|Date||February 23, 2016|
|Speaker||Paul SCHREYER (Deputy Director, Statistics Directorate, OECD)|
|Moderator||FUKAO Kyoji (Program Director, RIETI / Professor, Institute of Economic Research, Hitotsubashi University)|
|Time||12:15-13:30 (Registration desk and seminar room open at 12:00)|
Recent years have seen several broad economic shifts. One is globalization, with the accompanying interconnectedness of economies. Others are the emergence of digitalization and the growing importance of knowledge assets, which transform economies and shape competitive advantages. Many OECD countries are facing aging societies along with issues concerning financing of pensions and health systems. There are also increasingly broader quests for sustainability in economic, social, and environmental areas. All of these factors affect productivity measurement.
Globalization and interconnectedness in terms of value added can make it difficult to measure national value added (GDP) and productivity. Furthermore, there are new business models driven by digitalization that are sometimes hard to capture
Aging societies give rise to questions regarding how to measure health services. Sustainability requires measurement of capital—economic, environmental, and social—and a host of measurement issues arise.
Productivity is the ratio of outputs over inputs. In regard to measurement, it is useful to make a distinction between goods and services and then between market services and nonmarket services. A number of market services are hard to measure and one example is financial services. What unit of service do banks provide and how can this be measured in the accounts? Communication services, where rapidly changing goods and services are bundled, is another example. In addition, new business models and forms of utilization are being adopted.
In nonmarket services, health, education, and social housing are the three most important nonmarket services. Here, measurement of values and volumes is complex because there are no observable prices and no transactions that can reflect the whole value of the services being used.
Regarding inputs, a distinction is typically made between labor and capital. Labor quantity is typically measured as hours worked. We are reasonably well-equipped to measure this across the economy even though there are now cases where the distinction between consumers and producers is blurred—consumers can now check out their own goods at a supermarket, meaning they are actually providing the labor, and that does not appear in the business accounts. Labor quality is more difficult to gauge but is important as it is not appropriate to give the same weight to one hour of unskilled labor as one hour of skilled labor. To adjust for these quality differences, data on skills distribution and availability of compensation across the workforce are needed.
It is also useful to break capital down into two types. The first type consists of the produced assets included in national accounts: machinery, equipment, structures, and inventories. These assets are measured fairly well. The second type of assets composes the natural, non-produced assets that are part of balance sheets but are not often measured. The two biggest forms of these are subsoil assets, such as minerals and oil, and land. Land is a very important non-produced asset that is very poorly measured, in particular, in terms of value.
Beyond the asset boundaries in the national accounts is knowledge capital. This is more elusive, but is recognized by the business community as important for competitive advantages.
In OECD countries, health services account for a comparatively large share of GDP. In some countries they account for 10%, while in the United States, they account for around 16%. Therefore, any improvement or mistake made in measuring health carries over to GDP with a fairly big weight.
What is the accounting specificity of healthcare? Health is often provided by nonmarket producers, so a different process is needed to measure the value of nominal output. Revenues observed for market providers are not observable for nonmarket providers, so the value of output is estimated as the sum of intermediate inputs, wages and salaries, consumption of fixed capital, and taxes. A rate of return is not included because the national accounts do not recognize an imputation for a profit margin in the case of nonmarket producers.
Measurement in the health sector
The traditional method is to measure the volume or quantity of outputs by looking at that of inputs. To do this, an index reflecting the hours spent by nurses and doctors to produce the health services is produced. Of course, a measure of outputs based on inputs means zero productivity growth.
We are now moving from an input-based measure of output to an output-based measure by using a unit of output, e.g., the number of treatments that are being performed by the health industry; not the number of doctors, but the number of operations and interventions. This is increasingly possible as there is a strong drive in many OECD countries toward cost accounting because healthcare costs are increasing and health administrations want to account for procedures at some level of detail.
This type of measurement provides a great source of information for statisticians. Many countries have diagnosis-related groups (DRGs) providing indications of the costs and quantities of certain procedures in the health industry—the outputs of health administration—and enabling the gathering of information on medical treatments.
A project at the OECD is standardizing treatments so as to measure prices across countries, and derive purchasing power parities (PPP) for health services across OECD countries. This program defines standardized procedures and their prices across countries. Input-based measures are compared to output-based measures. Both overestimation and underestimation of health outputs are found, depending on the countries.
