Non-market output measures in the Australian National Accounts: a conceptual framework for enhancements, 2020

Jason Annabel (Economic Statistics Research Section)¹

The Australian Bureau of Statistics (ABS) is undertaking a program of research and analysis to enhance measures of non-market economic output in the Australian National Accounts². The central measurement problem is a lack of meaningful prices for non-market output. This paper outlines the approaches considered and the preferred approach of using costs of production data as a proxy for prices. This allows quality change over time to be captured but only implicitly. National Statistical Organisations (NSOs) accept that using costs as a proxy for prices is an acceptable solution to a difficult measurement problem.

Feedback on this paper is welcomed. To provide this feedback, or for further enquiries, please email economic.research@abs.gov.au.

Non-market output occurs when goods and services are provided free of charge, or at prices which are not economically significant. These are prices that have no significant effect on the amount that producers are willing to supply and the amounts purchasers wish to buy³.

The ABS examines each producing unit in the economy to determine whether it charges economically significant prices for its output. Units that predominantly charge prices that are not economically significant are classified as non-market producers.

While most industries in Australia contain only a small proportion of non-market production, the industries of Education and Training, Health Care and Social Assistance, and Public Administration and Safety contain significant amounts of non-market output, particularly through the activity of public schools, hospitals, universities, and the provision of general public services.

Market-based producers also operate within the industries mentioned above. For example, non-government schools, private universities and private hospitals are considered to produce market output, as do general medical practitioners, medical specialists, and other similar producers of health services.

When the non-market and market components of these industries are combined, they contributed around 18% of Australia’s gross value added (GVA) and provided 3,818,400 jobs in 2017/18⁴⁵. However, the relatively disproportionate contribution of non-market activity for these three industries creates difficulties with measuring their productivity. As a result, the ABS currently excludes all activity in these industries from published multifactor productivity statistics.

Improving measures of non-market output within these industries is therefore important as it accounts for a significant portion of GVA and also moves the ABS one step closer to being able to measure productivity within them. This, in turn, would help governments to assess the ‘value for money’ received through the provision of public services.

Funding was provided to the ABS to enhance measures of non-market output. This funding was used to compile experimental indexes for measuring growth in non-market output volumes for the health and education industries⁶. The details of these experimental indexes are being published as separate ABS research papers.

The experimental indexes aim to be: 

• robust and sustainable;
• not directly based on input costs;
• appropriately weighted; and
• reflective of innovation, technical progress and quality change over time.
   

When the full spectrum of products produced in an economy are aggregated to calculate total output and GVA, relative prices determine how much weight to assign to each product.

The forces of supply and demand interact with each other in open marketplaces. Under competitive market conditions, prices freely adjust to clear markets⁷. When markets clear, the price of each product reflects the value market participants place on it. If an apple costs twice as much as a banana, it reflects the fact that the market, through the interaction of supply and demand, has decided that apples are twice as valuable as bananas. In this example, each apple produced would be ‘worth’ two bananas in aggregation, because the relative price of apples is twice that of bananas. Such economically significant prices are the mechanism through which markets clear.

When an institutional unit disposes of the majority of its output at economically significant prices, it is considered to be a market producer. Corporations (whether privately or publicly owned, financial or non-financial) and not-for-profit institutions that effectively charge market prices are considered to engage predominantly in market activity.

In contrast, non-market output is distributed to final consumers free of charge, or at prices that are not economically significant. These prices are typically well below the prices that would clear a given market, and we cannot assume that prices for non-market output are an accurate reflection of value. This lack of price information for non-market output presents valuation and aggregation problems, and alternative compilation methods must be found.

While the delineation between market and non-market output is clear in concept, implementation is not straightforward. An assessment must be made for each statistical unit within the economy to determine whether it charges economically significant prices.

The 2008 System of National Accounts (SNA) is not prescriptive in terms of how to assess whether a unit is a market or non-market producer, and thus the rule for determining the boundary between market and non-market activity in practice is not clearly defined. It indicates that a possible rule of thumb is to compare the value of output sold to production costs incurred, and if the value of output is less than half the value of production costs over a number of years, the unit is a non-market producer. Non-market producers are those classified to the general government and non-profit institutions serving household (NPISH) sectors.

