Annual municipal energy consumption estimates

This method uses a population-weighted, area-based statistical concordance (published by the ABS) to transform the postcode-level data currently sourced from energy distribution businesses into data for Local Government Areas (LGAs) and Statistical Local Areas (SLAs). This means that a count of the population within an LGA can be estimated from the population count of postcodes whose boundaries do not exactly fit the LGA/SLA boundaries.

This methodology needed to be applied to data from DBs where postcodes overlapped different LGAs and/or changed through time.

Postcodes are seen to be an easy way to present data. It is assumed that they are easy to remember, and that postcode boundaries are clear and stable. However, neither is necessarily the case, and there are a number of inherent difficulties with the analysis of any postcode-level data. These include:

• Out-of-date postcode boundary information: Australia Post postcode electronic boundaries for use in spatial analysis and mapping software were last updated in published form in the early 1990s. Since that time, many postcode boundaries in both urban and rural areas have been altered to align with suburbs or as a response to revised rural addressing. However, as no 'official' updates have been issued, the spatial impact of the changes cannot be accurately assessed at this time. The final development of a set of stable geographic definitions for postcodes that may assist 'fit-for-purpose' statistical analysis of administrative data by postcode is still some time in the future.
• Lack of comparable variables at the postcode level: Few (if any) statistical agencies collect direct observations of demographic or other variables by postcode. The ABS produces the Australian Standard Geography Classification (ASGC), a set of population-oriented boundaries used not only for its Census of Population and Housing but also by other agencies (e.g. Victoria in Future projections released by the Victorian Government). Postcode data may be approximated from data collected at other spatial scales, although the concordance process affects the integrity of the data (see below).
• Poor fit between postcodes and other geographies: Postcodes often do not align with other boundaries – in this case, Local Government Areas (LGAs). Analysis of GIS boundary layers held by DSE suggests that 30% of postcodes in Victoria are split across two or more LGAs. It is also worth noting that postcodes do not match ASGC boundaries. These impact on the comparability of postcode-level data to broader statistical data.

Limitations of concordances in manipulating postcode-level data: While some population-weighted concordances between official statistical or administrative regions and postcodes are available (and have been used in the following analysis), such concordances are untested for manipulating certain data, such as the characteristics of businesses identified only by their postcode and not their address. Population weighted concordances assume a distribution across space based on residential patterns, which may not (and usually do not) apply to non-residential addresses. Use of concordances to re-allocate data from one geography to another can therefore result in data being allocated to the wrong area.

Some preliminary experimental statistics resulting from the methodology are published. Further analysis of the statistics to elicit trends, variance and other explanatory variables will be pursued in later stages of the project. This may be complemented with analysis by a range of stakeholder organisations once these benchmark energy consumption metrics have been made more widely available.

Top of page

Advantages

Top of page

Data issues

There are a number of issues that may impact on the quality of annual municipal energy consumption statistics. These should be kept in mind when reviewing the resulting data. The issues are outlined below, and our treatment of them is described.

Potential spatial error

Using a population-weighted, area-based concordance to transform original data introduces a level of spatial error.

Definitions of tariffs and classes

Specific definitions of tariff classifications vary between individual DBs and may also include secondary information which each DB uses to assign tariff, such as:

• consumption of energy for various purposes or by the activity (economic status) of the consumer, including:
  • consumption for residential purposes
  • consumption for rural or agricultural purposes
  • consumption for commercial or industrial purposes
  • consumption by non-profit organisations
    

Classes of consumption relating to residential tariffs have been harmonised across the DBs. This project has used energy consumption data defined as residential by each individual DBs.

In addition, many DB tariff definitions do not align well with data items, standards and definitions used by other statistical providers. This makes detailed analysis and the creation of composite measures very difficult. For example, the System of National Accounts (SNA93) supports data discoveries about the household sector in an agreed, documented and consistent framework. However, in an ideal statistical world, energy customers who have a residential tariff should not be equated to metrics for the household sector under SNA93, as the terms residential and household do not apply to the same populations.

Negative consumption

Negative entries corrected customer record summary data for billing purposes, it is likely that they were summed 'out-of-context' mathematically and therefore have affected data quality over subsequent reporting periods.

However, should energy consumption data continue to be requested at the postcode level, the algorithms used to compile the observations in any year would need to account appropriately for these negative values.

Creating Composite estimates

The term 'composite estimate' refers to any statistic that has both an observed volume of consumption and a count of the unit values for a specific type of consumption (e.g. total volume of energy per person/ per customer/ per property, etc.).

Estimates such as dwelling counts or households from other sources cannot be readily substituted for the counts of customers or properties derived by gas businesses. In some urban fringe municipalities, counts of residential gas customers are considerably lower than the total number of private dwellings or occupied private dwellings. This is to be expected (e.g. some of these private dwellings are not connected to the reticulated gas supply and use electricity exclusively or substitute other fuels e.g. wood or LPG tank/bottled gas) but it does mean you would arrive at a less accurate answer if you used the count of total occupied dwellings to calculate a per customer figure.

In other municipalities, counts of residential reticulated gas customers over-estimate the actual number of dwellings. This may occur because of double counting (e.g. a residential property that experiences a change of tenancy within a billing period), however this is only likely where properties are counted by account number (which will change with each successive occupant) rather than meter installation number (which will remain static unless the meter is replaced).

Estimates more frequent than annual

Obtaining energy consumption estimates for periods other than annual is difficult. Billing periods vary across the DBs and therefore it is difficult to observe a 'normal' number of billing days for a consistent periodic observation of consumption.

The following example illustrates the ramifications of varying billing periods on the usefulness of energy consumption metrics.

Example - The energy meter for a property in Postcode 1 is read on the first day of October, with the consumption observed relating to the previous 88 billing days. The energy meter at another property, in Postcode 2, is read on the last day of December, with the observed consumption also relating to the previous 88 days. While both records relate to a period of the same length, the observations actually refer to consumption in almost mutually exclusive time frames and with potentially very different climatic conditions.

While it is likely that a specific property has its meter read at similar spaced time frames throughout the year, there may be considerable lag between the period for which consumption is recorded and the period that the consumption actually took place, especially for metrics more frequent than annual. Any significant change to the way in which billing data are collected from meters may have longer term impacts on apparent trends.

Seasonality

Energy consumption estimates are directly influenced by a range of seasonal factors, particularly changes in temperature.

Top of page