V. Budget, Action Items and Conclusions

A. Budget
B. Action Items
- 1. Short Run
- 2. Long Run
C. Conclusions
Notes

A. Budget

There are several ways to consider the budget issue. One way is to ask how much money is going into economics at NSF now and discuss the appropriate percentage add-on for computational economics. The present economics program budget is $12 million, so if we are talking about a 50 percent increase, then $6 million per year is the appropriate number.

However, there is another approach. This begins from the notion that the United States faces important economic problems and that computational economics can over the long haul make important contributions to the management of some of these problems. We believe that our report has fully documented that computational methods have made significant contributions to economics in the past and that there is every reason to believe that the potential for future contributions is even greater.

This approach then continues by inquiring about (1) how many well-trained investigators there are who could make contributions if they had the necessary resources and (2) the size of the budget required for the typical investigator.

Consider first the number of investigators. One way to estimate the number of qualified individuals is to consider the top fifty departments in the country with an average of thirty economists in each department. This yields a pool of 1,500 economists. This is a very conservative estimate of the pool of talent since it does not include the many economists not in academics. Also it will be important to encourage interdisciplinary research in this field so that many of the qualified researchers are in other departments at universities.

If one-third of our conservative estimate of 1,500 economists are in the areas covered by computational economics then there would be 500 qualified investigators. In the strong competition for research support perhaps only a fourth of these qualified individuals would receive support. In this case there would be 125 investigators.

Next consider the question of the level of support for each investigator by calculating the annual support cost for a typical investigator.

Summer salary...................................15
Graduate student support........................20
Programmer or postdoctoral support..............20
Supplies and travel..............................5
          Subtotal..............................60

Overhead  (45% of subtotal).....................27
Hardware cost per year..........................20

          Total per investigator per year..... 107

These numbers are a rough estimate but they give us a basis from which to begin. So at 107 thousand dollars per investigator per year and 125 investigators we get a total of roughly 13 million dollars per year.

Then in addition, we believe that the budget should include funds for one or more large projects like the National Economics Server or other database and software projects. Estimate that we might need 3 million dollars per year for this plus 2 million a year for educational, postdoctoral and other programs. Then the grand total would be 18 million dollars per year.

A summary of the budget is presented below.

I.  Items awarded through grants to Principal Investigators
 
         Grants to 125 PI's  @  107.....................13,375

II.  Items awarded separately

     a.  Economics Data Server Project....................3000
     b.  Workshops and conferences.........................500
     c.  Education projects................................300
     d.  Research experiences for undergraduate............200
     e.  Postdocs in a national competition;  15 @ 40......600
 
                  Subtotal...............................4,600
 
                  Grand total...........................17,975

B. Action Items

We would also like to suggest a group of action items, some of which can be accomplished in the short run and others which will require a longer period of time.

1. Short Run

a. Software Purchases. The current policy in the Economics Program at NSF is to discourage investigators from adding software to their grant proposals. Software availability is an important constraint for many investigators so this policy should be reversed with investigators encouraged to add software to their grant request.
b. Provide access to the NSF network to small companies that are developing economics software.
c. Provide startup funding for the creation of one or more digital journals in economics.

2. Long Run

a. Facilitate the shift to a workstation and network-based infrastructure for economists. This includes funding purchases of workstations, shared servers, and network access fees as well as funding of hardware and operating systems support personnel for local networks. Also, NSF should encourage economist users groups to promote adoption of standards in communications and database design for economics software.
b. Explore development of an educational acceleration program in computational economics. The purpose of this program is to accelerate the transmission of computer tools developed in computational economic research to the teaching of economic analysis, especially in undergraduate economic programs. It might be expected that initial emphasis would be on funding proposals that develop computer-based visualization and simulation aids in teaching, use of national networks for access to economic data, freeware versions of "standard" micro and macro models, and for inter-campus competitions in simulations of multi-agent markets and decentralized decision making. Some portion of funding in this area would be targeted at training faculty at smaller colleges and universities in the use of national network facilities for research and teaching.
c. Create a National Economics Server (NES) system. The NES would serve as a network "front end" for economists to download economic data and computational software. The server would have links to the database systems of all the organizations which maintain major economic databases. Thus, the user would not need to know which agency compiled and maintained the data and would not have to know the conventions or the accounting procedures of each organization but, rather, only those of the National Economics Server.
The economist would query the NES and the national server would then, in turn, translate the user's request to the format required by the relevant database. The database would respond to the NES signal by forwarding the requested item by e-mail. There could be a number of servers on the system distributed at university or research institution sites, such as the NBER, with dedicated servers to accommodate NES inquiries. A simple initial pilot might be based on the model of the "netlib" gateway machine at AT&T Bell Labs, where the machine "reads" the user e-mail request by string-matching and automatically fills the user's software request from library files.
After experience with this pilot, a more adequate server system should be built on an object-oriented database design, such as those being developed in C++. The object-oriented design would accommodate not only the different "backend" filters that are required by the idiosyncratic file organizations of different databases, but permit the user to select from a variety of "front end" filters for mail delivery (such as file structures compatible with SAS, Splus, etc.). An object-oriented design would also facilitate pulling matched data from disparate database sources (often required in panel data analysis), edited or pre-processed data where data attributes include the transformation algorithms, or downloads of data with complex structures (such as those generated by models). Obviously, more efficient electronic transmission of large bulk data requests should also be explored.

C. Conclusions

Computational economic research is at a level of maturity at which breakthroughs are needed in order to raise this field to a new level of understanding. At the same time, vast amounts of data are located at different sources, and advanced computers can be accessed from even remote locations. Increasingly, one is becoming aware of economists who are taking on the challenges of capturing the technologies for use in their research. Although working, more or less in isolation, this community of "computational economists" is beginning to network via meetings and the establishment of a new journal. An initiative at NSF along these lines would help to create a critical mass for such research activities and exchanges. Indeed, the research questions being raised by this community and the intellectual excitement present, with added support, are sure to push forward the frontiers of economics.

Notes

1. The National Science Foundation supported SIPP ACCESS from November 1984 to the present (SES-8411785, SES- 8716448, and SES-8921213). The Bureau of the Census supplemented funding in 1989 (SES-8701911).

Seed money for SIPP ACCESS and development of the PC- SIPPTEST database and reference materials was supported by Sloan Foundation grant \#B1987-46.

[ Next | Previous | Table of Contents | Title Page ]