Using ABS statistics: Telling the right story

Using ABS statistics: Telling the right story

Using ABS Statistics: Telling the right story

Understanding ABS Statistics
Limitations of ABS data

Introduction

The Australian Bureau of Statistics (ABS) has a statutory duty to provide reliable information. The statistics produced by the ABS are used by governments, business and the general community to inform policy and business decisions or in supporting a news story. Statistics therefore must be used in the appropriate way, and presented in a manner that is meaningful and easily understood by readers.

Understanding ABS statistics

Ensuring accuracy and integrity in the reporting of statistics is a serious responsibility. In cases where there may not be an in-depth understanding of statistics in general, or of a particular topic, the use of glossaries, explanatory notes and classifications will assist in their interpretation.

Statistics can be presented and used in ways that may lead readers to draw misleading conclusions. It is possible to take numbers out of context.

Some common areas where mistakes can occur relate to survey sample size, non-response in census and surveys, margin of error and terminology. When these are not taken into account, conclusions drawn from statistics can be affected. Sometimes conclusions drawn from the use of the average or mean figure are not as useful as those drawn from the use of median. There are also pitfalls in using the wrong data. See How reliable are my conclusions?

All data have their limitations (see ABS Data Quality Framework, cat no. 1520.0) and it is important to establish these limitations with regard to the data that is being used. Giving due consideration to these limitations ensures that the right story will likely be presented.

Non-response

Non-response (not stating an answer to a question) is a problem that imposes a limitation on the data provided by the ABS. This can occur in censuses (for example, the ABS Census of Population and Housing) as well as in surveys (for example, the ABS Survey of Mental Health and Well Being). Refer toSurvey or Census: what's the difference?

In the 2006 Census of Population and Housing a large percentage of the population did not respond to the question on Religious affiliation or Individual income . This percentage was as high as 10%. The non-response rate for the question on Religion is always high because it is an optional question. Note this is the only optional question in the Census.

If the non-response is taken into account when drawing conclusions about either religion or individual income, the data becomes more meaningful. It may be more accurate to state that 'a certain % of those who stated their income were in a particular income range'. Similarly this applies to religious affiliation.

 Example Sydney (C) Local Government Area, Individual income Count of persons Population aged 15 years and over: 145,298 Of whom: Stated their weekly income 108,676 Of whom: Stated weekly income over \$2000 11,647 Misleading assumption: 11,647 out of 145,298 = 8% Correct assumption: 11,647 out of 108,676 = 11% Misleading conclusion: Only 8% of the population in the Sydney (C) had an income of over \$2,000 per week', Correct conclusion: 'About 11% of those who stated their income in the Sydney (C) had an income of over \$2,000 per week.'

If we look at one of the Statistical Local Areas within Sydney (C), the difference is even greater where 40% did not state their income.
The region or geographic area being used will affect the comparison of statistics. See Comparison Pitfalls.Terminology

In using terminology it is important to understand the definition of the terms that the ABS adopts in the dissemination of its data. Refer to Understand the context

 Example 1 Full-time/Part-time Employment: The Census data relating to full-time and part-time employment is based on the hours an employed person worked in the week prior to the Census. This is often not taken into account. As a result the data is taken to refer to full-time and part-time employment in the Census year. This also applies to assessment of unemployed.
 Example 2 Method of travel to work: The data for method of travel to work also has limitations because it is about the method used on census day. For the 2006 Census, this fact as well as the number of employed persons who did not go to work on Census day (10.9% of total employed), the number of persons who did not state their method of travel to work (1.8%) and the number of those who worked at home (a total of 4.7%, some of whom may have worked at home only on that day) impose limitations on the data relating to the Method of Travel to Work used by Employed Persons. In reporting this data, a more reliable picture could be provided by qualifying this data with the words ' On Census day, of those who went to work and stated their method of travel ...(4.9% walked to work)'.
Percentage (Percentages explained)

While percentages are summary indicators, sometimes they can be misleading.

 Example 2006 Census of Population and Housing Question: Of the NSW Local Government Areas, which has the highest proportion of unmarried (Registered Marital Status) women aged 18 to 40? Answer: If you look at percentages only, it is Brewarrina at 82%. However, if you look only at the numbers (ie. not the proportion), a different story emerges. As shown in the table below: Sydney (C) had the highest number at 32,241 out of 43,055 (75%) and Blacktown (C) had the second largest number at 21,982 out of 47,767 (46%). Whereas Brewarrina (C) had a low number though it amounted to a high percentage of 82%. Further, Brewarrina is not a typical NSW Local Government Area as it is comparatively small and in a remote location. If we look at the Social Marital Status of persons in this LGA, the percentage of women aged 18-40 years who are unmarried is only about 47%.
 Local Government Area Unmarried number of women aged 18 to 40 Total women aged 18 to 40 Percentage of unmarried women aged 18 to 40 Sydney (C) 32,241 43,055 75% Blacktown (C) 21,982 47,767 46% Brewarrina (C) 244 298 82%

Thus, when the Census data is placed in the proper context, it tells a more meaningful story.