ardo/Obsidian

category: literaturenote citekey: hakkertevaluationeffectsdriverbehavior2001 title: The evaluation of effects on driver behavior and accidents of concentrated general enforcement on interurban roads in Israel authors: "Hakkert, A.S.; Gitelman, V.; Cohen, A.; Doveh, E.; Umansky, T." year: 2001 date: 2001-00-00 2001 doi: 10.1016/s0001-4575(00)00014-2 publication: "Accident Analysis & Prevention" url: "https://doi.org/10.1016/s0001-4575(00)00014-2" zotero_key: 2ZWJUD85 zotero_storage: 86Q6AEQ8 collections: transpordi_planeerimine folder: 001_artiklid/Road Safety Indicators firstAuthor: "Hakkert, A.S."

status: converted

Abstract

reserved.

8225716.

PII: S0001-4575(00)00014-2

Keywords: Police; Enforcement; Evaluation; Behaviour; Attitudes; Accidents

* Corresponding author. Tel.: +972-4-8292376; fax: +972-4-

E-mail address: trivica@tx.technion.ac.il (V. Gitelman).

A.S. Hakkert ^a

Accident Analysis and Prevention 33 (2001) 43–63

The evaluation of effects on driver behavior and accidents of concentrated general enforcement on interurban roads in Israel

^a Transportation Research Institute, Technion-Israel Institute of Technology, Haifa ³²⁰⁰⁰, Israel ^b The Technion Statistics Laboratory, Technion-Israel Institute of Technology, Haifa ³²⁰⁰⁰, Israel Received 4 May 1999; received in revised form 12 January 2000; accepted 21 January 2000

The study deals with the evaluation of the effects of a new deployment of the National Traffic Police in Israel — a concentrated general enforcement on 700 km of interurban roads, which contain 60% of all rural accidents and about half of the severe accident locations. The enforcement project began in April 1997 and lasted for 1 year, aiming at a 10% reduction of severe accidents on those roads. The evaluation study consisted of three main parts: (a) monitoring of everyday police operations on the project roads; (b) periodic evaluation of the project's influence on drivers' behavior and attitudes; and (c) the evaluation of accident changes within the project area, at the end of the project year. Speed measurements, performed before the police project launch, demonstrated that, depending on road type, up to 85% of drivers violated the rural road speed limits. Two rounds of field observations, before and during the project, took place at 32 sites throughout the project area and comprised eight behavior types: turning performance; signaling while turning; compliance with 'stop' and 'yield' signs; safety belt use by drivers and front-seat passengers, at the intersections; 'keeping to the right' on dual carriageway roads; and 'not-crossing of the white separating line' at single carriageway roads. A comparison revealed a general reduction in violation rates of most behaviors, during-the-project, except for compliance with 'stop' signs and signaling. Concurrently, two rounds of driver surveys were performed which examined the project's influence on the drivers' attitudes towards the police enforcement effectiveness and risk connected with traffic rule violations. An improvement was identified in perceived general level of police activity, during the project, whereas a lower or similar risk of apprehension was attributed by drivers to most specific violations of the traffic rules. Three groups of indices of police activity were estimated monthly during the study: inputs; outputs; and efficiency indices, which characterized the police presence on the project roads, the citations produced and the usage rates of the vehicle fleet and the enforcement tools. Analyzing these indices, two periods in the project performance were determined, and three criteria were proposed to subdivide the project roads into two groups, according to the level of enforcement intensity. Both periods and enforcement levels, as well as geographic zone, were accounted for in the analysis of accident trends on the project roads. For the accident analysis, a statistical model combining the odds ratio and longitudinal methods, was developed and applied to the severe accident counts, all injury accidents and severe casualties. A statistically significant reduction in severe accidents and severe casualties, as opposed to the comparison group, was found on the highly enforced roads in the center of the country. At the same time, since the police project began, an increase in accidents was observed on most interurban roads, including the northern parts of the project area. However, it was noted that in four of the five project road groups the mean value of the odds ratio was much less than one. Thus, although the enforcement project did not attain its full purpose, it seemed to be a deterrent factor for the increasing accident trend that appeared that year on the interurban roads. The findings pointed out that the National Traffic Police did not exhaust its potential in the project's performance and needs more flexible enforcement and deployment tactics. © 2000 Elsevier Science Ltd. All rights

1. Introduction

The National Traffic Police (NTP) in Israel was established in 1991 as an operational branch of the

, E. Doveh ^b

, V. Gitelman a,*, A. Cohen ^b

, T. Umansky ^b

Table 1 Violation rates of driver behaviors at intersections in the second round as opposed to the first round of observations

Direction of travel	Round	Violation rate, %, in behavior
		Signaling for turning	Compliance to 'yield' sign	performance Turn	Compliance to 'stop' sign	Fastening safety belts: driver	Fastening safety belts: front-seat passenger
Type of site: signalized intersection Turning right from minor road	2 1	P=0.03 32.3 30.3	P=0.09 44.9 40.7	PB0.001 15.8 5.5	–	PB0.001 11.6 6.7	PB0.001 16.3 7.3
Turning left from minor road	2 1	PB0.001 34.3 30.2	–	PB0.001 7.4 2.4	–
Turning left from major road	2 1	PB0.001 27.9 40.8	–	P=0.08 5.9 5.1	–	PB0.001 13.1 4.9	PB0.001 11.8 1.9
Type of site: non-signalized intersection Turning right from minor road	2 1	PB0.001 27.6 23.5	PB0.001 30.2 8.9	PB0.001 6.3 1.9	P=0.04 38.2 37.4	PB0.001 13.8 3.9	PB0.001 14.1 2.0
Turning left from minor road	2 1	PB0.001 22.4 19.5	PB0.001 30.8 17.0	PB0.001 12.6 7.3	PB0.001 31.5 21.2
Turning left from major road	2 1	PB0.001 16.6 9.8	–	PB0.001 9.3 4.7	–	PB0.001 9.4 5.0	PB0.001 11.5 7.2
Average for all sites and travel directions	2 1	P=0.12 27.0 28.5	PB0.001 32.7 19.5	PB0.001 4.3 9.1	P=0.16 34.2 30.5	PB0.001 12.0 5.3	PB0.001 13.9 4.9

The driver survey performed in this study, examined the 700-project influence on the drivers' attitudes towards the police, the enforcement effectiveness and risk connected with traffic rules' violations and accident occurrences. The survey took place twice at the same sites, eleven fuel stations throughout the project area, and on the same working days, Monday to Thursday. In each round, some 1200 drivers were interviewed by means of a questionnaire, including 40 questions and brief details about the driver and the vehicle. Each question was associated with an attitude. The answers were measured on an ordinal scale of '1–10', where rank '10' represents the highest level of attitude/characteristic. A comparison between the two rounds of survey was performed aiming to reveal changes in drivers' attitudes caused by the project performance (i.e. the second round results were contrasted to the first round ones) and to check the differences in the attitudes in areas with different enforcement intensity. The values were analyzed in two forms: (a) as gain-values where, for each attitude, mean and other sample characteristics were estimated and a t-test was applied for testing hypotheses; (b) as ordinal values where a percentage of respondents having a high positive attitude (i.e. ranks 8, 9 or 10) was examined and a x² -square test was used. Due to the problem of multiple comparisons, only changes for which *P*B0.005 were considered as significant. To note, due to the large sample size, even a relatively small change (e.g. from 7.2 to 7.6) was statistically significant. Such a change was also treated as practically significant.

