Economist Ron Utt at the Heritage Foundation has provided a useful and accessible critique of the job creation estimates thrown around by advocates of new government spending. The analysis is a few years old (2004) remains a great primer on the weaknesses of claims that x spending on y program will create z number of jobs.
Ron's analysis is particularly relevant now because the critique was aimed at the debate over transportation reauthorization then. Most of these studies rely on a technique used by economsits called "input-output" analysis. The idea is you can trace expenditures from in one sector of the economy and map the dollars into other sectors. So, for example, $100 dollars spent paying a worker to make a steel rail will be spent by the worker elsewhere on clothes, food, entertainment, cars, etc. Similarly, the money spent to purchase the rail will be used by the manufacturer to buy the raw materials and equipment needed to manufacture it. The recycling of these dollars becomes a "multiplier" that is used to estimate "new" income (and jobs) when a dollar is spent in the economy.
But, as Ron notes, these studies are very inprecise, and definitive claims about job creation are not justified by the methods or types of data available.
Such qualifications are particularly justified given that the mathematical model used by the DOT--traditional I/O analysis--is little more than a comprehensive technical description of the quantities of materials, supplies, and labor that are needed to make a certain product. This model does not accurately describe the complex workings of a market economy in which, each moment, thousands of participants make millions of choices involving hundreds of thousands of services and commodities, all in limited supply. In the real economy, more of one thing means less of another in the short run as individuals and businesses substitute one product for another in response to changing prices. The DOT traditional I/O analysis does not consider such offsets and substitutions.
For example, using the job creation numbers provided by JOBMOD, an additional billion dollars in highway spending requires an estimated 26,524 additional workers3 to build and supply a billion dollars worth of new highways. In the real world, the additional federal borrowing or taxing needed to provide this additional billion dollars means that a billion dollars less is spent or invested elsewhere and that the jobs and products previously employed by that billion dollars thus disappear. Regardless of how the federal government raised the additional billion dollars, it represents a shift of resources from one part of the economy to another, in this case to road building. The only way that a billion dollars of new highway spending can create 47,576 new jobs is if the billion dollars appears out of nowhere as if it were manna from heaven.
Indeed, studies performed after the fact (ex post) often show much more modest impacts on the economy. Indeed, these limitations have been recognized by other assessments of government programs, and Ron discusses these as well. Here's a sample:
Of relevance to the potential impact of highway spending alone, the study also notes that "funds for public works programs, such as those that build highways or houses, were spent much more slowly than funds for public services." This is understandable given the long lead time between the decision to build and the time construction actually begins. For the typical federally funded road, environmental impact studies, construction plans, land acquisition, competitive bidding, and awarding of contracts can take several years. In some instances, the environmental permitting process can exceed five years.8 As a result of such delays, any employment effects related to additional highway spending would not occur for several years, thereby providing only a few jobs to those unemployed when the bill was enacted.
The full analysis is not that long and well worth a read (or re-read) in today's climate.