THIRD JOINT EUROPEAN COMMISSION

1

• THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
  ON INTERNATIONAL
• DEVELOPMENT OF BUSINESS AND CONSUMER TENDENCY
  SURVEYS
• BRUSSELS
• 12 13 NOVEMBER 2007
• Construction and Retail Trade ISAE surveys
  methodological aspects and cyclical features
• by Luciana Crosilla, Solange Leproux
• ISAE - Institute for Studies and Economic
  Analyses

2
INTRODUCTION

• ISAE has recently restructured the methodological
  framework for construction and retail trade
  surveys.
• For both sectors, the innovations specifically
  regard
• the frame list,
• the sample design,
• the weighting methods.
• The aims of this work is
• - to show the methodological and statistical
  aspects of the restructured ISAE surveys
• - to illustrate the capability of the surveys of
  tracking actual data.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
3
MAIN METHODOLOGICAL INNOVATIONS

• Main innovations concern
• Statistical Unit the firm has been adopted in
  place of the local Unit
• Frame list adoption of the official "ASIA"
  archives in place of telephone registers (Yellow
  Pages)
• Weighting Methods updating of the weights used
  in the data processing procedure.
• Further innovations are
• For the construction survey a new sample
  extraction criteria
• For the retail survey the theoretical
  predisposition of a unitary sample design and the
  adoption of a new rule distinguishing between the
  two kinds of distribution (traditional retail
  and large distribution).

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS
BRUSSELS 12 13 NOVEMBER 2007
4
THE TARGET UNIVERSE AND THE FRAME LIST

• Currently, the reference universe is made up of
  Italian firms classified in the following NACE
  (Rev. 1.1) sectors, as they result in the last
  Italian industry and service census.
• 45 for construction
• 50 and 52 for retail.
• Since its not possible to have a list (with data
  firms as address, name etc.) with all the firms
  surveyed in census, for extraction sample we use
  the official ASIA archives elaborated by ISTAT.
  ASIA is an official statistical register of
  active firms where we find all the structural
  information.
• Surveys are carried out
• with telephonic techniques for retail trade
  (since 2005)
• by mail for the construction sector.
• For the construction survey, the sample is over
  dimensioned with respect to the target (500
  firms) In order to assure a suitable response
  rate the sample is revised every two years in
  order to eliminate firms that failed to answer in
  the last 15 months.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
5
THE SAMPLING METHOD CONSTRUCTION

• The ISAE construction sample has been repeatedly
  revised and currently is represented by a
  reasoned panel of 500 firms, stratified into
  geographic regions (North West, North East,
  Centre, South) with proportional allocation of
  the units in the single strata (that is, in every
  single stratus there is a number of firms
  proportional to that of the correspondent stratus
  in the universe).
• However, a sample is not stratified according to
  the economic sectors in fact, ASIA is based on
  the NACE rev.1.1 classification that is different
  with respect to that proposed in the EU
  questionnaire (residential building, non
  residential building, civil engineering).

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
6
THE SAMPLING METHOD CONSTRUCTION
THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
7
THE SAMPLING METHOD CONSTRUCTION

• The firm selection is carried out by a mixed
  technique
• systematic-random extraction type with implicit
  stratification for small and medium firms (1-99
  employees)
• large firms (100 and more employees) have all
  been included in the sample.
• The selection, therefore, is not completely
  random that corresponds to the requirement to
  follow leader firms (-100 and more employees -
  which represent 0.1 of total firms and 5 of
  total employees in the sector).
• Leader firms represent the more reliable units in
  a sector where the birth-mortality rate of medium
  and small firms is high. Moreover, these firms
  are more sensible to markets change and they have
  economic importance more than proportional
  compared to their size.
• The new criterion of sampling - introduced in the
  course of 2005 and revised in 2007 - has produced
  a remarkable increase in the response rate, thus
  assuring a more representative sample.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
8
THE SAMPLING METHOD CONSTRUCTION
THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
9
THE SAMPLING METHOD RETAIL TRADE

• 
  Concerning the retail trade
  survey, the sample is a theoretical panel of
  1,000 respondents. The sample pattern, in
  particular, is complex and proportional. It is
  stratified by
• firm typology (traditional retail trade and large
  distribution)
• four geographical partitions (North-West
  North-East Centre South)
• five sectors of activity (Food Textiles and
  Clothing Household Articles among which Home
  Appliances and Other Articles Means of
  Transportation Other Products never surveyed
  before).
• In each sample strata, the number of respondents
  is proportional to both the reference population
  and to the weight of the strata in terms of
  turnover.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
10
THE SAMPLING METHOD CONSTRUCTION
THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
11
THE WEIGHTING METHOD