A comprehensive analysis of the survey was performed by David Shinar and will be published separately. The main findings of the analysis (Hakkert et al., 1998) essential for the general project assessment, are as follows:

(a) since the project began, an improvement was identified in all the indices of perceived level of police activity (Table 2, Questions 1, 2, 4). The drivers felt that they saw more police forces on the interurban roads and that the risk of apprehension had increased;
(b) despite what was mentioned above, when the questions concerned the enforcement of specific rules, it was found (Table 2, Questions 8–14) that the police was perceived as enforcing them at the same degree or even less, in comparison with the 'before-the-project' period. In both rounds, the respondents said that only speeding, license violations and red-light running were enforced to a great extent (see percentage of values 8–10 in Table 2);
(c) as happened with the perceived enforcement level, most respondents attributed a lower or the same value to the risk of being detected for specific violations, when 'during the project' answers were compared with 'before the project' ones (Table 2, Questions 33–38). Among all the violations mentioned, only speeding was

associated with a high risk of apprehension by about half of respondents. Concerning crossing of a white separating line, close following, not-keeping right and unfastened safety belts, less than 20% of drivers estimated the risk as high. Hence, except for speed violations, most drivers believed that for majority of violations, their chances to be apprehended were rather low. This is an 'unpleasant' finding, especially in light of the increased enforcement declared by the police and even perceived by drivers, in general (see clause 'a' above). On the other hand, as was demonstrated later by the analysis of the police activity indices, it is a very realistic finding, reflecting actual enforcement conditions;

(d) a comparison between the values attained in areas with different enforcement intensity (according to the NTP plan of 700-project — see Section 1) demonstrated that the level of enforcement perceived in highpriority regions was mostly lower or, at least, not higher than in others. Table 3 illustrates the attitudes (13 out of 40) where a significant interaction (with *P*B0.05) was found between the time and the region of surveys. In this sense, for example, the perceived level of general enforcement (see Table 3, Questions 1, 2, 4) demonstrated an opposite-to-expected trend as a higher positive change of attitudes was found on low- or regularly-enforced rather than on highly-enforced roads. This led to the conclusion that whereas the media announcements assisted to note a higher police activity, just signposts and some police presence were not influential enough to convince drivers that increased enforcement indeed took place, in specific regions;

(e) in spite of the low risk of apprehension noted for specific violations (clause 'c') and the low level of enforcement perceived for most specific rules (clause 'b'), the risk values accepted for AA-priority region during the second round, were sometimes higher than for other regions. In this sense, a considerable difference between the AA- and other regions, also supported by a positive change of risk value during the project performance, was found for crossing of a white separating line, unfastened safety belts and the use of cellular phone (see Table 3, Questions 34, 37, 38).

4. Monitoring of police activity

As discussed in Alouda and Jayet (1995) and Jayet (1996) the quantitative characteristics of conventional police activity are not well documented. Only selected studies of police operations provided data concerning the actual level of enforcement activity on the roads such as the number of enforcement hours per day (Vaa, 1997) or number of site visits per week and number of offenses detected (Leggett, 1988). Usually, limited in-

Table 2 Comparison of drivers' attitudes in before the project ('Before') and during the project ('After') surveys

| Question | | Mean value of attitude | Percentage of respondents
who gave value 8–10 | | Trend | | |---------------------------------------------------------------------------------------------|--------|------------------------|--------------------------------------------------|-------------------|-----------------------|--| | | Before | After | Before | After | | | | Perceived enforcement levela | | | | | | | | 1. Are more Police forces seen now on interurban roads? | 7.34 | 7.76
PB0.0001 | 52.13 | 61.57
PB0.0001 | Positive | | | 2. Does the Police give more citations on interurban roads? | 7.22 | 7.58
PB0.0001 | 48.59 | 57.90
PB0.0001 | Positive | | | 4. Did the risk of apprehension by the Police increase for
the traffic rules' violators? | 6.49 | 6.93 | 38.13 | 44.16 | Positive | | | | | PB0.0001 | | P=0.003 | | | | Perceived enforcement of specific rulesb | | | | | | | | 8. 'Keeping right' | 4.97 | 5.01
P=0.740 | 18.55 | 16.16
P=0.128 | No change | | | 9. 'Keeping distance' | 4.57 | 4.87
P=0.004 | 14.24 | 13.34
P=0.528 | Unclear | | | 10. 'Traffic sign compliance' | 6.70 | 6.30
PB0.0001 | 41.24 | 32.59
PB0.0001 | A reduction | | | 11. 'Speed limits' | 7.73 | 7.46
P=0.002 | 63.70 | 56.85
P=0.0007 | A reduction | | | 12. 'non-crossing of white separating line while overtaking' | 6.81 | 6.85
P=0.662 | 45.13 | 45.49
P=0.863 | No change | | | 13. 'valid license and insurance papers' | 7.55 | 7.63
P=0.322 | 58.73 | 61.16
P=0.229 | No change | | | 14. 'non-violation of red traffic light' | 7.44 | 7.64
P=0.041 | 57.68 | 62.25
P=0.024 | No change | | | Estimated Risk of Apprehension for Specific Violationsc | | | | | | | | 33. Excessive speed | 6.63 | 6.34
P=0.008 | 45.96 | 40.27
P=0.005 | No change | | | 334. Overtaking through white separating line | 4.95 | 4.78 | 23.34 | 18.06 | A slight
reduction | | | | | P=0.147 | | P=0.002 | | | | 35. Not-keeping right | 4.51 | 4.02 | 15.57 | 12.49 | A slight
reduction | | | | | PB0.0001 | | P=0.032 | | | | 36. Not-keeping distance | 4.29 | 3.97 | 15.42 | 12.50 | A slight
reduction | | | | | P=0.004 | | P=0.041 | | | | 37. Safety belt unwearing | 4.98 | 4.78
P=0.092 | 23.49 | 19.50
P=0.019 | No change | | | 38. Manual use of cellular phone while driving | 5.26 | 5.36
P=0.441 | 29.40 | 27.60
P=0.339 | No change | |

^a The questions ended by 'in comparison with the previous year'.

formation is supplied as regards to the absolute level of enforcement, so that 'low' or 'heavy' police presence cannot be interpreted numerically for a comparison purpose, for instance (Ostvik and Elvik, 1991). This also left untreated a more general question of how the police forces plan their everyday activity and which indices serve or should serve them for this purpose.