There is an increasing trend among OECD countries to carry out output-based measurement also at the national level. But not all countries have moved in this direction; e.g., in Japan, Korea, Chile, the United States, and Canada, the accounts still use input-based measures, while a number of European countries as well as Australia and New Zealand already use output-based measures. This may constitute a problem when it comes to comparability across countries, but progress is being made, and these new methods are being used more. For example, in the United States, a major economic analysis program is underway to set up health satellite accounts and move to an output-based measurement of production of health services.
Over the last few years, there has been much talk about the "sharing" or "Uberization" of the economy, referring to new business models introduced and enabled thanks to information technology (IT). These models may actually make measurements more difficult and cause important activities to no longer be captured in statistics.
These transitions from traditional business models include the move from equipment rental and hotel services to peer-to-peer services, or the shift from hotel services to online booking services. These new services take place between households, enabled by digital platforms that act as intermediators.
All of this gives rise to claims that production measurement is being lost, because this business is happening outside of traditional forms.
Measurement in the context of digitalization
Framing the discussion in accounting or measurement terms, the first issue is that the intermediation function originally provided by a traditional firm (e.g., a travel agent) is now enabled by a technology platform (e.g., Booking.com).
In these cases, the actor or company providing the service has changed, and the intermediation services are now carried out by the new provider. The revenue of these platforms is not the whole turnover, but just the commissions. In principle, there is no measurement problem here; the commissions simply need to be measured. The only practical issue here is whether a platform has moved out of the country, making it more difficult to identify the revenues of the commissions that accrue.
The second issue is the service provision itself. In the past, corporate firms such as hotels provided the services, but now households provide services (e.g., accommodation via Airbnb or car rides via Uber). If there is any revenue, then it would be the transaction for the service. In principle, these services fall within the GDP measurement boundaries unless they are provided on a very irregular, small-scale basis. For example, if a household signs up with Airbnb and makes one transaction a year, that activity would probably not be entered into the accounts. On the other hand, there are activities that are regular but may be undeclared, so there is also the issue of a non-observed economy. Similarly, small-scale barter transactions would also not typically enter the measurement system.
There is also the question of quality change. These new platforms allow much more granular and real-time searches for services. For example, when looking for a room to stay in Paris, you can select an apartment exactly where you want to stay because so many households operate on a particular platform, compared to the number of hotels. That increases consumer choice, so it is a quality improvement. On the other hand, it is unclear how the services are being provided and hotels may work more efficiently and have different standards than private providers, so this may be a case of quality decrease. Customization is closely related to these more granular choices and also leads to unique products and price comparisons that are more complicated. The direction in which quality change is moving is certainly not clear.
The third element in discussions on service sector measurement consists of free or seemingly-free products, which are traded via "triangular transactions." Examples include free communication with the Skype app, free music and videos on YouTube, and other smartphone apps. None of these services have explicit prices, so they have zero weight and are not part of GDP. However, there is almost always some financing behind them in the form of triangular transactions, where the software provider gets paid, for example, via advertisements that are placed on the free service or where the data generated by an app are sold to a third party. Many other business models exist, but some revenues have to be generated somewhere, at least in the future.
So there is an implicit transaction between the user and the software provider even in the case of free products. This means that the implicit valuation of these free apps is the value or revenue of the advertising that is being sold behind them. It also implicitly means that whatever deflator or volume index used to deflate the advertising services is also the implicit price index for the free services.
Of course, this is not necessarily a good measure of the marginal utility to the app user. What is clear is that there is a business model involved, and that is not completely unnoted in the national accounts. Whether some additional imputation for production or consumption is needed is a matter of debate.
Digitalization: In summary
It is too early to make a judgement on the quality of production measurement. Certainly conceptually, the existing GDP measures appear to cope relatively well with the digitalized economy. Challenges exist in some areas of implementation such as price measurement and when digitalization involves cross-border flows of revenues. In addition, a statement on production does not automatically imply a statement on productivity, because we do not actually know whether new business models that involve more production by the household sector are more productive than those provided by the commercial sector.
As a general point, however, households and their production have become more important and are moving center stage with these new business models, and this needs to be reflected by statistical methods. Finally, it is also good to recall that the ambition of GDP is not to measure welfare. GDP is a measure of production, and there are certain production boundaries that are well defined for good reason. Not everything that generates welfare should necessarily be reflected in GDP, because it is not necessarily production.