In Australia, the framework used to classify producing units as market or non-market is the Standard Institutional Sector Classification of Australia (SISCA)⁸. The primary indicator of a market producer is the expectation of the recovery of a considerable proportion of its production costs through sales income, though this is more nuanced for public sector units, which are units under the control of government.

The market/non-market assessment for public sector units is based on a manual review of the extent of cost recovery, extent of competitive pressure and the extent to which public policy influences underlying market forces. For the majority of private sector units, the assessment is based on an algorithm which considers the type of business entity, income tax exemption, charitable entity status and industry. Large, complex and economically significant private sector units are assessed manually using the cost recovery principle.

The 1993 SNA was the first edition of the System of National Accounts where the complexities of measuring non-market output were considered in detail. Up until this time, the majority of non-market production was measured by assuming that outputs were equal to the inputs consumed in producing it.

In 2001, the ABS commenced the development of basic volume-based measures of non-market output for the health and education industries, and published research around experimental output measures for the justice component of the public administration industry in the same year⁹. Incremental improvements have been implemented since then.

In 2003, the UK Office for National Statistics (ONS) commissioned a review to “advance methodologies for the measurement of government output, productivity and associated price indices in the context of the National Accounts”¹⁰. The intent was to design a public services measurement framework that allowed the quality of the estimates to be assessed. The Atkinson Review was based on a number of principles. These included that:

• Measurement of non-market activity should be as closely aligned as possible to that used to measure market activity (so that if economic activity moves from market to non-market and vice versa, there is no impact on headline economic aggregates)¹¹;
• Non-market output should be adjusted for quality change;
• Measurement of inputs and outputs should be as comprehensive as possible; and
• When measuring non-market productivity, independent evidence on productivity growth is required, rather than just assuming it equals the ratio of output growth to input growth (this is the principle of ‘triangulation’).

The Review primarily focused on enhancing measures of output for non-market services that are delivered and consumed individually. Examples of these types of services include the delivery of a course of education, or the performance of a surgical procedure. Some non-market services are consumed collectively and simultaneously by the entire population. Examples of these services include public policymaking and national defence. Measuring growth in the volume of collective services presents an additional range of conceptual challenges which are outside the scope of this paper.

The Atkinson Review is considered to be one of the primary reference documents for national statistical offices in enhancing estimates of public sector economic activity, and is a key resource for the ABS. The Review further notes that designing direct measures of non-market output requires considerable care and time, and it may take some experimentation to find a measure that is appropriate for National Accounts implementation.

The 2008 SNA extended previous national accounting frameworks by articulating three options for measuring volumes of non-market output. It is important to note that the focus of these measures was to capture the change in volumes over time; current price levels are derived elsewhere within national accounting systems. The three approaches proposed in the 2008 SNA are:

• Option 1: Output volumes are calculated independently of input volumes, either by (a) estimating ‘pseudo’ prices for non-market output, or (b) explicitly adjusting output to reflect quality change from the consumer’s perspective. (These two options are conceptually equivalent.)
• Option 2: Output volumes are calculated independently of input volumes, by deriving aggregation weights from production cost data. Whereas Option 1 explicitly captures quality change from the consumer’s perspective, Option 2 implicitly reflects quality change as measured from the producer’s perspective.
• Option 3: Output volumes are equal to input volumes.

While there is currently no viable alternative to using Option 3 for measuring non-market services that are consumed collectively, the SNA recommends Option 2 for measuring those that are consumed individually¹². This is because under Option 1, there is:

• a lack of meaningful prices for non-market output and the complexity/subjectivity in estimating ‘pseudo’ prices; and
• difficulty in measuring quality change from the consumer’s perspective by any other means.

In addition, Option 1 contradicts a stipulation found elsewhere in the SNA that “by convention no return to capital on assets used in non-market production is included when output is estimated as the sum of costs”¹³. It follows that the net operating surplus for non-market activity must always be zero. Estimating ‘pseudo’ prices and/or applying explicit quality adjustment permits a non-zero net operating surplus for non-market activity, which would imply a return to capital.

Ideally, national accounts statistics should capture non-market activity on exactly the same basis as market activity, but the stipulation that non-market production carried out by the public sector cannot give rise to a return on capital makes this impossible. To address this issue, some countries follow the convention on reflecting a zero net operating surplus for non-market activity within their national accounts aggregates, while allowing a return to capital when compiling estimates for productivity measurement.

