Semisequicentennial Transportation Conference Proceedings
May 1996, Iowa State University, Ames, Iowa

Pooled Estimates of the Effects of Through, Interstate, and Bypass Traffic on Kansas Towns

David Burress

Institute for Public Policy and Business Research,
University of Kansas,
607 Blake Hall,
Lawrence, Kansas 66045-2960.

Several studies of the economic effects of bypasses have compared bypassed towns with similar towns lacking bypasses. Using an improved approach, Anderson et al. (1) regressed sales tax data on traffic on bypasses and traffic on other highways. Their study had two limitations: it did not distinguish local traffic from through traffic, and it did not distinguish Interstate bypasses from bypasses on other routes. This is important because local traffic has different economic effects than through traffic, and freeways have different economic effects than ordinary highways. The present study created a new dataset showing four different types of traffic for each town in Kansas: local traffic, through traffic inside the town, through traffic on Interstate-type highways, and through traffic on bypasses of ordinary highways. These data were inferred from gross traffic counts using an Origin-Destination model of Kansas. Through traffic inside the town was found to have a positive long-term effect on employment and payroll (measured from unemployment compensation data). Diverting downtown traffic onto a non-Interstate bypass does not significantly reduce that effect. However, diverting down-town traffic onto an Interstate appears to reduce its employment effect by roughly one half, but the paper argues that this effect is due to a bias resulting from new traffic attracted by the Interstate. In contrast, by aggregating both types of bypasses Anderson et al. found a moderately negative effect on travel-related businesses. These results imply that Interstate bypasses should not be aggregated with ordinary bypasses. Key words: bypass, economic-impact, freeways, origin-destination.


Leaders in a small town are often wary when a bypass is proposed. They fear that downtown retail businesses depend on impulse decisions by through traffic drivers, which will dry up if the businesses are no longer visible to drivers. There have been many uncontrolled case studies on this question. A few studies have used paired comparisons between bypassed towns and similar towns without bypasses (reviewed in Federal Highway Administration (2)). These studies are persuasive both because of good sample sizes and because of the matched comparison, but the statistical control achieved this way is necessarily incomplete.

Using a greatly improved approach, Anderson et al. (1) regressed sales tax data for bypassed towns and other towns on two main variables: counts of traffic on bypasses, and counts of traffic on other highways. Their study used pooled time-series/cross-section data; in other words, many individual towns were tracked over several years. The great advantage of a pooled regression model is that it can control simultaneously for several important factors, including previous economic conditions in each town, contemporary economic conditions in other towns, and the year in which a bypass was built. Their dataset had two potential limitations: it did not distinguish local traffic from through traffic, and it did not distinguish bypasses on Interstates and similar freeways from bypasses on ordinary highways. One expects that local traffic has different economic effects on towns than through traffic has, and research has shown that freeways have different economic effects on towns than ordinary highways. Using a unique dataset, this paper shows these differences are important. (The study also had limitations in regression techniques, most importantly in not controlling for effects due to specific town and specific year.)

TRAFFIC COUNT DATA BY ORIGIN OR DESTINATION

The new dataset showed four different types of traffic for every town in Kansas:

These data were inferred from point traffic counts using an Origin-Destination model of Kansas. The model included a shortest travel time algorithm to predict routes, and a gravity model that related traffic counts to populations and travel times. The model was fit to traffic data (3 ) using a least squares algorithm. Modeled traffic estimates were weighted so that the four types of traffic added up to observed traffic counts on each highway segment.

Note that some of the traffic traveling on a bypass is actually local traffic making use of part of the bypass on the way into town from another place. The origin-destination model was used to remove that kind of traffic from the through traffic counts and place them in the local traffic counts.

UNEMPLOYMENT INSURANCE (UI) DATA

Measurements of economic conditions were based on detailed firm-level UI records for the first quarter of each year from 1988 through 1994. This is the first bypass study using UI data. UI data give a more complete picture of the economy than sales tax data, since many goods are sales-tax-exempt. The UI data include the employment of the firm, its total payroll for the quarter, and its Standard Industrial Classification (SIC) code. Both employment and payroll were aggregated by year, town, and SIC category for every town in Kansas. SIC codes were categorized into five categories and totals: travel-related retail, non-travel related retail, total retail, all other, and grand total.

