The Speed Limits Debate: Is Effective A Temporary Change? The Case Of Spain.

econstor Make Your Publications Visible. A Service of zbwLeibniz-Informationszentrum Wirtschaft Leibniz Information Centre for Economics Castro-Nuno, Mercedes; Castillo-Manzano, José I.; Pedregal-Tercero, Diego J. Conference Paper The Speed Limits Debate: Is Effective A Temporary Change? The Case Of Spain. 53rd Congress of the European Regional Science Association: "Regional Integration: Europe, the Mediterranean and the World Economy", 27-31 August 2013, Palermo, Italy Provided in Cooperation with: European Regional Science Association (ERSA) Suggested Citation: Castro-Nuno, Mercedes; Castillo-Manzano, José I.; Pedregal-Tercero, Diego J. (2013) : The Speed Limits Debate: Is Effective A Temporary Change? The Case Of Spain., 53rd Congress of the European Regional Science Association: "Regional Integration: Europe, the Mediterranean and the World Economy", 27-31 August 2013, Palermo, Italy, European Regional Science Association (ERSA), Louvain-la-Neuve This Version is available at: http://hdl.handle.net/10419/123873 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. www.econstor.eu 1 THE SPEED LIMITS DEBATE: IS EFFECTIVE A TEMPORARY CHANGE? THE CASE OF SPAIN. AUTHORS José I. Castillo-Manzano ([email protected]). University of Seville (Spain) Mercedes Castro-Nuño ([email protected]). University of Seville (Spain) Diego J. Pedregal-Tercero ([email protected]). University of Castilla-La Mancha (Spain) ABSTRACT Nowadays, speeding is one of the most relevant problems for traffic safety and most resistant to change in motorized countries. The key instruments in Speed Management Policy are speed limits. This road safety strategy is often established or changed, in order to save fuel during periods of rising prices. However, the relationship between speed limits and traffic accidents, is a topic widely discussed by researchers, and there seems to be some consensus about "speed kills." By applying advanced time series models of unobserved components, our study investigates the impact of a temporary reduction of maximum speed limits, implemented in Spain in 2011, in terms of fuel consumption and fatalities. Our analysis shows that this measure caused a positive effect, although with a limited statistical significance, on fuel consumption and a discrete reduction in road mortality. The costs associated with this temporary change seem to explain the discrepancies between these estimates and the forecasts that initially held the Spanish government. Keywords: Speed Management Policy, speed limits, road safety, fuel consumption, Unobserved Components Models. JEL Codes: C320, I180, R410. Acknowledgements This work was funded by a grant (reference 1671/0085) from the Spanish Road Traffic Directorate General (DGT: Dirección General de Tráfico), during 2012. The authors would like to thank the experts of DGT for their invaluable support and contribution. 2 1. INTRODUCTION. Over the past 50 years, citizens in motorized countries have benefited from the manufacture of vehicles capable of ever increasing speeds (ECMT, 2006). The speed of road transport has contributed to countries’ economic development and has increased wellbeing and the quality of life because of travel time savings (GRSP, 2008); this, according to Metz (2008), helps generate productivity gains and reduce opportunity costs in terms of user time. Nevertheless, speed also has very adverse impacts in the form of energy consumption, air pollution, noise emissions, uncontrolled urban growth and, above all, road traffic accidents (with negative effects on casualties and economic damage) as pointed out by ECMT (2006). Nowadays, excessive and inappropriate speed 1 is one of the biggest road safety problems (Elvik, 2010b; Wegman & Aarts, 2006), for both rich and highly motorized countries (Elvik, 2010a) and developing countries (Afukaar, 2003). However, despite speeding being a widespread issue and everybody being convinced that “speed kills” (GRSP, 2008), it is widely tolerated and, in the words of Elvik (2010a), “one of the road safety problems most resistant to change”. Most governments regard speeding as a priority in the context of road safety strategies, such as Vision Zero in Sweden, Sustainable Safety in the Netherlands and Safe System in Australia, where a range of tools are applied for developing an effective Speed Management Policy (GRSP, 2008). However, the key tool for speed management, and that which has been most widely addressed in the literature for decades, is speed limits (Ritchey & Nicholson- Crotty, 2011). And, despite the fact that there are some roads in motorized countries where no speed limits (i.e., German Autobahns) are in force, the need for legal speed limits on all types of roads is very widely recognized and