• For both surveys, ISAE adopts a two stage data
  processing procedure.
• In the first stage
• we weigh the answers of each firm with internal
  weights, that concur to reduce the distortions
  deriving from the different degree of firm
  influence on the total results. In the first
  stage of the process, the aggregated percentages
  of answers are obtained as a weighed average of
  the firm-specific answers, the weights being the
  number of the firm employees
• n n
• Xk (? Yi Xi) / ? Yi (1)
• i1
  i1
• where i indicates the single firm and n is the
  number of enterprises in the k activity sector.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
12
THE WEIGHTING METHOD

• In the second stage
• in order to obtain the results concerning to the
  overall sector, the data are aggregated using
  external specific weights. In particular, the
  answers of each sector (typology, in the case of
  retail trade) are combined using the
  corresponding turnover for the retail trade, and
  employees of the reference universe for the
  construction sector
• l l
• X. (? Wk Xk) / ? Wk (2)
• 
  k1 k1
• where Wk is the specific weight for the k
  sector (or typology) and l is the total of
  sectors (or typologies).
• In this second step, for construction a new
  weight system has been introduced since 2006. In
  particular, the results for total sector are a
  weighed average of every single sector
  (residential and non residential building, civil
  engineering), weighed on the basis of the
  investments estimated for 2004. For the new
  weights used, the series of balances for the
  total sector have been reconstructed since 1995.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
13
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS

• Two leading indicators have recently been
  elaborated for construction and retail sectors.
• In particular, the former is obtained as simple
  average of the seasonal adjusted balances of
  assessments on activity and expectations on order
  books, the latter is calculated as simple average
  of the expectations on business trend and on
  order books.
• Value added of construction Italian sector and
  households consumption are chosen as reference
  series.
• The aim of this analysis is to assess the in
  sample and out of sample properties of the two
  leading indicators in the period 1995, Ist Q-
  2007, IInd Q (construction sector) and 1990, Ist
  Q- 2007, IInd Q (retail trade).

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
14
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS THE APPLIED METHODOLOGY

• Turning point coherence turning points are
  calculated with the Harding e Pagan routine for
  both the LI and the reference series, calculating
  mean leads/lags for the LI.
• Correlation with reference series the
  cross-correlation coefficients are calculated
  between the cyclical components of the indicators
  and the reference series.
• In order to apply the econometric tests described
  below, the reference series are preliminarily
  subjected to testing for the presence of unit
  roots (ADF test).
• The quantitative reference series have been found
  to be non-stationary.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
15
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS THE APPLIED METHODOLOGY

• Granger Causality test We estimate the
  following equation
  l l
• ?logyt a ? ßi ?logyt-i ? ?i VIt-i et
  (3) i1 i1
• where
• i indicates the lag
• Yt indicates the reference variable in the
  first differences of the logarithms
• a indicates a constant.
• ßi indicates regression coefficients for the
  past values of the dependent variable
• ?i indicates regression coefficients for the
  past values of the independent variable VI (the
  leading indicators in question)
• et indicates the error.
• LI are considered not to Granger-cause the
  reference series if
• H0 ?i 0 for every i.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
16
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS THE APPLIED METHODOLOGY

• 4. In order to test the out-of-sample
  properties, we use the following equation
  
  l
  l
• ?logyt a ? ßi ?logyt-i ? ?i VIt-i et
  (4) i1 i0
• where the estimated values of the dependent
  variable are obtained using the static (one
  step) forecast.
• Performance of the indicators are then evaluated
  against the reference series by applying all the
  tests described.
• The in-sample and out-of-sample forecasting
  capabilities of the indicators are compared to
  those of the traditional EU-harmonised confidence
  climate with respect to the selected reference
  series.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
17
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS

• Leading indicator for the construction sector is
  characterized by
• three complete cycles, while the reference
  series has two cycles
• only six turning points are present for the value
  added while LVA is characterized by eight points
  
• According to the turning points alignment, the
  indicator has not got global leading ability .
  However, we mark that the indicator seems to have
  a good leading ability in the first years of the
  sample while it seems to have a lagging feature
  in the last years.
• Leading Indicator for the retail sector is
  characterized by
• Four complete cycles, while the reference series
  as three cycles
• Eight turning points for the reference series 10
  for the LI
• the leading indicator seems to lead the
  households expenditure, on the average, the
  reference series of two quarter. The leading is
  even better for downturns (-2.5 on average).

