Contents >>
Section 10. Statistics: Concepts, Sources, Methods and Usage
To assist in your understanding of the statistics presented in this book, some of the more important or regularly occurring statistical concepts, sources, methods and usage are explained in this chapter. However, the explanations provided here are very brief, so if you require a detailed understanding of a topic, you must be prepared to undertake further research.
The ABS has a range of publications that discuss the following issues in detail. Some of these are included in the Further Reading reference at the end of this chapter. In addition, the publications listed as sources contain information on concepts, sources and methods of the statistics they relate to and, in some cases, provide reference to publications which explain the issues in further detail.
STATISTICAL CONCEPTS AND METHODS
Time Series
A data set is a collection of observations relating to a variable or group of variables. For example, a set of data could consist of observations of the population for each State and Territory in Australia at a single point in time, say Census night 1991. This provides a snapshot view of the population of Australia which could be used to compare populations of the various States and Territories in terms of age, sex, etc.
A time series is a list of observations for the same variable or group of variables over a period of time. For example, a time series could consist of the population for Australia for each year from 1980 to 1990. Time series enable recent estimates to be placed in a meaningful historical perspective, which permits analysts to see if the current situation is improving, deteriorating or staying much the same.
When compiling time series for analysis, care should be exercised that data has not been revised. Many statistical series produced by the ABS, especially derived series like national accounts, are subject to revision as more information becomes available. Seasonally adjusted and trend series are always subject to revision.
Classifications
Classification is one the cornerstones of statistical collection and analysis. Without the accurate and systematic arrangement of data according to common properties, statistical output cannot be comparable. Classifications group data into classes or categories according to various characteristics. For example, retail businesses may be classified according to what they sell. Instead of just compiling data about 'retailers', data could be compiled separately for footwear stores, butchers, newsagents etc.
The ABS has defined standard classifications that are used to present a wide range of data. ABS classifications align closely with international classifications enabling comparability with international statistics. A wide variety of organisations (government, private sector, educational institutions, etc.) use the ABS classifications for a variety of purposes including the analysis of data and running their own surveys and censuses. This enables them to compare their data with data from the ABS and from other organisations which use the same standard classifications.
Two of these standard classifications which are often used in reporting economic information are the classification of industries and the classification of institutional sectors.
Australian and New Zealand Standard Industrial Classification (ANZSIC)
ANZSIC is the standard classification used in Australia and New Zealand for the collection, compilation and publication of statistics by industry. The objective of the industrial classification is to identify groupings of businesses undertaking similar economic activities. Subject to certain criteria being met (economic significance and compliance with international standards), each such grouping defines an industry. The similar economic activities characterising the businesses concerned are referred to as primary activities. Each individual business is assigned an appropriate industry category on the basis of its predominant activities.
The ANZSIC structure comprises four levels: Divisions (the broadest level), Subdivisions, Groups and Classes (the finest level). At the Divisional level the main purpose is to provide a limited number of categories presenting a broad overall picture of the economy and suitable for publication in summary tables in official statistics. The Subdivision, Group and Class levels provide increasingly detailed dissections of the broad categories.
Structure of ANZSIC example:
Division C : Manufacturing
Subdivision 22: Textiles, Clothing, Footwear and Leather Manufacturing
Group Title 224: Clothing Manufacturing
Class 2242: Women's and Girls' Wear Manufacturing
There are 17 ANZSIC Divisions each identified by an alphabetical character, as presented:
A Agriculture, Forestry and Fishing
B Mining
C Manufacturing
D Electricity, Gas and Water Supply
E Construction
F Wholesale Trade
G Retail Trade
H Accommodation, Cafes and Restaurants
I Transport and Storage
J Communication Services
K Finance and Insurance
L Property and Business Services
M Government Administration and Defence
N Education
O Health and Community Services
P Cultural and Recreation Services
Q Personal and Other Services
Standard Economic Sector Classifications of Australia (SESCA)
SESCA presents a system for classifying institutional units (such as enterprises and households) into appropriate sectors of the economy. A major component of the SESCA is the Standard Institutional Sector Classification of Australia (SISCA) which forms the basis for the institutional sectoring of a range of statistics in the Australian Bureau of Statistics, including the National Accounts, Balance of Payments, International Investment, Financial Accounts and Government Finance Statistics.