For detailed consideration of the police activity, in the current study, a special information system was established. This included the data of policemen's shift activity reports of all the NTP subdivisions involved in the project — a monthly input of some 4500 records. The policemen's activity report provides details on patrol car locations, activity types and citation categories produced during the shift. Every month, having accomplished a long list of data checks, eight summary reports of police activity were produced. These covered three groups of indices: inputs (the number of police officers, patrolling vehicles and devices per site in a definite time interval); outputs (the level of actual police presence and the citations given); and the efficiency indices (the performance-against-plan ratios and utiliza-

^b The questions started as 'How strong does the Police enforce'.

^c The questions' form was 'To your opinion, what is the risk of apprehension for this violation, on the road you are driving on?'

Table 3 Drivers' attitudes in different enforcement regions^a

| Question
number | Time of
Survey | |--------------------|-----------------------|----- | | | 1 | Before the
project | 7.39 | | During the
project | 7.27 | 2 | Before the
project | 7.34 | | During the
project | 7.32 | 4 | Before the
project | 6.48 | | During the
project | 6.55 | 10 | Before the
project | 6.45 | | During the
project | 6.27 | 12 | Before the
project | 6.54 | | During the
project | 6.91 | 13 | Before the
project | 7.52 | | During the
project | 7.55 | 14 | Before the
project | 7.17 | | During the
project | 7.68 | 33 | Before the
project | 6.65 | | During the
project | 6.38 | 34 | Before the
project | 4.70 | | During the
project | 5.15 | 35 | Before the
project | 4.22 | | During the
project | 4.05 | 36 | Before the
project | 4.06 | | During the
project | 4.01 | 37 | Before the
project | 4.79 | | During the
project | 5.35 | 38 | Before the
project | 5.09 | | During the
project | 5.83 | Mean value of attitude, in
enforcement region | | | | | | --------|--------------------------------------------------|------------|--|--|--|--| | AA-priority | A-priority | B-priority | | | | | | 7.22 | 7.34 | | | | | | 8.23 | 8.28 | | | | | | 7.05 | 7.18 | | | | | | 7.66 | 8.02 | | | | | | 6.48 | 6.50 | | | | | | 7.35 | 7.31 | | | | | | 7.11 | 6.80 | | | | | | 6.31 | 6.33 | | | | | | 7.24 | 6.94 | | | | | | 6.89 | 6.74 | | | | | | 7.77 | 7.38 | | | | | | 7.53 | 7.92 | | | | | | 7.0 | 7.82 | | | | | | 7.72 | 7.6 | | | | | | 7.21 | 5.95 | | | | | | 6.41 | 6.22 | | | | | | 5.77 | 4.65 | | | | | | 4.46 | 4.31 | | | | | | 5.09 | 4.57 | | | | | | 4.08 | 3.86 | | | | | | 4.81 | 4.26 | | | | | | 4.05 | 3.82 | | | | | | 5.58 | 4.79 | | | | | | 4.32 | 4.09 | | | | | | 6.12 | 4.73 | | | | | | 4.68 | 5.14 | | | | |

^a See question definitions in Table 2. A significant interaction between survey time and enforcement region was found for all the attitudes in the table, with *P*B0.05 (ANOVA).

tion of resources), and illustrated the police activity in the form of daily and average monthly figures, with respect to road sections, police regions and the entire project area.

The intensity of the police enforcement over the 700-project is characterized by the following figures:

1. the number of patrol units within the project area during a regular weekday shift was about 60; during a similar weekend shift — more than 30; during a night shift — 9.5 and 12, accordingly (Fig. 24 );
2. the average monthly amount of the patrol units in all the shifts was some 4000 in the whole interurban area of which some 2700 in the project-area (70%);
3. on average, in about 28% of the shifts an electronic enforcement device was applied (Fig. 3): a laser speed gun-some 720 shifts per month; a radar speed meter (KR11) — about 170; photo radar camera (F6) — over 150 shifts per month; and infra-red speed and gap-following camera (Marom) — over 60;
4. on average, every road section of the 700-project (15 in sum) was patrolled for about 1800 h monthly, with a 'production rate' of 0.83 citation per shift hour, or 1.2 citation per actual enforcement hour. As to stationary versus moving enforcement, no estimates can be elaborated based on the police reports. In general, most violations (especially, speeding) are apprehended from a stationary vehicle or by a policeman who stands outside the car. Usually, several hours during a shift are spent on a moving patrol of the road. The other output indices were:
5. the average monthly amount of citations in the project area was more than 24 000 (some 82% of the total) — Fig. 4;
6. the productivity of a patrol unit in the project area was, on average, nine citations per shift (7.7 in the whole territory under the NTP responsibility); these do not include automatic citations, produced by F6 and Marom;
7. at the same time, the efficiency of using enforcement equipment was rather low: on average, about 5.3 citations per shift were produced with a laser speed gun and only 2.7, per shift with a radar speed meter (KR11).

An average of 82% performance-versus-plan ratio was attained during the project. (The best month was July — 93%. The ratio approached 100% in certain police regions, during several months). As the NTP vehicle fleet numbers some 160 cars and motorcycles and for a regular shift on all roads some 90–100 are usually

⁴ No before data are available as these were not being systematically handled by the police. An overall comparison of monthly vehicle-shifts, during the project, with specially collected data for January 1997, demonstrated that the former was 1.4–1.6 times higher than the latter.

Fig. 2. Patrol units per shift on the project roads.

Fig. 3. Monthly number of shifts with enforcement tools (day shifts only).

planned, the index of vehicle fleet usage during the project year was surprisingly low — not more than 50%. Similarly, the stock of devices for electronic enforcement was utilized, in a best case, for 60% of their capacity (laser speed guns), in the worst case — for 15% (radar speed meters), and in other cases, for 40– 45% (radar cameras). In general, most of performanceversus-plan ratios for enforcement tools were unstable and varied over a wide range during the project year. According to the above figures, one has to conclude that the NTP did not exhaust its resources during the project performance and still has a potential for more intensive operations. During the shifts on the project roads, about 70% of the time was dedicated to field enforcement, some 4% — to traffic directing, 2% — to accident treatment. The rest of the time (24%) was spent on other operations (training, missions, desk-

work, etc.) which are not a direct part of the actual project performance.

