Abstract
Business surveys indicators represent an important tool in economic analysis and forecasting practices. While there is wide consensus on the coincident properties of such data, there is mixed evidence on their ability to forecast macroeconomic developments in the short term. In this study we extend the previous research on business surveys predictive content by examining for the first time the leading properties of the main business survey indicators coming from the Italian survey on inflation and growth expectations (SIGE). To this end we provide a complete characterization of the business cycle leading/coincident properties of SIGE data (turning points, average duration, synchronization etc.) with respect to the National Accounts reference series using both non parametric approaches (i.e. Harding and Pagan in J Monet Econ 49(2):365–381, 2002) and econometric models (discrete and continuous dynamic single equation models). Overall the results indicate that in both the approaches SIGE business indicators are able to early detect turning points of their corresponding national account reference series in almost all cases. Overall, the average lead of troughs is found to be higher than the average lead of peaks.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
ISM indicators are produced by the Institute for Supply Management on monthly basis. Respondents (business and executive) provide information on production, employment inventories etc. that is used to construct a diffusion index of economic activity.
The Bank of Italy survey has been designed to provide information on a wide range of business cycle indicators and is aimed at furnishing a timely outlook on the development of the Italian economy.
The Quadratic Probability Score (QPS) is a probability analog of mean squared error and it is defined as \( QPS = \frac{1}{T}\mathop \sum \limits_{t = 1}^{T} \left( {f_{t} - x_{t} } \right)^{2} , \) where \( f_{t} \) is the forecast probability made at time t, and \( x_{t} \) is the realization of the event at time t. T is the total number of the observations or forecasting quarters. The QPS ranges from 0 to 1 with a score of 0 corresponding to perfect accuracy, and is a function only of the difference between the assessed probabilities and realizations.
The weights are the inverse of sample probability of inclusion.
Survey indicators usually refer to industry and services sectors.
While most survey microdata are weighted with the inverse of sample inclusion probability, this variable is weighted by also considering the number of employees.
Currently the business survey data published in the Bank of Italy Statistical Bulletin are not seasonally adjusted.
The dating procedure for business cycle turning points consists in finding a series of local maxima and minima that allow segmenting the series into expansions and contractions. The algorithm requires implementing the following three steps on a quarterly series: (1) Identification. Using a window of 2 quarters on each sides, a local maximum yt is defined such that (yt − 2, yt − 1) < yt > (yt + 1, yt + 2); (2) Alternation rule. A local maximum must be followed by a local minimum, and viceversa. In the case of two consecutive maxima (minima), the highest (lowest) peak is chosen. (3) Censoring rule. A set of rules concerning the duration and amplitude of phases and complete cycles is applied in order to retrieve only significant series movements and avoid some of the series noise.
The algorithm can be considered an extension, for quarterly data, of the Bry and Boschan (1971) procedure used to date the US business cycle.
The classical business cycle definition considers slowdowns and increases in the absolute levels of the economic activity.
The choice of the detrending method for removing trend components from the data also implies some priors on the true business cycles length and therefore may introduce some distortions in the dating algorithm.
Although the year-on-year growth rates of a series are able to detect trend components in the data, they produce a cyclical component that contains the highest business cycle frequencies with respect to detrended series obtained with moving averages.
The definition of growth rate cycle that we adopt is based on a simple year-on-year growth rate and is different from that based on quarter-to-quarter growth rate used by ECRI (usually invoked in the literature) in which the growth rate is normalized with the previous six months cumulative growth rate of the series.
In a growth cycle perspective, a turning point occurs in a series when the deviation-from-trend series reached a local maximum (Peak) or a local minimum (Trough). Growth cycle peaks (end of expansion) occur when activity is furthest above its trend level. Growth cycle troughs (end of contraction/contraction) occur when activity is furthest below its trend level.
The turning points reported in the table are detected on the common sample.
See Croushore and Stark (2003) for a wide survey of shortcoming and advantages of using vintages.
In other words, conditional on the information set \( \varOmega_{t - 1} \),\( y_{t} \) has a Bernoulli distribution: \( y_{t} |\varOmega_{t - 1} \sim B\left( {p_{t} } \right) \).
More specifically, the forecast of the first observation in the 1997Q3 period was obtained with parameter estimates using data up to 1997Q2. Subsequent forecasts were calculated by re-estimating each model with the new data point and then forecasting the next observation.
Concerning investments survey data predictive content Osterholm (2013) finds that survey data on investment goods industry can improve the forecasts of business investment growth.
References
Abberger, S. (2007). Qualitative business surveys and the assessment of employment—A case study for Germany. International Journal of Forecasting, 23, 249–258.
Bergstrom, R. (1995). The relationship between manufacturing production and different business survey series in Sweden 1968–1992. International Journal of Forecasting, 11, 379–393.
Bruno, G., & Lupi, C. (2004). Forecasting industrial production and the early detection of turning points. Empirical Economics, 29(3), 647–671.
Bry, G., & Boschan, C. (1971). Cyclical analysis of time series: Selected procedures and computer programs. Cambridge: NBER. (technical paper no. 20).
Burns, A. F., & Mitchel, C. (1946). Measuring business cycles. New York: National Bureau of Economic Research.
