The Effectiveness of the «Checkpoint Tennessee» Program John H. Lacey*, R alph K. Jones*and Jam es C. Fell** *M id-a m erica R esearch Institute, Shepherdstow n, W est V irginia U SA **N ational H ighw ay T raffic Safety A dm inistration, W ashington, D C U SA A B S T R A C T A sobriety checkpoint dem onstration program w as conducted in the S tate o f T ennessee betw een A pril 1994 and M arch 1995. P rior to this period, the state typically conducted 10-15 checkpoints on an annual basis. D uring the experim ental checkpoint dem onstration period, a total o f alm ost 900 checkpoints w ere conducted involving close to 150,000 drivers checked and resulting in alm ost 800 arrests for driving under the influence (D U I) or driving w hile intoxicated (D W I). T he checkpoint program w as highly publicized over T V, on the radio and in the new spapers throughout the state. Surveys o f drivers throughout the state indicated aw areness o f the enforcem ent program. A n interrupted tim e series w as used to analyze the traffic safety im pact o f the program. T he results show ed a 21.6% reduction in the num ber o f drunk driving fatal crashes that w ould have occurred w ithout the checkpoint program. IN T R O D U C T IO N Sobriety checkpoints have long been know n to be an effective im paired driving enforcem ent m ethod. In a review o f the literature, it w as concluded that the accum ulation o f positive findings for visible and w ell-publicized checkpoints provide support for the proposition that sobriety checkpoints are capable o f reducing the extent o f drunk driving and o f deaths and injuries on the highw ays [Ross, 1992a], H ow ever, until recently, checkpoints have generally been im plem ented in the U nited States (U S) on a local level. A w ell-publicized sobriety checkpoint program held in B ingham ton, N ew Y ork, resulted in a 39 percent decrease in the num ber o f drinking drivers on the roads at night according to roadside surveys and a 23 percent reduction in late-night crashes in the m onths the checkpoints w ere held [W ells, et al, 1991]. In N ew Jersey, checkpoints w ere associated w ith a drop o f 10 to 15 percent in single vehicle nighttim e crashes (a com m only used m easure of alcohol-im paired driving) [Levy, et -969-
al, 1988], A year-long checkpoint program in C harlottesville, V irginia w as associated w ith a 13 percent reduction in the proportion o f crashes that w ere alcohol-related [V oas, et al, 1985], S im ilar results w ere obtained from a checkpoint program in C learw ater and L argo, F lorida, w hich experienced a 20 percent decrease follow ing checkpoint operations [Lacey, et al, 1986], W hile these results have been encouraging, for various reasons [Ross 1992b] very few states in the U S have em barked on statew ide sobriety checkpoint program s. B ased upon their potential effectiveness, and the strong evidence from A ustralia on their random breath testing (R B T ) program [H om el, 1990], the N ational H ighw ay T raffic Safety A dm inistration (N H T SA ) decided to conduct a dem onstration project in a state that w as w illing to change its philosophy about checkpoints. M E T H O D S In 1993, N H T S A entered into a cooperative agreem ent w ith the State o f T ennessee to conduct a highly publicized sobriety checkpoint program throughout the state and evaluate the effects o f that program. In M arch 1994, T ennessee initiated a statew ide im paired driving checkpoint program labeled «C heckpoint T ennessee.* W ith equipm ent purchases, som e logistics, and the evaluation funded by N H T SA, but personnel provided though diversion o f existing resources in the T ennessee H ighw ay Patrol, four sets o f three checkpoints w ere conducted throughout the state every w eekend using specially equipped vans w ith generators, lights, cones, signs, video taping and evidential breath testing equipm ent. Passive alcohol sensors in flashlights w ere also used to assist officers in detecting im paired drivers. Periodically, checkpoints w ere scheduled on the sam e night in each o f the 95 counties in the state. T hese, necessarily, did not involve as m any officers o r as m uch equipm ent per checkpoint as w as typical during other w eekends but served to reenforce the «blitz» concept. T h e checkpoints w ere coordinated and conducted prim arily by the T ennessee H ighw ay P atrol w ith support from local law enforcem ent agencies. P ublicity in support o f the program w as stim ulated by obtaining the special cooperation o f a single television station in each o f the four m ajor m arkets in the state. T hey each broadcast «C heckpoint T ennessee* as a special project. This publicity w as enhanced by «hard news* coverage from other outlets, a statew ide -970-