Among the offence reports given within the project area, some 18% were for speed violations; about 10% for safety belts not used; 4% for vehicle lights turned off. During the winter season, there was an increase of citations for lights turned off (when daytime running lights were obligatory), at the expense of citations for speed violations. Only about 16% of citations belonged to other moving violations (non-compliance to signs, dangerous overtaking, not-keeping lane, etc.) covering 24 kinds of maneuvers in sum. The remainder of citations (52%) was produced for technical and administrative violations, and unsafe vehicle conditions. As can be seen, less than half of the citations had a potential for direct influence on the road safety situation. On several road sections of the 700-project almost 70% of citations were of a 'technical' nature, whereas on only three sections out of 15, these citations presented less than half of the total amount. (The figures do not include the number of automatic citations as these are produced with a delay of up to 6 months).

A comparison between the data from the information system and the internal police summary-reports, during the project, discovered gaps in the numbers of patrol forces in the field, mainly in favor of the internal reporting. This raised questions concerning the manner of counting, accepted by the local police units, and the level of underreporting characteristic to policemen activity reports (the basis of the current study). Other defects revealed in policemen reports, were vehicle number faults; report duplications; speed citations given, seemingly, with no instrument; mismatch between the vehicle number and equipment, etc.

One of the apparent achievements of the 700-project was the monthly activity plans developed by the police. These plans contained everyday two-shift deployment of the police forces, i.e. vehicles and enforcement devices, on the project roads. However, the way of planning stayed unclear until the project completion because the performance was usually lower than the plan; a permanent discrepancy was found between the planned and the actual splits of regional patrol forces on the project roads; and the plan data never reflected the short-term (1-day) operations, declared by the police at the beginning of month. All this brings to a conclusion that inside the NTP there is a vital need for a management information system for planning and monitoring of everyday police activity.

As became evident from the evaluation of most input/output indices (see Figs. 2–4), the project intensity did not stay constant over the whole year. There was an initial period of increasing activity; a fall in September; some priority changes in October (it was a holiday month in Israel, and for various reasons, the police forces were instructed, mainly, 'to be present' on the roads but to give citations for severe violations only) and a return to routine in November, however, with a lower intensity in comparison with the initial project period. There were also other factors pointing to two periods in the project's performance: the intensive publicity accompanied the project only in the first 4 months; and in September, the central NTP subdivisions were restructured administratively. Finally, a detailed consideration of the police deployment — the force split between the project road sections, determined two periods: I, from April to August; II, starting from September 1997 till the project's completion. This was the first finding taken into consideration in the following accident analysis.

The second important point that influenced the accident analysis was the fact that not all the road sections were characterized by the same enforcement intensity during the project. Based on three criteria:

1. the number of patrol units per road-kilometer per month (the accepted threshold was four);
2. the rate of NTP forces allocated to the road section (5% was accepted as a threshold; it should be mentioned here that during the project year some 15% of units patrolled other roads which are outside the project area, and another 15% performed 'general' patrolling being not exactly responsible for a specific road section); and
3. the amount of pure enforcement hours on the section as opposed to the average value (above average was taken as considerable),

seven road sections out of 15, were chosen as having higher police presence during the project. Six of them had also the maximum of citations: more than 1500 per month, as a yearly average.

Fig. 4. Number of citations produced during the project.

The detailed consideration of the project performance revealed that shortly after the project beginning, the actual project territory was reduced, in comparison with the original plan, and comprised, in fact, only 600 km of roads. The accident analysis and the analysis of the police activity on road sections addressed this actual length of the project roads.

5. Accident analysis

In order to establish the NTP project's influence on safety, the trends in road accidents during the project year were analyzed. For evaluation of a treatment, two types of analysis are usually employed in observational studies (Hauer, 1997): the odds-ratio and a longitudinal (time series) analysis. The first method considers a relation between 'the average' after the treatment and before, for the treatment group, and compares it with a similar relation for a control (comparison) group. This does not take into account changes in accident trends and obviously depends on the adequacy of the selected control group to serve as a relevant comparison group. The second method analyses the time series of accidents without considering a comparison group, fitting separate models to the 'before' and 'after' periods, and then comparing the two models. In other words, a comparison is done between the fitted model based on the actual 'after' data with what is expected according to the fitted 'before' model, had the treatment not been applied. The disadvantage in the latter method is that it does not allow for confounding factors which may have affected the system in the 'after' period, and were not incorporated in the fitted model. Assessment studies of police enforcement projects, which evaluate their influence on road safety, usually apply one of these methods (Lamm and Kloeckner, 1984; Leggett, 1988; Elvik, 1997) or both of them but separately (Cameron et al., 1994; Hocherman et al., 1995). In order to cope with the deficiencies of each of the two methods, in the present study a modified technique was applied that combined the two methods, i.e. longitudinal models were fitted to the accident counts in the 'before' and the 'after' periods, for both the treatment and the control groups, followed by a comparison of the changes. Unlike the classical methods in which the odds ratio considers the average behavior 'before' and 'after', this method produces odds ratios for each time point (month) of the 'after' period.

For the accident analysis, all the roads under the NTP supervision were divided into eight groups, according to following three characteristics:

1. belonging to the 700-project (yes/no);
2. police activity level within the project area (high/low presence);
3. geographical zone (north, center, south).

The third characteristic was added as the geographical regions differ in their traffic patterns and, consequently, in the police activity modes. As outside the project area there was only one, non-project level of presence, overall, nine road groups were possible. Not nine, but eight groups were considered, as in the south there were two road groups only: 'non-project roads' and 'project roads with higher police presence'. The roads, which did not belong to the project, served as comparison groups in the corresponding geographical area.

Models for the 'before' and 'after' periods, were fitted for each road category. In this paper the results pertaining to severe accident counts are discussed in detail, since these were the most appropriate to the project objective. The same analysis was done also for the total injury accidents and for the number of severe casualties. A generalized linear model was fitted to the monthly accident counts, using the GENMOD procedure of SAS, assuming a Poisson distribution and allowing for over-dispersion. Each model includes a trend and seasonal components. The best fitted trend was a linear function on the log-scale, namely an exponential function in the original scale. Several alternative economic indices (such as private consumption) were examined to explain the trend, but none was found to improve the linear function (on the log scale). A possible explanation for this result is that the pattern of these indices did not resemble the accident pattern. The seasonal component was included as a dummy variable for each month. Based on the fitted model for the 'before' period for each month of the project period, the expected number of accidents, had there been no intervention, were predicted. Then, a model based on the actual data for the 'after' period, was also fitted. The monthly odds is the ratio of the predicted number of accidents according to the model based on the 'after' period and the forecast according to the model based on the 'before' period. These monthly odds were evaluated for all the roads (comparison and treatment groups).