Cesaroni, T. (2011). The cyclical behaviour of business survey data. Empirical Economics, 41(3), 747–768.
Cesaroni, T., Maccini, L., & Malgarini, M. (2011). Business cycle stylized facts and inventories behavior: New evidence for the euro area. International Journal of Production Economics, 133(1), 12–24.
Chauvet, M., & Potter, S. (2005). Forecasting contraction using the yield curve. Journal of Forecasting, 24, 77–103.
Christiansen, C., Eriksen, J., & Moller, S. (2014). Forecasting US contractions: The role of sentiment. Journal of Banking and Finance, 49, 459–468.
Claveria, O., Pons, E., & Ramos, R. (2007). Business and consumer expectations and macroeconomic forecasts. International Journal of Forecasting, 23, 47–69.
Croushore, D., & Stark, T. (2003). A real-time data set for macroeconomists: Does the data vintage matter? The Review of Economics and Statistics, 85(3), 605–617.
Harding, D., & Pagan, A. (2002). Dissecting the cycle: A methodological investigation. Journal of Monetary Economics, 49(2), 365–381.
Kauppi, H., & Saikkonen, P. (2008). Predicting U.S. contractions with dynamic binary response models. Review of Economics and Statistics, 90, 777–791.
Lahiri, K., & Monokroussos, G. (2013). Nowcasting US GDP: The role of IMS business surveys. International Journal of Forecasting, 29, 644–658.
Lemmens, A., Croux, C., & Dekimpe, M. (2005). On the predictive content of production surveys: A pan—European Study. International Journal of Forecasting, 21, 363–375.
Mintz, I. (1969). Dating postwar business cycles: Methods and their application to Western Germany, 1950–1967. New York: National Bureau of economic research. (Occasional paper n. 107).
Osterholm, P. (2013). Forecasting business investment in the short term using survey data. Munich: National Institute of Economic Research. (working papers 131).
Acknowledgements
The authors thank all the participants to the Joint OECD/EC workshop “Recent Developments in Business and Consumer surveys” held on 30 November 2015 in Paris for their comments. We would like also to thank Leandro D’Aurizio, Luigi Cannari, Riccardo De Bonis, Nicola Branzoli and all the participants of the Bank of Italy lunch seminar held on 28 May 2015 for their suggestions.
Author information
Authors and Affiliations
Corresponding author
Data Description
Data Description
Survey Data:
-
INFL_EXP: Inflation expectations.
“In October consumer price inflation, measured by the 12-month change in the Harmonized Index of Consumer Prices, was 0.0% in Italy and 0.4% in the Euro Area. What do you think it will be in Italy in March 2015?”.
-
SELLP_EXP: Expectations on firms’ own selling prices.
“For the next 12 months, what do you expect will be the average change in your firm’s prices?”.
-
EMPL_EXP: Expectations on number of employees.
“Your firm’s total number of employees in the next 3 months will be Lower, Unchanged or Higher?”.
-
INV_COND: Assessment on current investment conditions.
“Compared with 3 months ago, do you think conditions for investment are Better, The same, Worse?”.
-
BUSINESS_3M: Three-month firms’ business condition expectations.
“How do you think business conditions for your company will be in the next 3 months? Much better, Better, The same, Worse, Much worse”.
-
BUSINESS_3Y: Three-year firm business condition expectations.
“How do you think business conditions for your company will be in the next 3 years? Much better, Better, The same, Worse, Much worse”.
-
GENERAL_SIT: Expectations on Italy’s general economic situation.
“Compared with 3 months ago, do you consider Italy’s general economic situation Better, The same, Worse?”.
-
PROB_IMPROVE: Probability of improvement of the economy in the next 3 months.
“What do you think is the probability of an improvement in Italy’s general economic situation in the next 3 months? 0, 1–25%, 26–50%, 51–75%, 76–99%, 100%”.
National Accounts Data:
-
HICP: Harmonized index of consumer prices.
-
EMPL: Number of total employed population, adjusted for seasonality, total economy.
-
INV: Gross fixed investments, adjusted for seasonality, total economy.
-
GDP: Gross domestic product, adjusted for seasonality, total economy (Fig. 1).
Fig. 1 
Business cycle turning points of survey indicators (red lines) and official statistics (blue lines)—Red triangles are the peaks and troughs detected in survey indicators. The grey/white stripes are the contraction/expansion periods detected in official statistics series. Note Qualitative survey indicators (EMPL_EXP, INV_COND, BUSINESS_3M, BUSINESS_3Y and GENERAL_SIT) are expressed as balances between positive and negative answers. Quantitative survey indicators (INFL_EXP, SELLP_EXP and PROB_IMPROVE) are expressed as average values. Official statistics series (INFL, GDP, EMPL and INV) are expressed as year-on-year growth rates. Left scale: official statistics data. Right scale: business survey data
Rights and permissions
About this article
Cite this article
Cesaroni, T., Iezzi, S. The Predictive Content of Business Survey Indicators: Evidence from SIGE. J Bus Cycle Res 13, 75–104 (2017). https://doi.org/10.1007/s41549-017-0015-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s41549-017-0015-8