commonly legitimated by the fact that drivers’ choices of speed may not always be perfectly rational from a social perspective (in Elvik, 2012) and may be strongly influenced by how fast others are driving (Haglund & Åberg, 2000). Following ERSO (2006) and SWOV (2010), speed limits must also be safe and credible, reflecting the road design characteristics and environment and traffic composition. A number of research studies have been carried out on the implementation and effectiveness of the socalled Variable Speed Limits (VSL) or Dynamic Speed Limits, widely used in United States 1 According to ECMT (2006) terminology, we can define excessive speed as a speed over the legal speed limit and inappropriate speed as a speed much too high for the road, and the weather and congestion conditions, but within the legal speed limit. Excess speed covers both terms, both excessive speed and inappropriate speed. 3 and other safe European countries, such as Germany or UK, to warn drivers to adjust their speed according to the specific road situation. Since they became widespread at the beginning of the nineteen-seventies (Elvik & Vaa, 2004) speed limit strategies appear to have been linked not only to the goals of traffic safety, such as controlling speeding and reducing road accidents, but also very frequently form part of broader policies with environmental, health or economic purposes, such as reductions in fuel consumption for less foreign energy dependence during times of increased gasoline prices and reducing GHG emissions and their health costs. After an overview of the academic literature (see the meta-analyses and the reviews by Aarts & Van Schagen, 2006; Elvik, et al., 2004; Finch et al., 1994; McCarthy, 2001; Wilmot & Khanal, 1999), we find that governments may initially use speed limit laws with economic goals for saving gasoline and diesel consumption with road safety as a secondary objective. However, the relationship between speed limits and traffic safety is an issue that has been widely addressed by researchers worldwide, especially in the U.S. (Albalate & Bel, 2012; Dee & Sela, 2003; Forester et al., 1984; Friedman et al., 2009; Lave, 1985; Retting & Teoh, 2008). There are also studies for European countries (Burns et al., 2001; Johansson, 1996; Peltola, 2000; Richter et al., 2004), for Australia (Sliogeris, 1992) and, more recently, for Asia (He et al., 2012; Wong et al., 2005) and Africa (Afukaar, 2003). Since the initial estimates obtained by Solomon (1964), there has been a degree of consensus on speed having a significant effect on road safety (with certain causality according to Elvik, 2012) in the sense that both accident incidence and accident severity are expected to increase with higher speed limits (Ashenfelter & Greenstone, 2004). However, “…despite years of research, there is still no clear consensus in the literature on the impact that speed limit laws have on traffic fatalities” as pointed out by Ritchey & Nicholson-Crotty (2011). How changes in speed limits affect actual driven speeds and the consequent estimation of their effects on road accidents is a controversial topic addressed from a wide range of focuses: the influence of the individual speed chosen by the driver on the risk of a crash (well-known are the Power Model 2 by Nilsson, 1982, 2004 and its evaluations made by Elvik et al., 2004; Elvik, 2009; Hauer & Boneson, 2008); the influence of speed differences 3 on the risk of a 2 The original Power Model introduced by Nilsson (1982), explains the relationship between speed and crash rate as a power function: the crash rate increases more quickly when the speed increases and vice versa. For example, a 5% drop in mean speed caused by a reduction in the speed limits would lead to a 20% fall in fatal accidents and a 10% fall in accidents with injuries. Subsequent validations showed that the effect on rural roads was relatively greater than on urban roads (SWOV, 2012). 3 According to this approach, large speed differences between vehicles (speed variance) increases the likelihood of an accident. In addition, drivers driving much faster than the average driver have a higher accident risk 4 crash (with the first studies conducted by Solomon, 1964 and Cirillo, 1968, and the modern reformulations by Lave, 1985 and Kloeden et al., 2001). Bearing in mind all the foregoing considerations, we understand that this paper helps to shed light on this controversy. To be more specific, we analyze a recent temporary change to maximum speed limits allowed on free-public and toll-private dual carriageways and motorways in Spain between the months of March and June, 2011, which was put in place with the primary objective of obtaining savings in fuel consumption. This lowering of speed limits was the most striking measure in the so-called Energy Efficiency