1. Michael Pedersen Central Bank of Chile
Use of Chilean business surveys in conjunctural assessment and short-term forecasting
Michael Pedersen
Central Bank of Chile
Fourth joint EC-OECD workshop on business and consumer opinion surveys

2. The presentation Short presentation of the Chilean business surveys.
Case studies: Information in surveys about Chilean economic activity.
Evaluating information in aggregated indicators.
Using surveys for short-term forecasting.
Information in cross-checking answers.
Final remarks.

3. Some main characteristics: Frequency: monthly.
The Chilean business surveys are designed in line with the recommendations in the OECD handbook
Some main characteristics:
Frequency: monthly.
Sample: fixed panel of the largest firms, random selection of smaller ones.
Total sample: 607 firms (16% of GDP).
4 sectors: mining, manufacturing, retail and construction.
Presentation of results: diffusion indices calculated with simple balances.
En el contexto de reducir la incertidumbre en las proyecciones trimestrales de corto plazo juega un rol clave la incorporación de información mensual.
La iniciativa busca evaluar la capacidad predictiva de la información mensual en proyecciones trimestrales e incorporarla sistemáticamente mediante la utilización de modelos Bridge.
Benchmarks: arima agregado
P.e.: capacidad predictiva del imacec contrucción para proyectar PIB construcción 3

4. The series are available from November 2003
Difussion indices
Source: ICARE / UAI

5. Three case studies are presented to illustrate the information in the surveys
Because of the very small sample, the use of surveys in the conjunctural analysis has so far mainly been on an ad-hoc basis.
However, now almost six years of data are available and it is possible to analyze the statistical properties of the series – on a preliminary basis.

6. 1a. Relatively high coefficients of correlation for manufacturing and retail
Diffusion indices and annual growth rates
Retail sector (26%)
Manufacturing sector (39%)
Mining (17%)
Construction (18%)
Sources: ICARE / UAI, Pozo and Stanger (2009) and own calculations.

7. Cross correlation coefficients Tests for Granger causality
1b. Tests suggest that surveys of manufacturing and retail Granger cause the activity indicators
Cross correlation coefficients
Sources: ICARE / UAI, Pozo and Stanger (2009) and own calculations.
Note: Negative numbers on the first axis indicate that the business survey leads activity.
Tests for Granger causality
IMCE
ICOM
ICIN
ICMI
ICOT
Activity survey
0.2 1
0.95
0.62
0.87
0.26
Survey
activity
0.10
0.00
0.00
0.76
0.43
Source: Own calculations.
Note: p-values for the null hypothesis of no Granger causality tested in bivariate VAR models with the number of lags selected according to Schwarz information criteria.

8. 2. Preliminary estimations indicate that the general survey contains information which is useful for short-term forecasting
Out-of-sample one-step-ahead forecasting exercise
Source: Own calculations.
Note: aRMSE of the business survey model divided by the RMSE of the AR model. bPercentage of the twelve observations where the business survey model predicts better than the AR model. cp-value of the Diebold and Mariano (1995) test for the hypothesis that the models have equal predictive power against the alternative that the business survey model is better.

9. 3a. Why did the stock accumulation fall in the last three quarters?
Hypothesis 1: Because of restrictive financial conditions, firms could not borrow money to finance production.
Hypothesis 2: Because of expectations of lower sale, the optimal stock level implies a reduction.
Decompose the fraction of firms which replied that stocks were higher than desired:

10. 3b. Firms considered stock levels too high because of their expectation of future demand
Stock flow and cross-checking business survey answers
A: Stock flow / GDP (%)
B: Actual demand (%)
C: Future production (3M) (%)
D: Future general situation (6M) (%)
Source: Echavarría et al. (2009)
Note: Figure A shows the stock flows as a percentage of GDP. B-D show the proportion of the firms that considered that their stock levels were higher than desired and also considered that: (B) actual demand was low, normal and high, respectively; (C) future production will go down, not change and go up, respectively, and (D) the future general situation of the company will be worse, the same and better, respectively.

11. Several questions remain unanswered:
Final comments
The preliminary results are promising with respect to the contents in Chilean business surveys about the economic activity.
Several questions remain unanswered:
Are surveys affected by seasonality?
Do surveys lead annual growth rates, or rather monthly?
Levels or changes in surveys serve as leading indicators?
Ongoing research in the Central Bank of Chile aims at answering some of these questions.

12. Michael Pedersen Central Bank of Chile
Use of Chilean business surveys in conjunctural assessment and short-term forecasting
Michael Pedersen
Central Bank of Chile
Fourth joint EC-OECD workshop on business and consumer opinion surveys