Executive summary

Recent studies by Statistics Canada have focused on the importance of publicly owned infrastructure in the Canadian economy. Publicly owned infrastructure has also gained the attention of economists in other countries, who have asked how it supports growth and prosperity.

Despite the interest in public infrastructure, most papers in this area do not explicitly justify which assets should be called infrastructure or why a portion of infrastructure should be called 'public.'
 
Statisticians collect data that are meant to help inform debate. Classification is fundamental to this process of data collection. This paper asks whether there is a well-defined and defensible taxonomy that statisticians can use to organize their collection efforts to provide data on the capital that is invested in public infrastructure in Canada.

This paper first outlines problems in the present debate. It then proposes a taxonomy that can be used to assess the importance of different types of capital investments that researchers may consider infrastructure or the portion of the total that should be considered 'public.'

The paper starts by developing a classification system for the collection of data on infrastructure. While there is widespread agreement, in the existing literature on the functions of infrastructure, that it provides an important foundation required to support private investment and economic growth, quality of life and security, there is no universally accepted definition of what it actually is.

The second section of the paper proposes criteria for deciding which assets are infrastructure, and it summarizes the arguments that have been made as to why some infrastructure might be classified as 'public.'

When analysts do attempt to create a structure to help them define infrastructure, they commonly focus on functions of the products produced by industries in which infrastructure assets are located. We choose not to proceed in this direction, arguing that it is problematic for two reasons. First, it is difficult to obtain agreement on which functions are 'foundational' in the infrastructure sense. Lists that attempt to do so generally fail to establish any clear criteria as to what should be left out or what should be included. Definitions that refer to 'the public interest' leave the analyst with having to list areas where the public sector expends resources or where it legislates in the public interest. These areas vary, country by country and jurisdiction by jurisdiction, and therefore do not provide universal standards.

Instead, we define infrastructure as a set of assets possessing certain characteristics that give rise to specific economic problems that are pervasive across different economies. This approach relies on the economics of contract theory. In some situations, contractual problems lead to imperfect market outcomes like monopolies, which develop because of the characteristics of the production process that, in turn, are related to the nature of the capital assets that provide the foundation for production in these industries. In addition, the economic theory of contracts can be used to guide us in the choice of the characteristics of assets that create problems for the economic system — that either make for difficult and perhaps imperfect outcomes in the market system (natural monopolies) or that lead to public intervention.

We rely on certain physical characteristics of the assets (other than ownership) to define which assets should be classified as infrastructure. While we believe that clear criteria exist in this area for defining infrastructure assets, we conclude that coming up with convincing criteria that can be used to draw a line between what is private and what is public is difficult and probably brings a statistical agency into territory where it is inappropriate to tread. Furthermore, even if the distinction were clear in theory, existing data sources do not provide enough detail to draw it in practice. We thus construct a functional classification system that defines infrastructure but that will allow users to make their own judgements on where to draw the line between private and public infrastructure.

This paper defines infrastructure as a set of fixed structures with these specific characteristics — that have long useful lives, whose creation involves a considerable gestation period, that have no good short- to medium-run substitutes, that underpin the production of a flow of services, and for which it is difficult to maintain inventories. These assets also have a special foundational role, supporting other factors of production.

Fortunately, the existing collection systems of Statistics Canada already use a taxonomy that enables us to identify assets with these characteristics. They fall under the general category that we refer to as fixed structures — consisting of engineering structures and buildings; however, determining which of them have a foundational role is more difficult. Determining which of the structures have qualities that make them public in nature is an even thornier task. The latter requires an analysis of why the commercial sector cannot resolve the contractual problems associated with infrastructure assets that require public intervention to guarantee adequate supply. Since that is not the task of statisticians and its importance differs jurisdiction by jurisdiction, we forego any attempt to do so herein. We restrict ourselves to highlighting the main features of infrastructure assets in general and classifying each of the engineering and building assets as to whether they belong to infrastructure or not. We do not go beyond a discussion of the rationale used by others to argue that these assets may also be considered to be public in nature.

Since the characteristics associated with infrastructure assets are also often associated with government intervention, the fact that most structural assets attract government intervention suggests that many of these assets are candidates to be included in a definition of public infrastructure for this reason. That is, these assets do indeed have the characteristics that lead to contract failure — as evidenced by the amount of government intervention in these areas (see Baldwin 1989). Infrastructure tends to be long-lived, and thus takes a long time to build. It is inflexible in its use, and the services it produces have a low elasticity of demand. Their essential nature and their inflexibility often lead to contracting problems that create a demand for public sector ownership or involvement.

But although public sector involvement in some form is often a feature of public infrastructure, it is not a necessary or sufficient condition to designate an asset 'public infrastructure' as a result of contract failure. The government often owns or regulates for reasons other than the ones we have stated. Furthermore, since public sector involvement is often provoked by difficulties in writing contracts between private parties, it will depend on institutions (especially legal systems) that differ across countries and over time. Thus, in some countries, governments may be actively involved in certain sectors over certain periods of time; in other countries, governments take a hands-off approach to the same sector because the commercial sector is able to resolve contractual problems in the infrastructure arena. The physical characteristics that define infrastructure assets are only associated with contractual 'problems,' not with universal market failure.

We believe that a definition of infrastructure must be consistent, across jurisdictions and over time, in order to be useful. Therefore, we do not rely upon public sector ownership or government involvement in determining which assets are public infrastructure. Instead, we rely on the characteristics of the assets themselves that determine whether they are infrastructure.