METHOD

Pooled regressions were run to explain each of the five categories. In all cases, results using employment data were similar to results using payroll data, so this paper focuses on employment. Results were subjected to a sensitivity analysis with many variant specifications (details on request from the author). In general, the conclusions reached below were quite robust.

THE BASIC MODEL

Each regression model was of the form

Employmentti = a + b(local traffic)ti + [1]

g(total through traffic)ti + d(freeway traffic)ti +

e(ordinary bypass traffic)ti +f'(vector of other variables)ti + uti,

where the Greek letters represent constants to be estimated; t is the year and I is the town; uti is a disturbance term (which always included terms for each town and each year); and "total through traffic" is the sum of inside traffic, freeway traffic, and ordinary bypass traffic. In this model, we expect positive signs on b and g, because both local traffic and through traffic are associated with employment. g measures a "response to traffic" effect, in which local business receives some gain on average from each unit of through traffic. Note that if some traffic is diverted from downtown onto a bypass, then g still multiples the same quantity, but e multiplies a larger quantity than before. Therefore we expect a negative sign on e if bypasses reduce employment, and we might interpret e/g as the share of employment generated by through traffic that is lost when that traffic is diverted onto a bypass. Similarly, we expect a negative sign on g, and we might interpret d/g as the share of employment generated by through traffic that is lost when that traffic is diverted onto a freeway. In practice, however, the coefficients e and d are subject to two additional effects:

We should not view the "new industry" effect as a source of bias in the model. Rather, it is a real benefit to the town which is just as valuable as the "response to traffic" effect. The "new traffic" effect, however, leads to an unambiguously downward bias in the d and e coefficients; in other words, it causes this model to overstate the damage, or understate the benefits, from bypasses.

RESULTS

Selected results are shown in Tables 1 and 2. All estimates of g were positive and highly significant (p=.01), implying that through traffic inside the town does have a strong positive association with employment. In all cases, d is much more negative than e, implying that the "new traffic" effect far outweighs the "new industry" effect.

TABLE 1 Effect of Traffic Flows on Retail Trade Employment: Results for a Plausible Model

Dependent Variable
Independent Variable Total Retail Travel- Related Retail Non-travel- Related Retail
Intercept
-365.0779 -59.1055 -238.0023

(211.5983) (23.7826) (150.9437)
Local traffic 0.2037 0.0163 0.1482

(0.0754) (0.0084) (0.0542)
Total through traffic 0.4960 0.0573 0.3588

(0.0799) (0.0089) (0.0575)
Bypassed traffic, ordinary highways -0.0533 0.0023 -0.0466

(0.1106) (0.0124) (0.0797)
Bypassed traffic, freeways -0.2975 -0.0282 -0.2259

(0.0819) (0.0092) (0.0590)
R-squared .257 .264 .254
Ratios:
Bypass effect/effect of all through traffic -0.107 0.040 -0.123
Freeway effect/ effect of all through traffic -0.600 -0.493 -0.630

Note: Traffic is measured as average vehicles per day. Standard errors are in parentheses.

TABLE 2 Effect of Traffic Flows on Total Employment: Results for Some Alternative Models

Model Variant
Independent Variable Model 1 - Model 2 - Model 3 (includes very small towns) Model 4 (log-linear)
Intercept -128.4509 - -122.7755 -0.8988

(62.3633)
(38.6227) (0.4458)
Local traffic 0.0533 0.0485 0.0544 0.7422

(0.0218) (0.0217) (0.0163) (0.0544)
Total through traffic 0.1391 0.1128 0.1233 0.0504

(0.0231) (0.0193) (0.0166) (0.0113)
Bypassed traffic, ordinary highways -0.0076 0.0036 0.0066 0.0149

(0.0319) (0.0315) (0.0254) (0.0064)
Bypassed traffic, freeways -0.0727 -0.0527 -0.0702 0.0081

(0.0236) (0.02157) (0.0170) (0.0050)
Control variable:
Non-retail employment - - - 0.0769

(0.0116)
R-squared .251 .236 .250 .424
Ratios:
Bypass effect/effect of all through traffic -0.055 0.032 0.053 0.296
I-road effect/ effect of all through traffic -0.523 -0.467 -0.570 0.160

Note: Traffic is measured as average vehicles per day. Standard errors are in parentheses.