For some time, knowledge capital has been measured, and there is an increasing range of evidence and agreed definitions. Consider the different forms of investment in the European Union and the United States. Broadly speaking, there are three types of evolutions here. The first is the evolution of the knowledge-based assets that are recognized in the national accounts. They are essentially software and research and development (R&D), and these grew more rapidly over the last 20 years than other types of investment. The second type is the evolution of traditional investment such as structures, machinery, and equipment. This is leveling off, which is a big concern for many OECD countries. The third type is the evolution of other knowledge-based assets, including organization capital and training. These fall outside the national accounts and have been increasing fairly steadily.
These different evolutions are important because technological competitiveness in many countries is bound to knowledge, especially in advanced economies such as Japan, the United States, and the European Union. The measurement of this knowledge-based capital is not obvious, and often the only way to capture it is as the cumulative sum of costs. Almost by definition, there is no market for many knowledge-based assets, especially as firms produce these assets themselves and obviously want to protect them, so there is no place where the market value of some of these knowledge-based assets can be measured.
Depreciation is also an issue. Some say that knowledge does not depreciate as it exists forever, while others argue that, as there is much obsolescence in knowledge, it does depreciate. Surveys of firms in OECD countries found that the depreciation rates for research vary greatly between industries and are quite different from zero. Plausible patterns were found: in the chemical industry, depreciation is tied to the length of patents for particular drugs, but in other industries, especially IT, depreciation is much faster.
Price indexes for investment in knowledge capital also need to be considered. Typically, inputs that go into the investment are looked at and become the price indexes, so the wages of researchers and the prices of equipment used constitute the price index. Again, this is not ideal as it is an input-based measure.
The price indexes for software from various OECD countries over the last 20 years or so form an interesting example. They rise and fall depending on the countries. There is no reason why price indexes should be the same across countries, but it is reasonable to expect similar trends across countries in a service such as software production, which is produced and sold globally, as is the case for hardware. Here, it is not clear to what extent observations are true that differences exist in the prices of software and database investment, and to what extent they simply reflect the measurement assumptions that statisticians have to make. The OECD is working on this with statistical offices to ascertain where the differences come from and if they are real or simply reflect different methodologies.
Knowledge capital is becoming increasingly important. By its very nature, it is difficult to measure, but there is much work underway to move into broader measurement of assets and see how they affect measures of productivity.
Land productivity measurement
Land concerns all industries. As mentioned above, there are two major categories of capital assets in national accounts. In one category are the traditional, produced assets, including machines, equipment, and structures. In the other category are non-financial, non-produced assets, of which the three most important are oil and other mineral resources, land, and timber.
These non-produced assets are often unaccounted for during productivity measurement. Especially, land is often left out, and data on land are surprisingly far from complete, although all countries have some physical inventory of it. All OECD countries classify land according to surface and type, but few countries have a monetary valuation of land in their balance sheets. Yet it is a very big item, and those countries that do have an evaluation of land in their balance sheet have large figures.
One example is France. It has a complete balance sheet, and in 2014, the value of land accounted for 40% of its total wealth. Korea also has set up a complete set of wealth accounts recently and came up with large figures. Of course, the value of land goes up and up, and there is much revaluation, but that is fine.
Land is not only a value item but also a form of capital that provides services in volume terms, so it should be part of productivity measures as well. Clearly, land provides capital services, and we should measure them. The quantity of land changes very little, so including land in the measure of capital input will change the overall movement of the volume of capital and typically slows it down.
The inclusion of land in productivity measurement matters, and two examples show this. The first example is Korea. Consider the rate of return on capital. Dividing profits by capital will result in a very different figure if land is included in that measure of capital. In Korea, when land is excluded, there is a steady decline in the real rate of return on produced capital, but when land is included, the rate of return is actually stable or increases slightly as there are revaluations ongoing. This makes a big difference to the interpretation of economic figures.
Measuring multi-factor productivity essentially means measuring a residual; the change in volume of outputs is considered and then the change in volume of inputs is looked at. In this case, the inputs are labor and capital. Whether or not land is included in the capital measure clearly makes a difference to the resulting measure of productivity. Again, in Korea, multifactor productivity grew by about 1.2% in the last 20 years. However, if land is included, the measured productivity growth increases, meaning that the capital input growth has been overstated by not including land in the observations.