Option 3 is undesirable for volume measurement because it does not lend itself to measuring productivity growth. A further drawback of this approach is that technological innovation that facilitates reductions in expenditure will lead to a fall in measured output, when it is possible that only inputs have fallen.

Diewert (2010) expanded the three measurement options outlined in the 2008 SNA into nine possible measurement ‘cases’, which are depicted in Diagram 1 and explained in detail below. Some of these cases are conceptually pure but impractical to implement, while others are easier to implement but less conceptually ideal. The preferred ABS approach seeks a balance between conceptual purity and practicality of implementation.

Cases 1 to 4 rely on meaningful price information that either exists, or can be estimated, for non-market output. That is, where prices do not exist, ‘pseudo’ prices are calculated from the consumer’s perspective.

Case 1: Price and quantity information on outputs is available.

In theory, this is the simplest possible case and is conceptually equivalent to the approach used to measure market output. It requires observable market prices for the non-market output being measured, but this presents a conundrum: if there were observable market prices for the output being measured, it would no longer be non-market output. (Equivalent to Case 1 is the scenario where quality metrics measured from the consumer’s perspective are explicitly infused into the output volume index.)

Case 2: Quantity but not price information is available, however comparable market sector prices are available.

This differs from Case 1 in that while there are no prices for non-market output, ‘comparable market sector prices’ can be observed. These observed market prices are then used as aggregation weights in calculating non-market output. This assumes market-based entities are supplying output which closely resembles that being provided through non-market activity, which is an unlikely situation to find in practice.

Case 3: Quantity but not price information is available, and somewhat comparable market sector prices are available.

This differs from Case 2 in that prices for non-market output are ‘somewhat comparable’. Output provided by market-based entities and non-market entities are not closely related (for example, there may be a noticeable difference in quality between the market-based output and the non-market output), but they remain ‘somewhat comparable’. In this case, ‘pseudo’ prices for non-market output are estimated using market prices, where the estimated prices reflect the consumer’s willingness to pay for the non-market output. While this can be attempted via hedonic regression, there are a number of practical and conceptual issues involved, meaning the results can be subjective and potentially controversial.

Case 4: Quantity but no price information on outputs is available, and there are no comparable market sector output prices.

This differs from Case 3 in that it described an attempt to estimate ‘pseudo’ prices for non-market output where no ‘comparable’ market prices exist. The suggested tool is a general equilibrium regression model. Very detailed information on the rest of the economy, including consumer utility functions, is required. Again, this approach is likely to be subjective and controversial.

Note that Cases 1 to 4 would present the possibility of a non-zero net operating surplus for non-market activity, which as noted earlier, contradicts a stipulation found elsewhere in the 2008 SNA that returns on capital cannot be accrued through non-market production.

Cases 5 to 8 assume that price information for non-market output does not exist and/or cannot be reliably estimated from the consumer’s perspective, and cost information is used as a proxy. This approach generally relies on obtaining high-quality administrative datasets.

Case 5: Quantity data exists, and average cost data are available for both periods. No information exists on quality change.

This differs from Case 4 in that there are no meaningful prices that can be observed, nor can they be reliably modelled. Information on costs of production is available, and is used instead. ‘No quality change’ means quality change is not explicitly measured via a separate vector of quality variables. Instead, it is implicitly measured via changes in production costs (and cost differentials) over time.

Case 6: Quantity data exists. Average cost data is not available, but other cost estimates (for example marginal costs) are available. No information exists on quality change.

This differs from Case 5 in that complete information on production costs doesn’t exist, but other types of cost data are obtainable.

Case 7: Quantity data exists, and average cost data is available for both periods. Data on quality change, measured from the producer’s perspective, exists.

This differs from Case 5 in that full cost information is available, as well as a vector of variables that describe the quality characteristics of the output. The volume index is a composite measure of cost-weighted quantity growth that is explicitly adjusted for quality change, though the quality metrics are measured from the perspective of the producer and not that of the final consumer.

Case 8: Quantity data exists. Average cost data is not available, but other cost estimates such as marginal costs are available. Data on quality change (measured from the consumer’s perspective) exists.

This differs from Case 6 in that a vector of variables that describe the quality characteristics of the output exists, but again, this is measured from the producer’s perspective.

Case 9: There is no detailed information on output values, prices or costs.