Diverting down-town traffic onto a non-freeway bypass does not significantly reduce its measured effect on employment: in all cases e is positive or not statistically significant or both. In other words, bypasses on ordinary highways do not significantly harm the overall level of economic activity in a community. In a few variants, ordinary bypasses appeared to help total employment, though not retail employment. (This conclusion can be strengthened slightly if we are willing to assume that bypasses have attracted some appreciable amounts of new traffic in this sample of towns. If so, e is biased downward and these bypasses almost certainly had positive effects on employment.)

However, diverting down-town through traffic onto a freeway appears to reduce its effect on employment by more than 50 percent, depending somewhat on the variant model. Note however that this conclusion is negatively biased the large "new traffic" effect. Furthermore, the "new traffic" effect for freeways is probably at least as large as the difference between e and d ( because the "new industry" and "new traffic" effects affect (d-e) in opposite directions). If so, then all of the negative value of d (and perhaps more!) can be explained by the "new traffic" effect. It follows that diverting some existing traffic onto a freeway probably does not hurt employment.

RESEARCH IMPLICATIONS

Anderson et al. found that bypasses have a moderately negative effect on travel-related businesses. That is exactly what would be predicted from the present results if one aggregated ordinary bypasses with freeways. Yet the interpretations formed from the two studies differ. In the present study, bypasses on ordinary highways were found not to hurt travel-related businesses; in the Anderson study, all bypasses (treated as a group) did appear to hurt travel-related business. If the pattern in Texas is to Kansas, the Anderson et al. conclusion suffered from two kinds of bias: a "new traffic" bias, and an aggregation bias.

These results imply that Interstate bypasses should not be aggregated with ordinary bypasses in these studies. However they do not necessarily show that origin-destination models are needed. It remains to be seen whether the present result depends on the use of a distinct local traffic variable. On the other hand, use of a local traffic variable does make the present study more convincing.

POLICY IMPLICATIONS

According to these results, bypasses on ordinary highways do not hurt any category of business in the long-term. There is at least some evidence that both ordinary bypasses and Interstate bypasses are helpful to the town as a whole, in the long term. These results hold for non-travel-related retailing, for travel-related retailing, and for non-retailing. These results are somewhat more optimistic for small towns than previous research on the subject.

In all honesty, however, these results should not remove all cause for concern on the part of local town planners. First, these results refer to the aggregate and to the long term; there is much evidence in other studies including Burress (5) that some individual travel-related businesses are indeed hurt in the short-term by new bypasses. For example, some downtown businesses may decline while other new businesses form at the bypass interchange. If the new businesses have new owners, their creation provides no consolation to owners of older businesses.

Second, individual towns can certainly differ in either direction from the average town described by a regression model. In any case, the time when a new bypass is created is likely to be a time of economic change and challenge for any small town.

REFERENCES

  1. . S. Andersen, H. Mahmassani, R. Helaakoski, M. Eutitt, C. Walton, and R. Harrison. Economic Impact of Highway Bypasses. In Transportation Research Record 1395, TRB, National Research Council, 1993, pp. 144–152.
  2. Economic and Social Effects of Highways-Summary and Analysis. Federal Highway Administration, U.S. Department of Transportation, Washington D.C., 1972.
  3. 1990 Traffic Flow Map - State Highway System of Kansas, Kansas Department of Transportation, Bureau of Transport Planning, Topeka, Kansas,1990, 1992, 1988.
  4. T. Rephann, Highway Investment and Regional Economic Development: Decision Methods and Empirical Foundations. In Urban Studies, Vol. 30, No. 2, 1993, pp. 437–45.
  5. D. Burress. Impacts of Highway Bypasses on Kansas Towns, IPPBR Research Report No. 226, Institute for Public Policy and Business Research, University of Kansas, Lawrence, Kansas, 1996.

This research was supported by the Kansas Department of Transportation and the University of Kansas. Any opinions expressed here are the author's alone.

CTRE is an Iowa State University center, administered by the Institute for Transportation.

Address: 2711 S. Loop Drive, Suite 4700, Ames, IA 50010-8664

Phone: 515-294-8103
FAX: 515-294-0467

Website: www.ctre.iastate.edu/

Iowa State University--Becoming the Best