I am thankful to Ms Liljana Boçi and. Ms Alma Kondi, the Albanian Institute of Statistics (INSTAT), for their contribution and close cooperation during the selection
34 KB – 37 Pages
PAGE – 1 ============
BANK OF ALBANIA Methodology of Harmonised Confidence Surveys APRIL 2017 1 Ermelinda Kristo* 1 – * Bank of Alba nia, Monetary Policy Department, Short – term Forecasting Office. I am thankful to Ms Liljana Boçi and Ms Alma Kondi, the Albanian Institute of Statistics (INSTAT), for their contribution and close cooperation during the selection and application process of the best methodological practices of the surveys, some of which are explained in the second part of this paper.
PAGE – 2 ============
Content I. Introduction .. .. .. .. . 4 II. Methodology of Confidence Surveys .. .. .. 6 III. Confidence indicators and presentation of results .. .. 13 IV. Methodological changes after the inclusion in the EC programme .. 14 Annex 1. The harmonized questionnaires .. .. .. 20 Annex 2. Classification of economic activities for business surveys .. . 31 Annex 3. Treatment of outliers in quantitative questions .. .. .. 33 Annex 4. Confidence Indicators Charts .. .. .. .. 35
PAGE – 3 ============
Abbreviations: BCS – Business Confidence Survey BoA Bank of Albania CCS – Consumer Confidence Survey CI – Confidence Indicator EC – European Commission EU – European Union ESI – Economic Sentiment Indicator INSTAT Albanian Institute of Statistics SBR – Statistical Businesses Register
PAGE – 4 ============
I. Introduction The purpose of this document is to pr ovide detailed information on the methodological aspects of business and consumer confidence surveys, harmonized with the methodology of other European Union countries. Confidence surveys provide facts and information about business conditions and the stat us of households based on their opinion (unlike classical surveys that gather facts through figures). Indicators obtained from the confidence surveys are known as qualitative information, as opposed to the quantitative information obtained from the traditi onal surveys. These indicators complement the set of information received from the national accounts. Information obtained from confidence surveys is considered valuable even for advanced economies that have high quality national accounts because they: (i) provide timely information for the current situation of the economy; (ii) have information on aspects not covered by official statistics; (iii) are not revised. The scheme below shows the relation between the two measurements: traditional quantitative mea surements and confidence surveys. Source: Adapted from Cunningham (1997) 2 . 2 Cunningham, A. (1997): “Quantifying survey data”, Bank of England, Quarterly Bulletin, Article, August 1997.
PAGE – 5 ============
The results obtained from confidence surveys are considered valuable if they help explaining or predicting the developments of the official quantitative indicators. If the aspects of business and consumer behaviour measured through confidence surveys are correlated and explain the indicators from official statistics, then the survey data are considered valid to use (e.g.: the consumers propensity to consume is positively cor related to the consumption of population from national accounts; the businesses propensity for investments is correlated with the series of official investments series, etc.) The first confidence surveys in advanced economies date back to the 1920s, initi ally organized by chambers of commerce and later by statistical institutions and central banks. Among the earliest business confidence surveys are those organized by the Confederation of British Industries, the Institute of Economic Research (IFO) in Germa ny and the National Institute of Statistics and Economic Studies (INSEE) in France 3 . Regarding the consumer confidence, the first survey for measuring their sentiment was organized in the United States in 1946 4 . The confidence surveys of the Bank of Alban ia started in 2002, in cooperation with the Institute of Statistics (INSTAT) and initially with the assistance of the IFO. These surveys have provided important information to better understand economic developments. The results of the confidence surveys a re used in the periodic analysis of the Bank of Albania to supplement the information obtained from the official statistics. During the last six years, the information from selected balances in the surveys has been included in the short – term forecasting mo dels of economic growth. The Bank of Albania uses the confidence surveys to measure inflation expectations of both businesses and consumers, through direct questioning. Recently, detailed information at micro level obtained from the confidence surveys is u sed to construct economic uncertainty indicators. Since May 2016, the confidence surveys are carried out on a monthly basis, in line with the methodology of the European Commission. The quality of the information collected from the confidence surveys is p eriodically analysed at 3 4 https://data.sca.isr.umich.edu/survey – info.php
PAGE – 6 ============
