover labour costs such as wages, while some also cover non-wage labour costs and the cost of raw materials. The surveys report on both actual outcomes and expectations for the next quarter.

The results vary quite markedly across surveys, as well as across the variable being measured. The strongest results are for the ACCI-Westpac survey with a positive association evident between the survey's unit cost measure (for both actual and expected) and the ABS measure of Average Weekly Ordinary Time Earnings (AWOTE). This result is confirmed in the autoregressive model and the survey measure appears to be forward-looking as measured by the test of Granger causality. (This may reflect, to some extent, the long sample period, which includes large swings in wages growth.) Strong results were also recorded for the raw material cost component of the AIG-PWC survey and the ABS measure of manufacturing input prices, for both quarterly and year-ended growth. There also appears to be some information in the purchase cost and overhead components of the NAB survey when measured against the ABS measures of manufacturing input prices and non-wage labour costs respectively.

5.9 Selling Prices

Correlation coefficients of the various survey measures of selling prices (referred to as ‘final prices’ in the NAB survey and ‘prices charged’ in the Yellow Pages survey) with the consumer price index (CPI), the weighted median CPI and manufacturing output prices are shown in Tables 11a and 11b. With the exception of the ACCI National Survey of Business Expectations, correlation coefficients are quite positive for both quarterly and year-ended growth rates over both the full and common sample periods. The correlations are stronger for the CPI than for the weighted median CPI. For those surveys focusing on the manufacturing sector, the actual quarterly outcomes are highly correlated with the ABS manufacturing output price series, particularly the ACCI-Westpac series. In terms of expected quarterly growth rates, most surveys appear to be useful. These results are consistent with the results for the autoregressive model, suggesting that the selling price component of business surveys is providing information about current inflation in addition to information provided by lags of inflation. A number of the survey measures of selling prices are also shown to Granger-cause the CPI (Table 11e).

Footnotes

The ‘business conditions’ grouping which is used in the NAB survey is a simple average of three net balance indices relating to firms' trading conditions, profitability and number of employees. [9]

This measure is reproduced in the ABS ‘Business Indicators, Australia’ release (ABS Cat No 5676.0), and is based on the ABS Quarterly Economic Activity Survey (QEAS). [10]

Investment in livestock and intangible fixed assets accounts for the remainder. [11]

Results from this survey are reproduced in the ABS ‘Business Indicators, Australia’ release (ABS Cat No 5676.0). [12]