In this case, non-market output is measured as an aggregate of the inputs consumed in producing it. Because output volumes grow in lockstep with input volumes under this approach, it is not possible to measure productivity growth.

The current approach to output measurement in the Australian National Accounts is a combination of Cases 5 and 9.

For the foreseeable future, non-market output that is consumed collectively by the community (such as national defence services) will continue to be measured as equal to the inputs consumed in producing them (Case 9). For the remainder of non-market activity, the ABS is working towards direct output volume measurement as described in Case 5. In addition, where some kind of direct volume measure is already in place, these measures are being reviewed and enhanced, as data permits.

Volumes of market-based economic activity are generally calculated indirectly via price deflation. When outputs and inputs measured in current price terms are deflated via appropriately defined and weighted price deflators, and where prices are measured to constant quality, the volume growth measures reflect quality change as well as quantity change.

Measuring non-market activity in the same way is difficult because prices do not have the same meaning. This, in turn, makes it difficult to calculate volumes indirectly through price deflation, because the lack of prices makes it difficult to calculate appropriate deflators.

An alternative, to create deflators from input costs such as wages, is generally considered to be inappropriate because it deflates the output not by its own value, but by the value of the inputs used up in producing it. Another alternative is to estimate ‘pseudo’ prices that reflect the consumer’s valuation of the output, but this approach generates conceptual and measurement problems that have been described earlier.

Volume measures can also be calculated directly via an appropriate set of quantity indicators, where the relative values of products serve as aggregation weights. Direct calculation of volumes through the application of quantity indicators is the general solution for measuring non-market volume growth.

For example, an appropriate indicator of quantity for school education might be an hour of teaching services delivered, whereas quantity indicators for hospital output might be the number of surgical procedures undertaken. But we cannot just add up the number of hours taught or the number of surgical procedures performed, because the outputs provided by a school or hospital can have different characteristics. Hours of teaching for different levels of education are not the exact same type of service, and there can be wide differences across the spectrum of surgical procedures delivered by hospitals.

Two essential pieces of information are required: indicators of quantity (for example, the number of surgeries performed in a given time period); and indicators of cost (for example, how much each surgery costs to perform). Costs are intended to reflect the marginal value placed on the services performed. These datasets need to be disaggregated to the same level of classification so that the cost and quantity data can be matched. Ideally, the classification will ensure that ‘like’ services are grouped as homogeneously as possible and classified as narrowly as possible.

The level of differentiation between services is also important. Rather than looking at the totality of the service provided, the intention is to define a set of individual products, to which cost weights are applied. Hospital admissions data is highly detailed, and statisticians are able to measure substitution between services over a range of disaggregations. Finding the right level of detail requires experimentation and reflection. If the dataset is not sufficiently stratified, there is a risk that finer details of substitution may be missed – using a total quantity metric can generate misleading results. By contrast, disaggregating too deeply runs the risk of introducing ‘noise’ into the statistics.

Combining quantity and cost information attempts to capture the impact of technical progress, which leads over time to improvements in service quality. Existing services are transformed and improved over time, new services appear and old ones disappear, and this process is implicitly captured through changes in relative costs¹⁴. This occurs because changes in cost relativities over time highlight the rates of substitution across the spectrum of services produced. The drawback is that quality change is only ever captured from the supply side.

There are a number of further extensions to be considered in measuring non-market activity. Two of these are addressed below.

There is extensive discussion in the national accounting literature on the difference between measuring outputs and outcomes¹⁵. Outputs are goods and services that result from processes of production, whereas outcomes are situations that are valued by consumers.

In terms of health care, outputs consist of the provision of medical care through processes such as hospital treatments and medical consultations, whereas desired outcomes from the consumer’s perspective might include properties such as longer life expectancy, faster recovery from surgery, and shorter waiting times.

For education, output is the organised communication of knowledge from teacher to student within schools and universities, whether delivered face-to-face or remotely. Outcomes can include improvements in examination scores, stronger intellectual engagement, and enhancements in estimated lifetime earning capacity.