the Bank of Albania. The empirical studies have proven the validity of information taken from them over the years 5 . At the same time, the confidence surveys have evolved, thus meeting the need to enhance the quality of these surveys , in the national and international plan 6 . Starting from May 2016, the confidence surveys are carried out under the Joint Harmonized EU Programme of Business and Consumer Surveys carried out in member and candidate countries. This programme aims to improve the quality of survey data and to align the practices used from different countries. Information from a common methodology is comparable and may be used to study the cyclical economic development of various countries. This document explains the methodolo gical characteristics of the confidence surveys (second part), describes and presents the confidence indicators (third part) and outlines the changes that resulted after participating in the EC programme (fourth part). The detailed questionnaires that are used for these surveys and more technical aspects of the methodology are presented in the annexes. II. Methodology of Confidence Surveys The main characteristics of every survey are the questionnaires, the selection of a representative sample, the way information is collected, and lastly the aggregation and presentation of the results. In contrast to traditional quantitative surveys, the process is faster in the case of confidence surveys. The qualitative nature of the questions facilitates the answerin g process of businesses or households. Also, the data collection and their aggregation in quantitative indicators is a rather easier process compared to quantitative surveys. The main characteristics of the confidence surveys are explained in the following . The questionnaire. One of the main characteristics of the questionnaire of confidence surveys 5 a new informati – 6 Cooperatio n and technical assistance from IFO until 2006 and with the European Commission from 2015.
PAGE – 8 ============
operates. Target population and sample. The accuracy of results drawn from the confidence surveys depends on the sample design and on the quality and completeness of the lists of the businesses and consumers. The Statistical Businesses Register (SBR) of INSTAT serves as the population list for t he businesses survey. The register includes every legal units registered at the National Centre of Registration and Tax Authorities. Table 1 shows the main statistics of the population for all four sectors of the business surveys based on SBR 2015, compile d by INSTAT experts. Firms operating in the industry and services sectors employing less than 5 employees and firms operating in construction and trade sectors employing less than 2 employees are excluded before proceeding further. This is called the frame after cut – off and will serve as the base for randomly selecting the businesses to form the representative sample. Table 1: Data on population frame and sample (nr. of businesses) Sector Population (nr. of businesses registered in SBR) Frame before cut – off Frame after cut – off Sample Industry 10257 9978 2264 404 Construction 4946 4821 2858 203 Services 52125 51470 4695 342 Retail Trade 45529 45177 15823 342 For the industry , the frame after the cut – off includes 23% of the number of the enterprises in the sector, which account for 93% of the total turnover of the sector and for 80% of the employees. For construction , the frame after the cut – off includes 59% of the enterprises , which represent 99% of the turnover and 94% of the employees in the construction sector. The frame for the services survey includes only 9% of the total enterprises in this sector, demonstrating a very fragmented sector. On the other hand these enterprises represent 80% of the total turnover of the sector and 48% of the employees in the service sector. Finally, in the trade sector, the number of businesses in the frame accounts for 22% of the whole population from
PAGE – 9 ============
SBR (whole trade sector). These enterprises account for 85% of the turnover and 64% of the total employees. which serves as population f or CCS, is taken from the Population and Housing Census 2011 data. This register contains the complete count of all persons, households and dwellings in Albania. On 1 January 2016, the number of population over 15 years old, which serves as population of C CS, consists of 2.376 million individuals. S ampling method. Theoretically, there are several methods used to extract information from a given population. The direct way is to collect the desired information from all the individuals within the population being surveyed. In practise this is a costly method which takes too much time. Other methods rely on a sub – population being surveyed. The extracted information from the sample represents the whole popula tion. The two main methods to create a sample are: purposive and random sampling. In the case of confidence surveys of the Bank of Albania, the sample is chosen through random selection, since it provides representative results for the whole population wit hout making further assumptions. In the case of CCS, the sample is created to represent all consumers aged above 15 years. For the BCS, the