billboard cam paign and press releases announcing individual checkpoints, follow ed up by reports o f their results in term s o f arrests, etc. T w o w aves o f a paper and pencil survey w ere adm inistered in several d riv er s license renew al offices to m easure know ledge and attitudes about the program. T he first w ave was adm inistered in M arch 1994 prior to the form al announcem ent and initiation o f the C heckpoint T ennessee program. T he second w ave w as adm inistered in the sum m er o f 1994, four m onths after program initiation. T he first w ave yielded 1,305 respondents w hile the second w ave yielded 1,071. T he results o f several questions indicated increased aw areness o f the «C heckpoint Tennessee* program. T he driver license office survey data indicated overw helm ing public support o f checkpoint program s w ith 88.1% o f respondents supporting the use o f checkpoints at w ave 1 and 91.6% at w ave 2. T here seem ed to be a slight im provem ent in self-reported drinking driving behavior w ith 8.6% o f respondents adm itting to having driven after having too m uch to drink at w ave 1 and 7.3% at w ave 2. Sim ilar changes w ere observed in reported driving w ithin tw o hours of drinking even though the perceived risk o f arrest eroded slightly betw een waves. B etw een A pril 1, 1994 and M arch 31, 1995, a total o f 882 checkpoints w ere held. T his com pares to the typical 10-15 checkpoints conducted on an annual basis for the 5 years prior to the dem onstration project, yielding quite a contrast in program s. A total o f 144,299 drivers passed through these checkpoints w ith 773 arrested for driving under the influence o f alcohol (D U I) or driving w hile intoxicated (D W I). A n additional 201 drivers w ere arrested for drug violations, 84 for youth offender violations, 35 felony arrests w ere m ade, 49 w eapons w ere seized, 1,517 w ere cited for safety belt or child restraint violations and 7,351 w ere given other traffic citations. T elevision, radio and print m edia coverage w as extensive during the 12 m onth operations phase o f the program. E V A L U A T IO N A P P R O A C H A n interrupted tim e series approach w as used in analyzing the traffic-safety im pact o f the checkpoint program. In this approach, a tim e series o f the data o f interest is studied to see if an «intervention» occurring at som e point in the series is a statistically significant factor in a -971 -
m athem atical m odel o f the series. T he intervention analyzed here is the «C heckpoint T ennessee* program. The independent variable and m easure o f effectiveness in the m odel w as «drunk driving fatal crashes.* A drunk driving fatal crash w as defined as a fatal crash in w hich one o f the involved drivers had a blood alcohol concentration (B A C ) o f 0.10% or m ore eith er through direct B A C test results or through an algorithm developed by N H T SA [K lein, 1986], Ideally, all classifications w ould be through direct B A C tests, how ever no state as yet obtains a B A C test o f all drivers in fatal crashes and this approach is considered to be the best available alternative. T he data used in the m odel w ere retrieved from N H T S A s Fatal A ccident R eporting System (FA RS). The data covered the period 1988 through 1995. T w o techniques w ere used to guard against attributing any changes in drunk d riving fatal crashes to the program w hen they m ight have been due to som e oth er events that ju st happened to coincide w ith the program. First, a m odel o f drunk driving fatal crashes in five states surrounding T ennessee (K entucky, G eorgia, A labam a, M ississippi, and L ouisiana) w as developed using the sam e procedures to see if an effect occurred coincident w ith T ennessee intervention. Such an effect m ight be indicative o f a regional or, possibly, a national factor having nothing to do w ith the intervention. All fatal crashes w ere also included as an explanatory variable in the m odel for T ennessee and the m odel for the five surrounding states. Fatal non-drunk driving crashes w ere also considered as an explanatory variable but the resulting m odel did not provide as good a fit to the data. N om inally, it w as assum ed a program start date o f A pril 1, 1994, but w e also studied the effect o f assum ing several other start dates to account for a possible lag betw een the tim e the program w as started and the tim e an im pact occurred. It w as assum ed that a step-function intervention w as appropriate for the m ajority o f the analyses, and the effect o f interventions o f other tim e profiles, for exam ple, a ram p function, w as studied. T he A R IM A analysis m ethod developed by Box and Jenkins in the 1970s, and incorporated in the SAS statistical package as PRO C A R IM A w as used. -972-
R E S U L T S T he best fit to the T ennessee series w as obtained through a m odel using all drunk-driving fatal crashes as the dependent variable. A ll fatal crashes w as used as an input series. T he transfer function fo r the input series w as a sim ple scalar o f value equal to 1. The model showed a significant effect fo r the intervention variable (a step function coincident with the checkpoint program start date) amounting to a reduction o f about nine drunk-driving fatal crashes per month (t ra tio -6.3 0 ). T his w as a 21.6% reduction over the num ber o f drunkdriving fatal crashes that w ould have occurred w ith no intervention. T he results are depicted graphically in Figure l.t h e m odel for the com parison series used 12-span differencing o f the dependent variable (drunk-driving fatal crashes), and used the sam e differencing o f the dependent variable (all fatal crashes). A gain, the transfer function w as equal to 1. T he m odel show ed an insignificant increase in drunk-driving fatal crashes coincident w ith the T ennessee intervention (t ratio= 1.02, F igure 2), lending support to the hypothesis that the checkpoint program w as responsible for the positive results observed in Tennessee. D IS C U S S IO N W hile other statew ide sobriety checkpoint program s have recently been initiated in the U S (in N orth C arolina and N ew M exico, to nam e tw o) this dem onstration in T ennessee is o f interest because it resulted in a significant decrease in alcohol-related traffic fatalities w ith relatively low im plem entation costs. T he total cost o f the tw o year dem onstration project w as only $927,594, w ith federal funding at $452,255 and state funding at $475,339 in the cooperative agreem ent. T he state contribution covered police salaries, publicity costs and other program expenses. T he police salary contribution w as accom plished by a reallocation o f effo rt to this endeavor rather than through additional funding from other sources. N H T S A funding covered design, equipm ent and program evaluation. T he State o f T ennessee has elected to continue w ith the checkpoints on a w eekly basis, although not at the sam e frequency or intensity as the 12 m onth operational phase described in this paper. T hat is also considered a successful outcom e w ith federal funding stim ulating the continuation o f a program the state deem s to be effective. F urther analyses including costbenefit and im pact on non-fatal crashes are underway. -973 -
M any o f the logistical reasons for the non-use o f sobriety checkpoints (e.g., they are too expensive, require too m uch m anpow er, do not yield enough D W I arrests) [Ross, 1992b] are being overcom e by the results o f this program and o f those in N orth C arolina and N ew M exico [Lacey, et al, 1995], A recent study [Stuster, et al., 1995] show s that sobriety checkpoints yield greater public aw areness o f the program and greater decreases in alcoholrelated crashes than an enforcem ent program involving roving patrols. T he prem ise o f highly visible, highly publicized, frequent sobriety checkpoints conducted on a statew ide basis appears to be a viable, effective deterrent to im paired driving. O ther states w here D W I checkpoints are legal should consider im plem enting statew ide program s and states w here they are not perm itted should undertake m easures to rem ove those legal barriers. R E F E R E N C E S H om el, R., R andom breath testing and random stopping program s in A ustralia. In W ilson, RJ and M ann, R. (Eds), Drinking and Driving: Advances in Research and Prevention. N ew York: G uilford Press, pp 159-204, 1990. Klein, T., A method fo r estimating posterior BAC distributions fo r persons involved in fa ta l traffic crashes, N ational H ighw ay T raffic Safety A dm inistration, W ashington, D C, D O T HS- 807-094, 1986. L acey, J., S tew art, J., M archetti, L., Popkin, C., M urphy, P., L ucke, R. and Jones, R., Enforcement and Public Information Strategies o f D W I General Deterrence: Arrest Drunk Driving: The Clearwater and Largo, Florida Experiences. T echnical R eport. W ashington, DC: N ational H ighw ay T raffic Safety A dm inistration, D O T H S-807-066, 1986. L acey, J., Jones, R. and Fell, J., A com parison o f blitz versus continuous statew ide checkpoints as a deterrent to im paired driving, In Proceedings o f the 13th International Conference on Alcohol, Drugs and Traffic Safety, Volume 2, K loeden and M cl ean (Eds), pp 845-848, 1995. L evy, D, Shea, D., and Asch, P., Traffic safety effects o f sobriety checkpoints and other local DW I programs in New Jersey. A m erican Journal o f Public H ealth, 79: 291-293, 1988. -974-
R oss, H.L., The deterrent capability o f sobriety checkpoints: Summary o f the American Literature. T echnical R eport. W ashington, DC, N ational H ighw ay T raffic Safety A dm inistration, D O T H S-807-862, 1992a. Ross, H.L., R easons for non-use o f sobriety checkpoints. Police Chief, V ol. L IX, No. 11, pp 58-63, 1992b. Stuster, J. and B low ers, P., Experimental Evaluation o f Sobriety Checkpoint Programs, T echnical R eport., W ashington, D C, N ational H ighw ay T raffic Safety A dm inistration, D O T H S-808-287, 1995. V oas, R., R hodenizer, E., and L ynn, C., Evaluation o f Charlottesville Checkpoint Operations: Final Report. T echnical R eport., W ashington, D C, N ational H ighw ay T raffic Safety A dm inistration, DO T H S-806-989, 1985. W ells, J., Preusser, D., and W illiam s, A., Enforcing alcohol-impaired driving and seat belt use laws, Binghamton, NY. Insurance Institute for Highw ay Safety, A rlington, VA, 1991. A C K N O W L E D G M E N T T his dem onstration program and evaluation w as funded by the N ational H ighw ay T raffic Safety A dm inistration (N H T SA ) and the State o f Tennessee. T he opinions and conclusions are those o f the authors and not necessarily those o f N H T SA or the S tate o f T ennessee. -975 -
- 9 7 6 -