and Saving Plan passed by the Spanish Government to counteract the effects that the high price of oil was having on the weakened Spanish economy at that time. This analysis is appropriate in as much as, firstly, although speed limit laws have existed in Spain since the mid-nineteen-seventies, to date there has still been no evaluation of their effects, neither of the actual driven speed, or on traffic safety, or in terms of fuel consumption or environmental impact. Secondly, despite the short period of time that it was in force, only three and a half months, limiting speed like this resulted in intense debate in Spain on the pertinence and real effectiveness of speed limits: in public opinion, between environmental associations, accident victims’ associations, and vehicle manufacturers and in all the media. The core question that surrounded this debate and that this paper attempts to answer is whether it is worth changing a country’s speed limits temporarily with the major costs this entails in order to save fuel. To shed light on this debate, this study analyzes the impact of limited changes in the Spanish speed limits measured by road accident indicators and fuel consumption. In order to isolate the impact of the provisional reduction in speed limits, a method based on advanced time series models was used, of the discrete linear time transfer function type, with multiple explanatory variables. In the line of other preceding studies, such as Balkin & Ord (2001) in the U.S. and Johansson (1996) in Sweden, this methodology answers the need stated by Dee & Sela (2003) and Shafi & Gentilello (2007) of pursuing findings that control potential confounding factors and eliminate the biases that may be due to unobserved or specific aspects of traffic safety, such as other simultaneous policies implemented. The paper is structured in the following way: following this introduction, Section 2 explains the data, variables and methodology used; the findings are stated and discussed in (“variance kills” as stated by Dee & Sela, 2003); it is not yet evident that this is also the case for the slower driver, and neither has any clear relationship been established to date to quantify the effect of speed differences and the crash involvement rate (ERSO, 2006). 5 Section 3 and the conclusions and resulting policy implications are analyzed in Section 4. Finally, we include the references. 2. DATA AND METHODS. The data used to measure the effect of the changes in speed limits in Spain can be divided into three groups that will be used later in the estimated models. A) Endogenous variables: namely fatalities in road accidents (FATAL in later table) and gasoline and diesel consumption for transport (CONSUMPTION). In order to use consistent time series that were as long as possible, we used the definition of deaths within the first 24 hours after the accident, instead of the Vienna Convention definition (30 days after the accident). The available time series span from January 1998 to August 2012 (Source: Spanish National Statistics Institute (INE) and DGT). B) Dummy exogenous variables: The most important, with their definitions, are: b.1) SPEED: takes into account the change in speed limits on highways (from 7th March 2011 to 30th June 2011). Several options have been tried out empirically, but the final version is one step over the whole period with a value of 75% in the first month, in order to take into account the fact that enforcement took place after the first week in March. b.2) EASTER: Traffic campaigns around this vacation period are especially intense in Spain. In fact, authorities launch special police operations to minimize problems on the roads. Accordingly, the moving Easter festival variable is defined by assigning different weights to the days in question depending on the expected traffic density (these weights must add up to one). b.3) TRADING: The number of trading days in a month in excess of weekend days, assuming that in each week there should be 5 working days and two days at the weekend. For each month this variable takes a value that equals the number of working days minus 2.5 times the weekend days. b.4) LEAP: Dummy variable to take into account the effect of 29-day Februaries. b.5) Two legal changes: firstly, the introduction of the Penalty Points driving license system in 2006 (PPS; Castillo-Manzano et al., 2010), which will be modeled as a transitory change in accordance with Butler et al (2006) and Farchi et al (2007) on the Irish and Italian cases, respectively. Secondly, a dummy variable has been included to 6 estimate the effects of the 2007 Spanish Penal Code reform. To be more precise, although the reform came into effect in December 2007, its effects started to be felt earlier, in November 2007 (NOV07), given the huge impact that the passing of the