The calculations described so far are analogous to the stage in the classical method of evaluating odds ('after' compared to 'before'). The next stage of evaluating odds ratio is required to take account of possible changes which were not necessarily caused by the intervention (project). Thus, for each month the odds in the treatment group were divided by the corresponding odds of the comparison group to obtain a monthly odds ratio. This odds ratio was expected to be significantly less than 1 had the project been effective. Indeed, in a quite substantial number of months, this ratio was smaller than 1. It should be noted that the calculation of confidence intervals for these ratios is not a trivial task and that the technical difficulties could be solved due to recently updated advanced software released by SAS (GENMOD). The gain (or loss in accident number

Fig. 5. Smoothing splines for the numbers of severe accidents on the road groups.

Table 4 Odds ratios and estimated 'gains' for severe accidents on the project road groups

Evaluation a period	After/before ratio (odds) for the project roads	After/before ratio (odds) for comparison- group roads	Odds ratio	Estimated 'gain'b	Observed accident countc
	Road group: north, higher police presence
I	1.75	1.29	1.35	−7.74	29
	(0.93; 3.27)	(0.90; 1.86)	(0.66; 2.79)	(−19.09; 15.67)
II	1.37	1.41	0.97	1.02	36
	(0.75; 2.50)	(0.99; 2.00)	(0.49; 1.95)	( −18.18; 39.48)
Whole year	1.50	1.37	1.10	−5.92	65
	(0.88; 2.54)	(1.00;1.86)	(0.60; 2.02)	(−33.88; 45.61)
	Road group: center, higher police presence
I	0.95	1.42	0.67	28.58	58
	(0.70; 1.29)	(0.94; 2.15)	(0.40; 1.12)	(−6.07; 86.64)
II	0.76	1.32	0.57	58.02	76
	(0.57; 1.02)	(0.89; 1.96)	(0.35; 0.93)	(5.52; 143.62)
Whole year	0.82	1.36	0.61	88.06	134
	(0.64; 1.06)	(0.96; 1.92)	(0.39; 0.93)	(9.85; 208.44)
	Road group: south, higher police presence
I	0.85	1.06	0.80	4.18	16
	(0.47; 1.53)	(0.69; 1.64)	(0.38; 1.66)	(−6.53; 26.50)
II	0.95	1.23	0.77	6.48	22
	(0.54; 1.70)	(0.80; 1.89)	(0.38; 1.59)	(−8.24; 36.67)
Whole year	0.91	1.17	0.78	10.74	38
	(0.55;1.51)	(0.80;1.70)	(0.42; 1.46)	(−12.24; 53.72)
	Road group: north, lower police presence
I	0.97	1.29	0.75	14.16	41
	(0.63; 1.48)	(0.90; 1.86)	(0.43; 1.31)	(−9.91; 56.26)
II	1.25	1.41	0.89	11.58	90
	(0.86; 1.81)	(0.99; 2.00)	(0.53; 1.47)	(−29.21; 79.39)
Whole year	1.14	1.37	0.83	26.51	131
	(0.82;1.59)	(1.00; 1.86)	(0.53; 1.31)	(−31.67; 118.13)
	Road group: center, lower police presence
I	0.79	1.42	0.56	10.56	13
	(0.31; 2.00)	(0.94; 2.15)	(0.20; 1.54)	(−4.66; 52.51)
II	0.65	1.32	0.49	19.58	19
	(0.27;1.60)	(0.89; 1.96)	(0.19; 1.31)	(−4.57; 83.76)
Whole year	0.70	1.36	0.52	30.38	32
	(0.32; 1.55)	(0.96; 1.92)	(0.22; 1.22)	(−5.94; 116.43)

^a The project periods: I (first) 5/97–8/97; II (second) 9/97–3/98.

due to the treatment) was expressed as the difference between the product of odds ratio times actual count and the actual count.

The data file for the analysis consisted of the accident records from January 1995 till March 1998 (the observed monthly counts of severe accidents are given in Appendix A). Fig. 5 displays the data, with smoothing splines that enhance the data pattern. To note, smoothing splines is one of several available techniques for smoothing scatter plots (accidents in our case). This method is used to highlight the typical response as a function of an exploratory variable (time in our case). The objective in using this method is to display the main trend without giving too much weight to the 'random noise' (Hastie and Tibshirani, 1990). (The name 'spline' refers to a draftsman's tool, which is a flexible rod used to interpolate smoothly points in a design drawing. The Smoothing Spline method is available as a part of any advanced statistical software such as SAS.) On Fig. 5, the time axis includes 39 points: t=1 for January 1995; t=39 for March 1998. The point t=28 corresponds to April 1997 — the project starting month, and t=33 to September 1997 — the beginning of the second project period. These two time points are indicated by vertical lines. In the fitting stage the data of April 1997 were excluded, since the project began on the second week of this month. Hence, this month could be assigned neither to the 'before' nor to the 'after' period.

^b 'Gain' corresponds to loss in the accident number due to the project.

^c The observed accident counts for the before period and for the comparison group roads are given in Appendix A.

Fig. 6. Odds ratios for severe accidents on the project road groups (mean values and 2s-boundaries).

As can be seen in Fig. 5, since the project started, an increase in severe accident numbers was observed, in most road groups, except for the project roads in the central and southern zones. However, the initial comparison of the 'during the project' accident counts with the 'before' period (Table 4, 'after/before ratio') revealed that none of the changes appeared to be significant. A further comparison of the changes observed for the project roads with those that occurred in the proper comparison groups, provided the following results (Fig. 6, Table 4 — 'odds ratio'):

1. a statistically significant reduction of severe accidents, as opposed to the comparison group, was found only on the highly enforced road sections in the center of the country, mainly, during the second project period;
2. a statistically significant (but hardly explained from the viewpoint of project performance) jump in severe accident numbers appeared in August in the northern group of highly enforced project roads;
3. no other statistically significant results were obtained but as can be seen in Table 4, in most cases, the mean value of the odds ratio is much less than 1 and, for the central and southern project roads, the mean values of the index of effectiveness (Hauer, 1997) — a ratio of 'after' and 'before', i.e. expected, accident counts — is also less than 1.