The same reasoning applies to a country with natural resources and subsoil assets, such as Australia. Calculations including minerals in the mining industry significantly change the picture of Australian productivity growth. It is always important to think about land, because there are so many interesting facets, but land is rather neglected in our statistics.
Progress is being made in measuring services, particularly such non-market services as health, education, finance and communication services, but more work is required.
Digitalization and new business models are disruptive in terms of their economic effects, and they are a challenge from both a policy and a statistics perspective. They do, however, cast new light on household activities. While we do not know whether some measures of productivity are biased, just because some data is not yet part of GDP, it does not necessarily mean that we are incorrectly measuring production.
On the input side, ongoing work on knowledge-based capital and land should be continued and extended to non-produced assets in particular, because they are so important economically and environmentally.
Q1. In terms of efficiency moving into the household sector, what approaches would you like to see being used to get a better grip on data in the household sector? Regarding capital, in an aging and shrinking society, the value of land comes from the services the land provides, and this might be the root of the differences that we are seeing. Also, KLEMS projects already are providing better measures for capital services. What is your impression of progress in the measurement of capital services? And what do you think we should focus on now?
I agree with your observations. Regarding approaches for household information, new surveys would certainly be good, as would adopting existing surveys. However, statisticians are often reluctant to expand surveys too much, because of declining response rates and respondent fatigue. We need to gather new information through better exploitation of administrative data. Increasingly, statisticians are actively looking into Big Data and new data sources, which can facilitate a good and interesting indication of what is happening in households.
When it comes to capital, in aging societies, capital in conjunction with structures and housing is indeed mainly a measure of wealth but it is also a source of capital services. Appreciation will affect the demand for housing as the price level changes, but that is different from the quantity of capital services that are being provided from houses.
I am fully in favor of measuring capital services along the lines of the KLEMS projects. They are actually part of the national accounts now, but I think the scope of services should be extended to non-produced assets.
Q2. [Moderator FUKAO Kyoji] Asymmetric treatment, in regard to choosing whether to include real return to capital in production measurement, is related to the current discussion, and some trials are being conducted to take account of real return to capital in health services. What do you think about that?
This is somewhat of a technical issue, but, yes, the treatment of capital is asymmetric in that it depends on whether the capital is in the market or nonmarket sector. In the nonmarket sector, there is no imputation for rate of return. I am in favor of such trials considering that, for example, if a hospital were to move from the nonmarket sector to the market sector, it would suddenly have completely different treatment of exactly the same assets.
There was a big debate during the revision of the national accounts about whether to include the rate of return, but the conclusion was against the idea, simply because it would mean another imputation that directly affects the measure of GDP. European countries do not like imputations that directly affect GDP because they are directly linked to the EU's administrative procedures and affect budget contributions to the EU. That involves all manner of implications beyond research, which makes statisticians particularly prudent. From a conceptual or research point, however, I think it should be done.
Q3. Regarding the differences among countries in terms of input and output, do you think that some information on the productivity of healthcare is forthcoming, or are the differences mostly due to the measurements?
We can get more information if we also improve our input measures. The problem is actually that the input measures are not so good, either, and it is difficult to get good numbers, especially in a cross-country comparison. It is very difficult to standardize, for example, doctors' wages across countries, so we simply obtain the same level of qualifications, which is needed to construct the input-based index. In principle, if we have both input-based measures and output-based measures the difference is productivity.
Q4. Regarding health services, I think that a perfect health index would be 100%, with no health expenditure. In Japan, many people are very old, but the medical expenses are very low, and that is what we need to aim for.
Regarding land, if we were to incorporate the value of land in GDP, we would have to be careful not to double-count it. Office rent and land rent are already reflected in GDP, so if we were to add the land price or another valuation it would be double-counting.
Health expenditure is indeed not necessarily a good predictor for health outcome, and that distinction is very important. However, in the national accounts, we do want to measure the production of the healthcare industry, which may or may not produce good health. Good health depends on many different factors. Life expectancy in Japan is so high, but that is not necessarily because the medical service is so good; it is because people live healthily and have a good lifestyle. That is a big determinant of health outcomes and is totally independent of the medical industry.
Concerning land, I am not proposing that it be part of GDP. I only want it to be present in the balance sheets. In the flow measures I would like only the volume of services from land to be included.
*This summary was compiled by RIETI Editorial staff.