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
18
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
19
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS

• As for the 2nd step of analysis we highlight in
  table 5
• For the construction sector, the cross
  correlation between the new leading indicator and
  the reference series peaks at lead -2 (0.54)
  while the contemporary correlation is equal to
  0.23.
• For retail trade, the new indicator shows a good
  contemporary coefficient (0.52) and a quite high
  maximum correlation one quarter in advance
  (0.60).
• Table 5 also provides a first analysis of in
  sample properties of the leading indicators.
• We note that both leadings granger-cause
  reference series.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
20
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS

• Looking again at table 5, we can see the out of
  sample properties of the leading indicators.
• In estimating the equation (4), we considered up
  to 4 lags of the independent variable for
  construction and up to 8 lags for retail trade.
• Construction leading indicator has an acceptable
  forecast out-of-sample ability. Theil inequality
  coefficient is 0.54 with bias and covariance
  equal, respectively, to 0.013 and 0.95.
• As for retail trade, the leading indicator really
  seems to possess acceptable out of sample
  forecasting capabilities too the value of the
  Theil coefficient is 0.54 with a bias and
  covariance of 0.07 and 0.91 respectively.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
21
LEADING INDICATORS AND EC CONFIDENCE CLIMATES

• Encouraging results have been however obtained by
  comparing the cross correlation, the in-sample
  and out of sample forecasting capabilities of the
  indicators with those of the EC confidence
  climates monthly released for Italian
  construction and retail trade sectors. This is
  done using, of course, the value added and the
  households expenditures as reference series for
  construction and retail, respectively. Looking at
  table 5
• EC confidence climate for construction
• - cross-correlation peaks at lead -7 but the
  function value is similar to that of leading.
• - EC confidence climate does not Granger- cause
  the value added.
• - We notice that the forecasted values by the
  model including the confidence index, have a
  slightly higher Theil coefficient but they are
  more biased than the model including the leading
  indicator.
• All in all, the leading indicator seems to have a
  stronger correlation (two quarter in advance) and
  better forecasting capabilities than the EC
  confidence climate.
• EC confidence climate for retail trade
• - cross-correlation peaks at lead -1 but the
  function value is lower than that of the leading
  indicator.
• - EC confidence climate does not Granger- cause
  the households expenditure.
• - The Theil Inequality Coefficient is higher
  than that of the indicator but the forecasted
  values are more biased.
• . To conclude, the indicator seems to possess a
  stronger correlation (one step) and better out of
  sample forecasting capabilities than the EC
  indicator.

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
22
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
23
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
24
LEADING INDICATORS FOR THE ITALIAN CONSTRUCTION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS

THIRD JOINT EUROPEAN COMMISSION OECD WORKSHOP
ON INTERNATIONAL DEVELOPMENT OF BUSINESS AND
CONSUMER TENDENCY SURVEYS BRUSSELS 12 13
NOVEMBER 2007
25
CONTENTS
INTRODUCTION
THE RELEVANT INNOVATIONS
THE TARGET UNIVERSE
THE SAMPLING METHOD (1)
THE SAMPLING METHOD TAB1 (2)
THE SAMPLING METHOD (3)
THE SAMPLING METHOD TAB2 (4)
THE SAMPLING METHOD (5)
THE SAMPLING METHOD TAB3 (6)
THE WEIGHTING METHOD (1)
26
CONTENTS
THE WEIGHTING METHOD (2)
CONSTRUCTION SECTOR LEADING INDICATOR (2)
LEADING INDICATORS FOR THE ITALIAN CONSTRUCION
AND RETAIL TRADE SECTORSTHE APPLIED METHODOLOGY
(1)
LEADING INDICATORS FOR THE ITALIAN CONSTRUCION
AND RETAIL TRADE SECTORSTHE APPLIED METHODOLOGY
(2)
LEADING INDICATORS FOR THE ITALIAN CONSTRUCION
AND RETAIL TRADE SECTORSTHE APPLIED METHODOLOGY
(3)
LEADING INDICATORS FOR THE ITALIAN CONSTRUCION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS
TAB 4 LEADING INDICATORS FOR THE ITALIAN
CONSTRUCION AND RETAIL TRADE SECTORS ANALYSIS OF
THE RESULTS(1)

27
CONTENTS



LEADING INDICATORS FOR THE ITALIAN CONSTRUCION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS
LEADING INDICATORS FOR THE ITALIAN CONSTRUCION
AND RETAIL TRADE SECTORS ANALYSIS OF THE RESULTS
LEADING INDICATORS AND EC CONFIDENCE CLIMATE
TAB5
Construction fig1
Retail fig2