The supplementary analyses performed for all injury accidents and for the numbers of severe casualties (serious injuries and fatalities together) provided similar results. To note, separate consideration of fatalities did not bring a significant contribution to the findings as, due to scarce statistics, the confidence intervals for the odds ratio values were very wide. None of the project road groups demonstrated a statistically significant change of all injury accidents. Whereas, for both northern road groups, the mean value of odds ratio was higher than 1, indicating an increasing trend in accident counts, all the other project roads, in the center and in the south, were characterized by a decreasing trend in accident numbers in both project periods. A statistically significant reduction of severe injuries, as opposed to the comparison group, was found on the highly enforced central project roads on average, up to 40% reduction. In most cases, except for both project periods on the highly enforced northern roads and the second project period on the southern road group, the odds ratio was less than 1, indicating a reduction trend in severe casualties.

Thus, it can be concluded that, in spite of a general deterioration in safety on the interurban roads during the project year, the phenomenon was less tangible on the project roads, and this result ought to be, due to the concentrated police enforcement applied in this area.

6. Discussion and conclusions

The concentrated general police enforcement project took place for a whole year on most heavily-traveled interurban roads in Israel. The project was supposed to achieve a 10% reduction of severe accidents on the roads in focus and, concurrently, to improve the Traffic Police working modes.

The project results can be called rather moderate:

(a) In the drivers' behavior observations, during the first project months, a reduction of the violation rate was observed in several behavior types, such as compliance to 'yield' signs, proper turn performance, use of safety belts, keeping right on a dual-carriageway road, non-crossing of the white separating line on single-carriageway roads, whereas no change was found in compliance to 'stop' signs and in signaling while turning. At some directions of travel at the intersections, the violation rates of 'yield' and 'stop' signs and of proper signaling, stayed rather high, 30–40%.
(b) During the first project months, an improvement was identified in perceived general level of police activity, whereas a lower or the same risk of apprehension was attributed by drivers to most specific violations of the traffic rules. Four months after the project beginning, most drivers still believed that, except for speed violations, their chances to be apprehended were rather low. The differences found for the perceived enforcement level and the perceived risk of apprehension, between the enforcement regions with various priority levels, were rather subtle.
(c) The project did not attain its full purpose the reduction of severe accidents, on all the project roads, as a significant reduction of severe accidents (as opposed to the comparison group) was found only on one of the five project road groups. In general, since the police project started, an increase in accidents was observed on most interurban roads, including the northern parts of the project area. However, positively thinking, we can see that in four of the five project road groups the mean value of the odds ratio is much less than one. Thus, although the enforcement considered did not reduce the accident numbers during the project year, it seemed to be a deterrent factor for the increasing accident trend that appeared this year on the interurban roads.
(d) Despite the increase in productivity of the patrolling units during the project year, the efficiency of using the police resources remained rather low — not more than 50–60%, especially in the part concerning the application of electronic enforcement tools. Except for a change of the force deployment, no special activity modes were applied for the shift performance. The major part of citations was still of a 'technical' nature, not having a direct connection with drivers' behavior. The follow-up findings pointed out that the NTP did

not exhaust its potential in the project performance and needs a more flexible and more effective use of its resources.

In general, the positive changes that were seen in drivers' behaviors and attitudes should be attributed to the general deterrence effect (Cameron et al., 1996), which stemmed from a combined influence of the publicity campaign, somewhat increased police activity and the scale of the project. At least during the first project months, even those who were not caught for a violation still believed that the general risk of apprehension was considerable and behaved respectively. In the course of time, especially when the accompanying publicity died down, the drivers frequently could not appreciate that the special enforcement project indeed took place.

The current study of the 700-project and the study of the accompanying publicity campaign (mentioned in Section 2) were consistent with the following: the publicity influenced the public's outlook but, apparently, from the project beginning there was a gap between the publicity announcements and the field police activity. The low estimate given by the drivers already in June 1997 to the effectiveness of the police project and to the ability of the police to reduce accidents (as found by another study), was rather surprising, but indicated this phenomenon. In the current study, having accomplished the analysis of the police activity indices, we could see one of the reasons in that the police working modes did not change significantly during the project's performance.

When an enforcement project is evaluated, the accident and injury reduction is accepted to be the most meaningful indicator of the project effectiveness. However, as known, the police activity influences the road accidents not directly, but through a complex of relations and intermediate variables. Moreover, the police enforcement is only one of a whole gamut of factors affecting road safety, where part of them, such as weather or economic development, is out of control. Due to a wide spectrum of unmeasured influences, on the one hand, and, conversely, due to the relatively small number of accidents which produces large confidence intervals, it is often difficult to obtain definitive conclusions, especially within a limited time interval and large-scope project. Hence, the lack of significance in most statistical results of the current study is not surprising. Also, in drawing the conclusions, one should remember that statistical significance is not equivalent to practical significance. Then, the full results of analysis are of a great assistance in deriving lessons as to the project's achievements and faults. One more note relevant to this context is that the statistical method developed in the study, namely the combination of longitudinal and odds ratio methods, enabled to strengthen the findings, and to make them more definite than could be attained by each method applied separately. For example, the traditional odds ratio would not recognize a breaking of the downward accident trend on most project roads which concurred with the project's beginning. Conversely, a merely longitudinal method would not account for the general deterioration of safety in the rural area during the project year. It is also remarkable that a significant accident reduction was found for the central project roads, as opposed to the comparison-group roads and not on the step of initial after/before comparison.

The fact that during the project year the increase of productivity of the patrol units, in comparison with previous years (Hocherman et al., 1995), was observed, proves the efforts invested by the police in enhancement of the field enforcement. However, it is still about one citation per shift-hour, on average. Comparing this index with the amount of violations observed in drivers' behavior, at the same year: for instance, 30–60% of drivers exceed the speed limits if the traffic density enables this, i.e. there could be 200–400 violations per hour at some locations; some 30% of drivers do not comply with 'stop' and 'yield' signs, i.e. there are 20–60 violations per hour; one can see that the citations produced refer to a few percent of violations at most. From this, it follows that the present police activity modes, i.e. manual enforcement with a scattering of attention over all the violation types, are unacceptable to considerably affect any specific behavior. A similar conclusion stemmed from the analysis of drivers' attitudes (given by D. Shinar in Hakkert et al., 1998): to improve a specific behavior, it is necessary to create an impression of a specific enhanced enforcement; both the publicity and the enforcement should be focused on a small number of the same specific violations. This conclusion is also commensurate with the international experience saying that police operations usually succeeded when they treated a limited group of violations (e.g. Leggett, 1988; Noordzij and Mathijssen, 1991; Cameron et al., 1996; Vaa, 1997). Also, stemming from the latest reviews of enforcement practice (Managing Speed, 1998; OECD, 1999), intensive speed enforcement implies a wide application of automatic enforcement tools.