SISCA is often used in conjunction with a number of other supporting classifications such as Public/Private classification, Level of Government classification, Jurisdiction classification and the Type of Legal Organisation classification (TOLO). This group of classifications and other related classifications are together presented as the SESCA. These associated classifications allow further identification and detail of particular characteristics of institutional units classified under the SISCA.
The structure of SISCA is as follows:
1 Non-financial corporations
2 Financial corporations
2.1 Central Bank and other supervisory authorities
2.1.1 Reserve Bank of Australia
2.1.2 Other supervisory authorities
2.2 Depository corporations
2.2.1 Banks
2.2.2 Other depository corporations
2.3 Insurance corporations and pension funds
2.3.1 Life Insurance
2.3.2 Pension funds
2.3.3 Other insurance corporations
2.4 Other financial institutions
2.4.1 Central banking authorities
2.4.2 Financial intermediaries n.e.c
2.4.3 Financial auxiliaries
3 General Government
4 Households
5 Non-profit institutions serving households
6 Rest of the world
Chain Volume Estimates
Chain volume estimates provide a convenient way of measuring changes in quantities (or 'real' change) in various economic statistics because they remove the direct effects of price changes.
Many economic statistics, such as gross domestic product, relate to a wide range of goods and services. Our difficulty is how to aggregate different units of measurement, e.g. the number of cars produced with tonnes of steel produced. If we use a common unit of measurement, i.e. money values (or dollars), we can express transactions for a range of goods and services as a single aggregate.
However, change in money values from one period to another is generally a combination of change in price and a change in quantity. In most cases, we are interested in changes in the physical quantities underlying the dollar values, e.g. the change in the number of cars produced. As a result, estimates are adjusted to remove the direct effects of price changes. Such estimates are said to be chain volume estimates (or in real terms).
The current price value of a transaction may be thought of as being the product of a price and a quantity. The value of a transaction in chain volume terms can be derived by linking together movements in volumes, calculated using the average prices of the previous financial year, and applying the compounded movements to the current price estimates of the reference year. The reference year for our chain volume measures is the year prior to the latest complete financial year (currently 2000-01).
It is not possible to derive chain volume estimates for items such as interest rates or profits that do not have price and quantity components. Nevertheless, such items can be expressed in real terms by deflation using a price index in order to measure changes in the purchasing power of the item. This involves dividing the current price values by a broad indicator of price change such as the CPI or the chain price index for GDP. The underlying assumption is that these price indexes are representative of price change of the goods and services that could be purchased with the money earned from profits, interest, etc.
Up until 1998 we produced constant price estimates to measure changes in quantities. However, the quality of constant price estimates deteriorate as relative prices change. Chain volume measures deal much better with this, hence our decision to adopt these measures and to discontinue the production of constant price estimates.
Indexes
An index number measures the value of a variable in relation to its value at a base period. The essential idea of index numbers is to give a picture of changes in a variable much like that drawn by saying ‘the price of petrol rose 5% from June 1992 to December 1993’. Index numbers measure change without giving the actual numerical value of the variable. Change is measured from a base period which is expressed as 100.0.
The index number = current value
------------------ x 100
base value
Because indexes summarise change, they are useful in economic analysis.
Movements in index numbers from one period to another can be expressed either as percentage changes or as changes in index points. It is important not to confuse the two methods because unless the comparison is with the base period, the two yield different results.
Seasonal Factors
Some data are influenced by the nature of the period to which they relate. For example, sales of sunblock are higher for January than for July. Normal seasonal influences on data are those effects that recur regularly one or more times a year. Data that are seasonal may reflect the influence of the seasons themselves (such as farm production) or social convention (such as the incidence of holidays) or economic factors (e.g. timing of tax payments and financial year timing). Some data reflect differences in the composition of the months or quarters in terms of the number of trading days in the period or accounting practices used.
This feature of the data can make interpreting monthly, quarterly and yearly changes difficult and so the ABS uses a special statistical tool called seasonal adjustment to standardise the data. Seasonally adjusted data has had all the calendar-related influences removed.
Seasonally adjusted data still contains the effects of irregular influences on the data. For example, sales of beer may have been affected by some large, one-off event such as a strike in several large breweries. Seasonal analysis does not remove such effects but the ABS is able to significantly dampen such irregular influences in seasonally adjusted series by producing smoothed seasonally adjusted or trend estimates.
