A look at the difficulties in writing specifications for services

Until the most recent very few decades, government purchasing was pretty much concerned with acquiring things, not services. Indeed, for the US, acquiring things was often a question of build or buy, and building was common from the earliest days of the pre-Revolution Union, particularly as regards arms and heavy equipment and transportation networks.

As an aside, the nation's use of its own armories to manufacture arms and create the machine tools for doing so, according to James F. Nagle, underwrote the industrialization of America. In his epic overview of the "History of Government Contracting", Nagle makes the point that the phrase "close enough for government work" began as a form of high praise, rather than derision as it is now used, because the government set the highest standards for precise manufacture of interchangeable parts. It's a truly enlightening book and a must read for any student of procurement.

But back to the story here, the trouble with much contracting has been in writing sufficiently detailed specifications to obtain what the government needs without unduly limiting the sources from which it can obtain it. (As Nagle pointed out, we have come a long way from the three page contract awarded to the Wright Brothers to produce the Army's first airplane.)

The problem with drafting effective specifications has been compounded by the recent trend towards a preponderance of contracting for services. As hard as it is to write effective specifications for things that can be measured for physical characteristics as well as performance qualities, the problem with writing specifications for services is almost vague by definition due to inability to objectively identify, quantify, measure and reward the subjectivity of service criteria. History shows that wherever subjective criteria are relied upon, favoritism creeps in.

Which brings me to the following economic study, which tries to bring some economic objectivity to an evaluation of municipal government contracting for services vs privatization, in-house vs outsourcing services. I get the sense that the study started with a bias toward privatization, but it is nevertheless instructive for its findings that show correlation, if not causation, between certain characteristics associated with build vs buy. It is also useful for showing how the relative difficulty in drafting specifications (and administering contracts based on them) affects decisions related to service (or supply, by extension) delivery mechanisms.

Given that the article discussed below focuses on municipal government contracting, it should have particular relevance for places like Guam and for places, again like Guam, which have adopted the ABA Model Procurement Code, a procurement model particularly intended for state and local procurement regimes.

The paper is "CONTRACTING FOR GOVERNMENT SERVICES: THEORY AND EVIDENCE FROM U.S. CITIES" by Jonathan Levin and Steven Tadelis, in THE JOURNAL OF INDUSTRIAL ECONOMICS. The authors are affiliated with Standford University and the University of California, Berkeley. Excerpts from the Introduction to the paper follow (footnotes omitted; refer to the article for more detail and explanation):

The debate [whether the private sector can provide a variety of public services more effectively than the government] has touched on services ranging from education, healthcare and transportation to trash collection and street repair. In addition to the normative question of what role government should assume in providing services [see, e.g., the discussion on "inherently government services" in this post], it has also raised the positive question of what determines government privatization decisions in practice.

We start with a simple model of procurement in which a government must arrange delivery of a service from an agent. The government can write a contract that specifies the time the agent must spend on the job and a set of performance requirements. Assume that specifying and enforcing a time requirement has minimal cost, but there are non-trivial costs to establishing and maintaining a set of performance requirements. Provided the government cares only about what is actually delivered, we show that an optimal contract must take one of two forms. The government either pays the agent for meeting a minimal time requirement or for meeting a performance requirement, but not both. These forms of contracting capture, in a rough way, the two most common ways that governments provide services: inhouse provision using salaried city employees and performance requirements contracts with private sector firms.

In our model, inhouse provision suffers from productive inefficiency due to the weak incentives of employees, but enjoys low contracting costs. In contrast, the productive efficiency of performance contracts comes at the cost of specifying and implementing performance requirements. This leads to predictions about how privatization decisions will vary across services. Services for which it is harder to write, monitor or adjust performance standards are more likely to be provided inhouse. The same will be true of services for which city administrators are more sensitive to the ultimate quality provided.

For obvious reasons, we cannot approximate a large-scale randomized research design that would allow us to quantify precisely how changes in contracting costs or city characteristics affect contracting practices. Rather what we can do is document broad patterns in contracting practices and relate them to our model and other ideas proposed in the theoretical literature.

Of course, a central prediction of efficiency based theories is that difficulties in specifying and administering performance requirements are likely to reduce privatization. To quantify these difficulties, we surveyed a set of city administrators, asking them to assess twenty-nine city services along a number of salient dimensions.1 We use this data to construct a measure of performance contracting difficulty.2

Our main empirical findings can be summarized as follows. First, services for which it is harder to write and administer performance contracts are less likely to be privatized. The effect is substantial. A one standard deviation change in contracting difficulty is associated with a change in the probability of being privatized of eight percentage points – that is, a forty percent reduction in the likelihood of privatization.

We also find that services ranking lower in terms of resident sensitivity to quality are more likely to be privatized.

Overall, our results indicate that a transaction cost view of privatization provides a useful framework for explaining local government contracting patterns. Notably, however, our results do not allow us to distinguish very clearly between the distinct sources of transaction costs that have been suggested in the theoretical literature. When we try to separate out problems with performance measurement, the potential for holdup, and the desire for control and flexibility, we find our survey measures of these problems to be so highly correlated across services as to be essentially impossible to disentangle.

In light of this, our view is that decomposing specific sources of transaction costs is likely to require more detailed data, most likely on contracting outcomes as well as choices.

The bulk of our analysis focuses on identifying city and service characteristics that are associated with privatization. We document that, controlling for the mix of services that cities provide, cities that do more private sector contracting spend notably less per capita.

The authors qualify that last observation with the following statement, which hopefully speaks more to their methodology than bias:

Though it is difficult to infer causality given the available data, the result is consistent with our modeling approach and suggests possibilities for future research.