Beside the number of citations, another important indicator of police activity, frequently used in comparisons, is the enforcement factor or the police presence level, on the project roads. During the 700-project, the enforcement factor increased, on average, by 1.7 in comparison with before the project level (measured in January 1997). With only a 70% increase in activity, one could hardly expect a serious reduction in the accident numbers (Ostvik and Elvik, 1991; Elvik et al., 1997). At the same time, considering the roads where the 700-project did succeed (center, higher police presence), we did not see a 'jump' in the police presence on these roads during the project as opposed to before the project period, but the enforcement intensity on them all was high — more than four patrol units per roadkilometer per month. This brings us to a cautious conclusion that as to conventional enforcement, this level of police presence seems to be sufficient, for the roads in the center of the country, to moderate hazardous drivers' behavior and to contribute to local road safety. Analyzing field enforcement activity on the project roads, it was also found that positive accident trends appeared on the roads with a higher productivity of patrol units, as opposed to the remainder of the roads of the same region, and on those, where more attention was given to moving (speeding, dangerous maneuvers) and not to technical or administrative violations.

Reviewing the 700-project development, some general reasons of its limited success can be conceived of. First, the basic level of traffic law enforcement in Israel forms more than 160 patrol officers per million registered drivers and about 12 per million vehicle-kilometers traveled, which is rather high in comparison with many European countries and the USA (see, for example, OECD, 1999). This means that, at least during the 1990s, traffic law enforcement in Israel was given relatively high priority within the National Police forces and, then, the 700-project did not present a major change neither in the traffic police role nor in its image within the driving community. Second, the Traffic Department within the Israeli Police which is intended to conduct the strategic command over the traffic police and to attain a professional specialization of its forces, was established only in the middle of 1997. Varying enforcement tactics and more tight evaluation of operations performed became the real issues within the traffic police only following this event, i.e. in 1998–1999. Previously, a 'more of the same' approach was mostly employed by the regional police subdivisions. The 700-project fell on a 'transition period '' and was, in a sense, a victim of the old approach, in spite of high motivation and apparent efforts undertaken by the police subdivisions, at least during the first project months (that was seen, for example, in enhanced productivity of patrol units and increasing number of shifts within the project area). Third, the project was initiated in a short time: the first announcement appeared at the end of January 1997, the project started on April 6, where no specific basis was developed by the police, as to the target setting, force deployment or enforcement tactics involved. The assessment team was attached only at the last step before the project launch. Later, during the project performance, the plans to carry out several experiments with specific enforcement tactics (e.g. different shares of marked/unmarked vehicles; 'the fist' several times increased enforcement on several roads) were not realized due to bureaucratic reasons similar to those which failed the repeated speed measurements. All this left mostly an observational and not an intervention role to the research team. Fourth, as known, it is important to create conditions in which traffic enforcement is not vulnerable to day-to-day disturbance (OECD, 1999). This rule was not strictly working during the 700-project as sometimes whole regional subdivisions were sent to perform escort or security rather than enforcement duties. In general, we would say that in the 700-project failures reflected both sluggishness of the bureaucratic structures, which should have coordinated the interaction of the parties involved (the police, the publicity, the measurement and assessment teams), and an objective lack of effective large-scale strategies and proper tactics for management of conventional enforcement. There is a general agreement that a good enforcement practice should be based on sound accident analysis and on a scientifically supported insight of enforcement effectiveness (OECD, 1999). However, there is still a gap when a project target, even properly based, should be converted into everyday enforcement activity. In this context, certain reporting discrepancies, found out by the 700-project information system, indicate that management of the police forces also needs a more founded, systematically updated basis.

To conclude, the approach applied to the 700-project when the available police forces are spread over a large road network, acting in the same routine way for the whole year, cannot be considered as effective enough. It is reasonable to conclude that focused activity that is shorter in time, more concentrated in area/enforcement subject and more flexible in performance of police operations, will gain advantage over the 700-project results.

Appendix A. The observed counts of severe accidents for the road groups considered

| Year | Month | | Non-project road groups | | Project road groups | | | | | | |------|-------|-------|-------------------------|-------|----------------------------|--------|------------------------|--------|-------|--| | | | North | Center | South | Lower police pres-
ence | | Higher police presence | | | | | | | | | | North | Center | North | Center | South | | | 1995 | 1 | 33 | 19 | 11 | 17 | 4 | 6 | 11 | 5 | | | | 2 | 18 | 9 | 15 | 12 | 5 | 11 | 9 | 3 | | | | 3 | 22 | 9 | 13 | 15 | 5 | 9 | 12 | 4 | |

	4	25	7	14	13	3	8	11	4
	5	23	12	22	12	5	8	10	3
	6	32	9	15	20	3	8	13	7
	7	26	11	22	16	3	12	13	9
	8	31	8	12	16	2	1	11	5
	9	25	13	10	13	0	3	8	3
	10	25	12	19	9	5	5	8	4
	11	25	11	17	7	2	5	13	4
	12	30	14	13	16	3	3	15	5
1996	1	28	10	14	10	4	8	13	7
	2	29	9	16	9	4	9	7	2
	3	20	12	9	15	1	4	11	7
	4	18	4	17	15	2	7	9	3
	5	22	17	15	11	0	4	14	5
	6	27	8	21	11	4	6	10	9
	7	27	12	14	8	5	4	13	4
	8	29	7	21	7	6	4	16	3
	9	11	11	7	15	5	5	13	5
	10	22	12	10	13	4	6	16	4
	11	23	6	13	10	7	7	12	3
	12	21	9	4	14	0	6	16	4
1997	1	18	8	11	12	2	6	9	2
	2	15	5	19	14	3	4	17	1
	3	19	6	10	10	0	3	12	5
	4	22	9	12	11	1	3	9	2
	5	21	14	14	11	2	2	18	4
	6	30	8	19	6	6	9	10	3
	7	34	13	18	10	4	9	12	5
	8	26	7	17	14	1	9	18	4
	9	27	6	19	10	3	4	13	3
	10	11	5	15	7	0	3	9	4
	11	24	7	13	18	1	1	10	3
	12	36	10	10	13	5	9	14	4
1998	1	29	9	16	14	3	7	11	4
	2	17	18	16	12	1	7	11	2
	3	23	13	7	16	6	5	8	2