Trend Estimates
The smoothing or trending procedure used by the ABS is based on a set of moving averages known as Henderson filters. These moving averages dampen the irregularity of data without distorting the timing, level or shape of turning points, i.e. peaks and troughs. Trend estimates provide a simple yet very effective measure of the underlying growth or decline of a time series. They also provide a much wider basis for analysis than the more erratic seasonally adjusted or original data.
National Accounts
With separate indicators, particular aspects of economic activity can be monitored. Another important use of this information is as the building blocks of a set of accounts for Australia, called the national accounts. Just as a set of accounts for a business consolidate a lot of information about the business and present it in a set format, national accounts consolidate a range of statistics, from those involving individuals to those involving the whole nation, into a consistent format which describes the overall economic position of the nation.
The concept of national accounting is quite old, having been developed as far back as the 17th century. However, its current look is relatively new, with welfare economists led by Pigou in the 1920s producing the first effective modern measurement of national income. A fundamental re-direction of emphasis in economic analysis and policy occurred after the acceptance and adoption of principles set down in John Maynard Keynes’ 1936 publication The General Theory of Employment, Interest and Money.
As a result, national accounting has developed as an integral part of economic analysis and policy advising. Macro-economic policy, concerned with the maintenance of income, price and employment stability, is dependent for much of its effectiveness on timely and accurate information on the components of domestic production. The national accounts provide conceptually consistent information and illustrate the relationships between the components.
Australia’s national accounts are compiled in a manner which closely accords with the recommendations of the System of National Accounts 1993 (SNA93). This document was produced jointly by five international organisations: the Commission of the European Communities, the International Monetary Fund, the Organisation for Economic Co-operation and Development, the United Nations and the World Bank. SNA93 is expected to provide a framework for national account statistics into the 21st century.
Australia's national accounts record the essential elements of the Australian economy: production, income, consumption (intermediate and final), accumulation of assets and liabilities, and wealth. The starting point for the system is production, which is recorded in the gross domestic product (GDP) account. The GDP account has two 'sides': income and expenditure.
On the income side of the account are the incomes accruing to the factors of production: compensation of employees (earned by wage and salary earners), gross operating surplus (profits) (earned by corporations, general government and the ownership of dwellings) and gross mixed income (earned by owners of unincorporated businesses), as well as net taxes on production and imports accruing to government. On the expenditure side of the account are final consumption expenditure, investment (represented by gross fixed capital expenditure and changes in inventories), plus the value of Australia’s exports (which are part of Australia’s total production) minus the value of imports (which represent part of the production of other nations).
It can be seen from the above that the familiar Keynesian identity Y = C + I + X - M (where Y is income, C consumption, I investment, X exports and M imports) is apparent on the expenditure side of the GDP account.
Complementing the GDP account are an income account, a capital account, a financial account, and a balance sheet, which shows the nation's wealth.
National accounts estimates attempt to account for every monetary transaction of every economic agent in the economy, as well as imputing a value for a range of transactions that do not involve the exchange of money (for example, when producers consume their own products). The quality of national accounts statistics depends to a large degree on the quality of the original records maintained by businesses, governments and other institutions from which data are obtained.
INTERPRETING STATISTICS
Definitions
It is important that your understanding of relevant terms correspond to the ABS definitions. This ensures that interpretation of terms is uniform and the information is used in the right context. For example, how do you define `unemployment’? Compare your definition with the ABS definition. ABS publications contain definitions of the information they include.
Footnotes
Footnotes are used to add comments and/or explanations to the tables or graphs. Footnotes are indicated by the inclusion of a letter in brackets e.g. (a), (b), (c), etc. beside the figure or heading which requires explanation. This letter and its footnote are presented under the table or chart.
The position of the footnote reference is important in the table or graph. If the footnote reference is in the title of the table or graph, then the message in the footnote relates to the whole table or graph. If it appears next to a column heading, then the message in the footnote applies to the data within that column. When analysing statistics, it is important to give attention to the footnotes as they often point out limitations in the data which could significantly affect interpretation.