For education, the range of economic concepts of interest include: 

• Inputs: labour, capital, goods and services
• Productive activities: classroom teaching, setting and marking homework and examinations
• Outputs: education services delivered
• Outcomes: a higher level of education

For health care, the equivalent examples include:

• Inputs: labour, capital, goods and services
• Productive activities: surgical procedures performed
• Outputs: health care services delivered
• Outcomes: enhanced health status

Professionals in the education and health care industries care deeply about improving the outcomes of the services they provide, as do students and patients. But there is no consensus in the national accounting community on how to estimate outcomes for non-market production beyond implicit incorporation of quality as described earlier. In particular, there is little agreement on which specific aspects of quality should form part of an explicit total quality measure.

For example, metrics that could be used in assessing changes in the quality of hospital output over time could include changes in waiting times for surgery, changes in patient recovery rates, changes in life expectancy, changes in patient quality of life, changes in the technical knowledge of health care professionals, and changes in the rate of surgical failures (readmissions and deaths). It is far from clear which of these criteria should be included in a total quality measure, how best to measure them, or how to weight them together.

It is also possible that indicators of outcomes might not be entirely driven by, or even moving in the same direction as, outputs. For instance, the use of standardised test scores to assess the quality change of education over time does not account for the possibility that some cohorts of students may be more naturally gifted or dedicated than other cohorts, meaning test scores can vary over time in ways that have nothing to do with the quality of education services produced. Similarly, in the case of health care, improvements in rates of surgical successes and recovery times could be related to the possibility that the community is becoming fitter and healthier over time, a property which is unrelated to the quality of health care output. Extrapolating this point, a fitter and healthier community may have less cause to demand hospital services: an improvement in health outcomes could give rise to a decrease in the quantity of health output. (Contemplate the extreme case where a patient dies on the operating table: output has still been produced by the hospital, even though the outcome for the patient is tragic.)

A number of National Statistics Offices (NSOs) have compiled some experimental outcomes-based measures for non-market activity, and the United Kingdom has formally incorporated some outcomes-based measures into their productivity statistics.

The lack of prices data makes it difficult to measure productivity growth for non-market output¹⁶. It is important to measure productivity growth, because over the long term, productivity is commonly seen as the main way to lift real incomes over the longer term and hence to improve Australia’s living standards. It is particularly important to measure productivity in the case of health care, where large-scale technical progress has occurred and is ongoing.

To estimate productivity growth for a particular industry, sufficiently detailed measures are needed for:

• inputs – labour, capital and intermediate inputs such as energy, materials and services used up in the production process; and
• outputs – the value of goods and services produced.

The current ABS research into non-market production is focused on enhancing measures of outputs.

Diewert (2008) published a mathematical proof showing that using relative costs as aggregation weights allows the measurement of output growth independently of input growth, and hence productivity growth becomes measurable.

The proof assumes that each type of output is produced with perfect allocative efficiency (there is no ‘waste’, such that costs are a true reflection of the resources required to produce the output). It also assumes constant returns to scale (so that average cost equals marginal cost), and that technical progress exists – in other words, the methodology allows for new processes and technologies to be introduced that deliver the same services using fewer resources.

It is unlikely that the assumptions around allocative efficiency and constant returns to scale hold true in reality for non-market activity. Issues such as the principal-agent problem, X-inefficiency and the fact that ‘political’ considerations can impinge on the efficiency of public sector output must also be considered¹⁷ ¹⁸.

In addition, the stipulation in the 2008 SNA that capital deployed in non-market production cannot earn a return needs to be considered when interpreting productivity measures. Assuming this stipulation remains unchanged, and following the approach taken in the UK, productivity estimates for non-market industries would need to break this stipulation, thus placing them on a different conceptual footing to the rest of the national accounts.

Notwithstanding these considerations, the ABS will continue to discuss the value of feasibility of such measures with stakeholders

This paper has outlined the approach the ABS is taking in developing enhanced measures of non-market output volumes for the Australian National Accounts. This sets the scene for a number of papers to be published over coming months outlining new experimental output volume indexes for schools, hospitals and universities.

The ABS is also planning to review the inputs of labour, capital and intermediate goods and services used by schools, hospitals and universities to produce their output. This would facilitate the derivation of experimental measures of multi-factor productivity for the first time, and it is envisaged that a progress paper on this work will be published by June 2021. The ABS may also look at other areas of non-market output, such as the delivery of ambulance services.

The ABS will be seeking the input of a range of external stakeholders as work proceeds. Feedback on this paper is welcomed. To provide this feedback, or for further enquiries, please email economic.research@abs.gov.au.

This release previously used catalogue number 5900.0.00.000