References

Alouda, A., Jayet, M.-C., 1995. Police control and driving offences: a space-time and socio-demographic analysis of police control enforcement. Proceedings of International Conference 'Road Safety in Europe and Strategic Highway Research Program', Lille, France; VTI konferens 2A, Part 4, pp. 101–113.
Bjornskau, T., Elvik, R., 1992. Can road traffic law enforcement permanently reduce the number of accidents? Accident Analysis and Prevention 24, 507–520.
Blackburn, R.R., Gilber, D.T., 1995. NCHRP Synthesis of Highway Practice 219: Photographic Enforcement of Traffic Laws. Transportation Research Board, National Academy Press, Washington, DC.
Cameron, M., Newstead, S., Vulcan, P., 1994. Analysis of reductions in Victorian road casualties, 1989 to 1992. Proceedings 17th ARRB Conference, Part 5, Australian Road Research Board, pp. 165–182.
Cameron, M., Newstead, S., Gantzer, S., 1996. Effects of enforcement and supporting publicity programs in Victoria, Australia. Proceeding of International Conference on Traffic Safety on Two Continents, Prague, Czech Republic, VTI konferens 4A, part 4, pp. 244–253.
De Waard, D., Rooijers, T., 1994. An experimental study to evaluate the effectiveness of different methods and intensities of law enforcement on driving speed on motorways. Accident Analysis and Prevention 26, 751–765.
Elvik, R., 1997. Effects on accidents of automatic speed enforcement in Norway. Transportation Research Record 1595, Transportation Research Board, Washington, DC, pp. 14–19.
Elvik, R., Borger-Mysen, A., Vaa, T., 1997. Trafikksikkerhekshandbok (Traffic Safety Handbook). Institute of Transport Economics, Oslo, Norway.
Hakkert, A. S., Yelinek, A., Efrat, E., 1991. Police surveillance methods and police resource allocation models. In: Koornstra, M.J., Christensen, J. (Eds.) Enforcement and Rewarding: Strategies and Effects, SWOV, Leidschendam, The Netherlands, pp. 98–101.
Hakkert, A.S., Gitelman V., Cohen, A., Doveh, E., Umansky, T., Shinar, D., 1998. A Follow-up Study of a New Deployment of the National Traffic Police in 1997 — Focused Police Enforcement. Research Report No. 98-268, Transportation Research Institute, Technion, Israel (in Hebrew).
Hastie, T.J., Tibshirani, R.J., 1990. Generalized Additive Models. Chapman and Hall, New York.
Hauer, E., 1997. Observational Before–After Studies in Road Safety. Pergamon, Oxford.
Hocherman, I., Zaidel, D., Hakkert, A. S., 1995. Monitoring traffic enforcement effectiveness on a national scale. Proceedings of International Conference 'Road Safety in Europe and Strategic Highway Research Program', Lille, France; VTI konferens 1A, Part 1, pp. 247–268.
Holland, C.A., Conner, M.T., 1996. Exceeding the speed limits: an evaluation of the effectiveness of a police intervention. Accident Analysis and Prevention 28, 587–597.
Jayet, M.-C., 1996. Traffic law enforcement in built-up areas: what do we learn from offences reported by the police? Proceeding of International Conference on Traffic Safety on Two Continents, Prague, Czech Republic; VTI konferens 4A, Part 4, pp. 168–183.
Lamm, R., Kloeckner, J.H., 1984. Increase of traffic safety by surveillance of speed limits with automatic radar devices on a dangerous section of a German autobahn: a long-term investigation. Transportation Research Record 974, Transportation Research Board, Washington, DC, pp. 8–16.
Leggett, L.M.W., 1988. The effect on accident occurrence of longterm low-intensity police enforcement. Proceedings 14th ARRB Conference, Australian Road Research Board, Part 4, pp. 92– 104.
Makinen, T., 1988. Enforcement studies in Finland. In: Rothengatter, T., de Bruin, R. (Eds.), Road User Behavior: Theory and Research. Van Gorcum, the Netherlands, pp. 584–588.
Makinen, T., Beilinson, L., Salusjarvi, M., 1991. Traffic enforcement strategies and tactics. Presented at 2nd International Conference on New Ways and Means for Improved Road Safety and Quality of Life, Tel-Aviv, Israel.
Managing Speed, 1998. Review of current practice for setting and enforcing speed limits. Special Report 254, Transportation Research Board, National Academy Press, Washington, DC.
Nilsson, G., 1991. Speed limits, enforcement and other factors influencing speed. In: Koornstra, M.J., Christensen, J. (Eds.), Enforcement and Rewarding: Strategies and Effects. SWOV, Leidschendam, the Netherlands, pp. 46–50.
Noordzij, P.C., 1996. Progress in the field of urban traffic law enforcement. Proceeding of International Conference on Traffic Safety on Two Continents, Prague, Czech Republic, VTI konferens 4A, Part 4, pp. 238–242.
Noordzij, P.C., Mathijssen, M.P.M., 1991. Police enforcement and road user behavior. In: Koornstra, M.J., Christensen, J. (Eds.) Enforcement and Rewarding: Strategies and Effects. SWOV, Leidschendam, the Netherlands, pp. 105–107.
OECD, 1999. Enforcement. Safety Strategies for Rural Roads. Road Transport and Intermodal Research, Chapter 5. Organisation for Economic Co-operation and Development, IRRD No 491006, Paris.
Ostvik, E., Elvik, R., 1991. The effects of speed enforcement on individual road user behavior and accidents. In: Koornstra, M.J., Christensen, J. (Eds.) Enforcement and Rewarding: Strategies and Effects. SWOV, Leidschendam, the Netherlands, pp. 56–59.
Rothengatter, J.A., 1991. Normative behavior is unattractive if it is abnormal: relationships between norms, attitudes and traffic law. In: Koornstra, M.J., Christensen, J. (Eds.) Enforcement and Rewarding: Strategies and Effects. SWOV, Leidschendam, the Netherlands, pp. 91–93.
Shinar, D., Mcknight, A.J., 1985. The effects of enforcement and public information on compliance. In: Evans, L., Schwing, R.C. (Eds.) Human Behavior and Traffic Safety. Plenum Press, New York, pp. 385–419.
Staysafe-5, 1985. Appropriate Strategies for Police Traffic Law Enforcement. Parliament of New South Wales, Joint Standing Committee on Road Safety, NSW, Australia.
Vaa, T., 1997. Increased police enforcement: effects on speed. Accident Analysis & Prevention 29, 373–385.
Zaal, D., 1994. Traffic Law Enforcement: A Review of the Literature. Report No.53, Accident Research Centre, Monash University, Australia.
Zaidel, D.M., Hocherman, I., Hakkert, A.S., 1994. Evaluation of a National Traffic Police Force, Transportation Research Record 1401, Transportation Research Board, Washington, DC, pp. 37– 42.

hakkertevaluationeffectsdriverbehavior2001.md 75 KB 履歴 Raw

status: converted

4. Monitoring of police activity

5. Accident analysis

6. Discussion and conclusions

References

hakkertevaluationeffectsdriverbehavior2001.md 75 KB

履歴 Raw