Explanatory Notes
Explanatory notes are designed to assist the user in understanding the data in the publication. They provide information on the data collected and the method of collection and are useful in highlighting the limitations of the data. For example, explanatory notes generally include descriptions of the methodology and scope used to collect the data, data definitions, reliability of estimates, seasonal adjustment and comparability with other data.
Averages
An average (arithmetic mean) provides a useful summary measure of the contents of a set of data. However, averages can give a very deceptive picture of the meaning of statistics if they are misunderstood or misused. The average is affected by extremes in data (highest and lowest values) and unequal distributions. It may be beneficial in analysis to also examine the mode (most frequently occurring value) and the median (the value in the middle of an ordered data set) as a guide to the characteristics of the data.
Composition of Totals
Analysis of totals will give you an idea of overall trends in time series data. To gain a more complete understanding of the data, however, an analysis of the components making up the totals is necessary. For example, there were more women than men in Australia at the 1996 Census. However, further analysis shows men outnumbered women in each age group up to the 25-29 years age group. Women then have greater numbers until the 40-44 years age group. There are more males in each age group until 60-64 years, however women then consistently outnumber men in the older age groups.
Graphs
Graphs are an excellent way of presenting data. They enable the user to get a quicker feel for the data than when using tables or from text.
Graphs, however, can very easily mislead and care should be taken in interpretation. Care must be taken to understand what the title and axis headings mean and what data series are actually represented in the graph. Attention must be paid to the units (e.g. millions of dollars, persons) and the scales used.
Surveys and Censuses
Ideally, if we want to find out something about a group of people or businesses we would approach every person or business in the group (called the population). This is called a census. The best known census is the Census of Population and Housing, which collects information from every household in Australia. However, by applying the rules of sampling, a reliable picture of a population can be drawn from a selection or a sample of that population. The key lies in selecting a sample that is representative of the whole population.
An advantage of sample surveys over censuses is that they are cheaper and are easier to run. However, one main disadvantage is that the results contain sampling error, which is the difference in the results obtained by using a sample of the population rather than the whole population. In some instances this error can be quite large. Where information is being analysed from sample surveys, the size of this error should be taken into account when assessing the credibility of results. Sample survey and census results can also contain non-sampling error, which is error resulting from collection and processing errors, e.g. respondents being unable to accurately recall information or mistakes made in recording or coding.
STEPS IN ANALYSIS
Although there are no hard and fast rules to the correct approach, the following steps may give you a starting point for analysing time series data.
(a) Determine what data are available that are relevant to your topic. This publication is a good place to start. The references shown in the Further Reading part of each section will assist you in identifying sources of more detailed or related information. The ABS Catalogue of Publications and Products (1101.0) could also help you to determine what is available.
(b) Look at the layout of the table in order to understand how the data are arranged. Check the row and column names to obtain a clear idea of the variables being displayed.
(c) Scan the totals in the tables for an overall idea of the trends in the data. A graph is often the most appropriate tool for this analysis. If no graph is presented, consider graphing the data yourself to get a clear picture.
(d) If the data are available by different frequencies (e.g. annually, monthly), decide which of the available frequencies is most appropriate for your purpose. Annual data may be appropriate for examining data over a long time; quarterly or monthly data may provide a better picture of more recent developments.
(e) Make sure you have a clear idea of the questions for which you seek answers in the data. For example:
- are the values of the variable rising or falling over time?
- when was the last peak (high point) or trough (low point)?
- has the rate of change risen or fallen over time?
- have the shares of components in the total changed over time?
It is important to conduct your analysis one logical step at a time. Do not try to take all the information in at once and try not to get side-tracked with minor issues as you do your analysis.
Further Reading
Statistics - A Powerful Edge! (Cat. No. 1331.0)
A comprehensive guide to understanding statistics - designed for the reader to gain confidence in using statistical information.
Surviving Statistics: A User’s Guide to the Basics (Cat. No. 1332.0)
A comprehensive basic guide to understanding and using statistics.
Information Paper: A Guide to Interpreting Time Series - Monitoring Trends (Cat. No. 1349.0)
Explains why, in ABS publications, the main features and commentaries sections concerning most time series are increasingly emphasising the trend series rather than the seasonally adjusted or original data. It also explains how these trend estimates are obtained as well as how they may be used more effectively for informed decision making.