We construct a dataset on infrastructure assets for Canada from 1961 to 2002, and find that the overwhelming majority of tangible, produced structural assets are located in the public sector and in a handful of other economic sectors — electrical utilities, transportation, communications and water. These commercial sectors produce outputs that either are, or traditionally have been, the subject of government regulation in terms of their quantity or price.

As our data show, whether one uses the assets, or the sectors in which they are found, to examine and define infrastructure, one ends up with a roughly similar picture. There are a small number of three- and four-digit North American Industry Classification System (NAICS) industries that have most of the assets that can be characterized as infrastructure. In fact, the only decision that has a significant impact on the data is whether to consider only public sector assets, or whether to consider assets in the small number of commercial sectors where these assets are concentrated. We argue for the latter measure, for reasons of cross-country comparability and time consistency cited above. However, we provide enough detail in our dataset to allow for alternative interpretations. We prefer the asset route because it allows us to provide a picture of the importance of infrastructure across the entire economy, not just in those industries where most of the infrastructure is found.

Another useful feature of the method used here is that most of the engineering construction assets and many of the buildings can be classified by function. In this paper, we adopt a specific mapping that allows us to describe which of the assets are devoted to transportation; communication; electricity; waste, water and sewage; natural gas and oil distribution; public safety; defence; education, recreation and culture; and health and social protection. This functional classification system is derived, with minor modifications, from a mapping that we propose from asset types to the Classification of Expense by Function of Government (COFOG) that permits us to examine infrastructure, not just by industry classification but also by purpose in the economy. Transportation, electricity and water make up most of the engineering infrastructure outside of mining.

The functional breakdown has two applications. It allows us to ask whether the patterns displayed by infrastructure are a function of changes in the economy, or of changes in specific sectors. It also allows us to specify different relationships between different types of infrastructure. It is likely, for example, that transportation performs a different function in the economy than sewage treatment. Distinguishing between the roles of the various types of infrastructure should thus help empirical enquiries into the nature of infrastructure's role in supporting other factors of production.

The paper's main purpose is to propose a rigorous definition of infrastructure and to provide the facts on the amount of infrastructure in Canada. It does not address the adequacy of the present state of the infrastructure stock. But it does examine the evolution of Canada's infrastructure capital since 1961. In doing so, it examines the role of infrastructure relative to general economic activity. It is important to recognize that infrastructure is just one of several types of capital and to ask how changes in infrastructure compare with changes that have occurred in other factor inputs (capital and labour). In this regard, the paper compares trends both in the commercial and the public administration sectors to see whether use of infrastructure differs in the public and private sectors.

Canada's total capital stock was roughly steady relative to output until the mid-1990s, after which it declined slightly. The drop was led by a relative decline in engineering construction, which stopped growing in the 1980s. The output-to-capital ratios of building construction remained roughly steady while that of machinery and equipment increased. In combination, the relative stability of total capital, along with increases in the shares of machinery and equipment and declines in engineering capital, are consistent with changes in relative prices over time that have made the former relatively less expensive than the latter and has led to substitution among different forms of assets. That this substitution also occurred in the United States indicates that similar forces were at work within the two economies.

The study also compares Canada's long-term infrastructure growth history with that of the United States. The period from 1960 to 2000 divides into two approximately equal periods. Prior to 1980, gross domestic product (GDP) and capital grew more quickly in Canada than in the United States. The relative performance in Canada slowed thereafter. But the slowdown in spending on structures was only part of the overall slowdown. The spending on machinery and equipment in Canada slowed even more than that on structures.

Since changes at the level of the economy can hide differences within sectors, we also examine the changes that occurred within the public sector and certain other infrastructure industries.

The relative decline in engineering construction in the aggregate economy is not — —with the exception of the transportation sector — due to the relative decline in the importance of the goods and services these sectors produce, but because almost all these sectors are using less engineering construction to produce their output.

The within-sector decline could be due to one of three reasons. The first possibility is that these sectors have invested too little, and are producing higher output by working their infrastructure too hard, thus wearing it out. The second possibility is that firms have substituted other types of capital (i.e., machinery and equipment) and labour for infrastructure capital. The third possibility is that the firms in these sectors are making better use of their infrastructure capital. It could mean that these firms have found ways to use their existing capital better. Alternatively, they may in the past have built excess capacity in anticipation of higher future demand.

Given that the downward trends in infrastructure-to-GDP ratios have persisted over a long period of time (almost 20 years) and are consistent across infrastructure sectors, the first possibility seems unlikely. Moreover, the Canadian trends were broadly duplicated in the United States. A more plausible explanation is that firms in these sectors have become more efficient in using their capital, either because they started from a position of over-capacity or because they have increased their productivity.

With regards to the public sector's infrastructure, its importance in the economy (infrastructure capital relative to total GDP) has been declining steadily since the early 1970s, in spite of the increase in the public sector's share of output until 1993. This trend can be found in both Canada and the United States. This is due to an increase in expenditures on labour in the public sector. In a sense, public administration is becoming more of a service sector, and requires relatively less infrastructure capital and more labour services to produce the output that is being demanded of it.

While infrastructure capital has been declining relative to other forms of capital, it has not declined in relative terms, compared with the population that it is servicing. Indeed the amount of public sector infrastructure capital per capita has been relatively constant over the last 20 years.

In conclusion, infrastructure assets not only have characteristics that differentiate them from other types of capital, but their usage patterns distinguish them from other assets. They are highly concentrated in a small number of sectors that are either non-commercial or where there is either regulation or substantial public ownership. Moreover, the growth pattern of these assets has followed quite a different trajectory than other assets over much of the last 40 years. Whether in the commercial or the public administration sector, these assets have grown considerably slower than other assets.

View the publication Infrastructure Capital: What Is It? Where Is It? How Much of It Is There? in PDF format.