population represents almost all the businesses registered in the Statistical Business Register of INSTAT. Annex 2 c ontains more details on the BCS population. The list of the best methodology practices for the confidence survey, suggests that, before continuing with the random selection, the population should be separated into groups or strata with similar characteris tics, e.g. population of similar size group. The reason for this grouping is that the variation inside the group is smaller than the variation of the whole population. Because of this, random selection becomes a more efficient method, ensuring thus a more accurate evaluation of the population indicators. In order to define the strata, other official indicators known for the surveyed population are used. After stratification, the random selection is applied on each of these groups. For businesses, the strata are chosen based on two
PAGE – 10 ============
criteria: size and type of economic activity. For the first criteria, the size, the information on the number of the employees from the Businesses Register of INSTAT is used. For dividing by economic activity, the classification of businesses registered according to the Nomenclature of Economic Activities Rev.2 is used. In accordance with the best practices, the allocation of businesses into strata is done using the Neyman optimum allocation, based on the number of employees. Regard ing consumers, the criteria for the selection of strata are: geography and the density of the region where they belong to. Data collection. The field work is done between day 1 and 15 of each month when businesses and households respond to the questionnai res. Interviewers are trained both periodically and every time new questions are introduced in the questionnaire . The data entry, the quality control, the weighting and the aggregation process are done in 5 to 7 working days before the end of the month. B efore starting the aggregation process, the outliers are identified and removed. Annex 3 explains the method used for this process. Aggregation. The process of aggregation means that the qualitative information taken from the confidence surveys is quantif ied in a single number. The aggregation process is carried out in three main levels, starting with each question (net balance), then in sector level (confidence indicators) and finally extracting a representative indicator for the whole economy (Economic S entiment Indicator). In the first level of aggregation, the responses of businesses and consumers are quantified through the balance indicator. This is the difference between the percentages of businesses/consumers that report an improvement to those that report deterioration. The balance values range between – 100 and 100 percentage points. Although, in the case of businesses they are required to report changes compared to the previous quarter accounting for the seasonal fluctuations, the balances show sea sonal behaviour. Because of this, the time series are adjusted for seasonality.
PAGE – 11 ============
After the first level of aggregation, net balances are evaluated to check how informative they are. The most informative balances, which have the strongest linear connection w ith the reference series, are aggregated at sector level, in order to obtain the confidence indicators. After the inclusion in the EC programme, the composition of the confidence indicators was revised. More information on the changes incurred after the pr ocess of adapting to a common methodology may be found in part IV of this methodology document. Finally, all selected Weighting. ghted after the questionnaire is filled out and data entry has finished. The weighting criteria are: geographic area, age of the respondent, gender and size/density of the population. The weighting process is done after the data collection, because parts o (e.g. age or gender). In the case of businesses, weighting is done in two phases. The weighting in the first phase reflects the sample structure, whereas weighting in the second phase reflects the structure of the economy. The second phase weights are in proportion with the relative size of the different branches of the economy. By using the terminology of the OECD manual, the first phase weights are called sample weights, while the second phase weights are called size weights. The sample weights are calculated as the inverse of the inclusion probability for an firm to be selected in the sample. In the case of large enterprises, where the probability of inclusion in the sample is 1, the weight is also 1. The small and medium – sized firms weights are more than 1, being that the probability of selection in a given strata is smaller than 1. The larger the strata is, the smaller is the probability of the firm to be selected, the larger is the weight that is going to be assigned to this firm. So the firm that belongs to a larger group has the biggest weight in the sample. Somehow its response will represent the answer of other firms within that group. If the sample would be the simple ra ndom selection, not stratified, every firm would have the same probability of representation and weight.
34 KB – 37 Pages