Bill through Parliament had on the media (Castillo-Manzano et al., 2011). b.6) There are other outliers, often related to bad weather conditions (like JAN06) and other causes that have been detected by statistical tools (September and October 2000). In the last two cases this was perceived as being due to the truck drivers’ and retailers’ strike of October, 2000. The procedure followed to look for outliers of this type consisted of selecting the residuals out of four times standard deviation and including them as potential candidates in the models under different specifications (LS, TC as explained above or additive outliers AO for sudden changes that affect just one observation). The outliers are included in final models with the specification that provides the best fit when they result statistically significant. C) Other exogenous variables: we assume that gasoline and diesel consumption and the number of fatalities depend on a set of common causes, of which the most important are gasoline prices and the level of economic activity (or economic cycle or income, see e.g., Castillo-Manzano et al., 2010 and García-Ferrer et al., 2007). The price of gas consumption used in this paper is approximated by the Brent oil price measured in Euros (variable PRICE in table). Although some statistics on gasoline and diesel consumption prices for transport were available, we preferred Brent prices because they represented a very good approximation and mainly because the time series is longer. Income is approximated by economic activity in the Spanish economy, measured by the Activity Synthetic Index (ASI) estimated by the Spanish Ministry of the Economy (http://serviciosweb.meh.es/apps/dgpe/default.aspx). The models used in this paper are of the multivariate Unobserved Components (UC) model class that allow for a time series to be broken down into economically meaningful, though unobserved, components, see equation (1). ttttt vSTz DI (1) t z , t T , t S and t v denote the endogenous time series and trend, seasonal and irregular components, respectively. t DI measures the effects of explanatory variables in matrix t I through a linear regression model. 7 One appropriate set up in which the UC analysis may be carried out is the State Space framework, in which the dynamic system is split into two types of equations, i.e., State and Observation equations. Discrete-time, stochastic State Equations reflect all the dynamic behavior of the system by relating the current value of the states to their past values as well as to the deterministic and stochastic inputs, while Observation Equations define how the state variables are related to the observed data (as a matter of fact, equation (1) is the observation equation of the UC model, see below). There are a number of different formulations of these vector-matrix equations, but the one favored here is as follows: Equationsn Observatio: Equations State: 1 tttt tttt vz DIHx wΓIΦxx (2) where t x is an n dimensional stochastic state vector; t I is a k dimensional vector of dummy exogenous variables; t w and t v are an n and scalar dimensional vectors of Gaussian system disturbances, i.e., zero mean white noise inputs with covariance matrix Q and R and independent of each other; and Φ , Γ , H and D are the so called system matrices, some elements of which are known while others need to be estimated. Given model (1), the well-known Kalman Filter (KF, Kalman, 1960) produces the optimal estimates of the first- and second-order moments (mean and covariance) of the state vector, conditional on all the data in a sample as it minimizes the mean squared errors (MSE). An algorithm that is used in parallel with the KF but which is not as well-known in certain contexts is the fixed interval smoothing (FIS) algorithm, which allows for an operation similar to that of the KF but with a different information set. The estimation of the unknown parameters in the system matrices RQDHΓ and , , , , may be tackled in several ways. Maximum likelihood (ML) is the most common because of its good theoretical properties. Under the Gaussian assumption, the loglikelihood function can be computed using the KF via ‘prediction error decomposition’ (see details in Harvey, 1989; Pedregal and Young, 2002). The unobserved components model in equation (1) fits naturally in the SS framework in equation (2), since the observation equations show the breakdown of the time series into its components, and the state equations specify the components dynamics. A description of the 8 full SS system for a simplified case with just one input variable is shown in equation (3) (see Harvey (1989), Pedregal and Young (2002)). tttttttt t tt t t vIbvfSTz w w w w w I baaf S S S F T 1 1 2 * 1 1 * 0 0 111 2 11 11 1 1 2 * 1 1 1 1010101 0 0 0 0 0 0 00000 0cos0000 00cossin00 00sincos00 000010 000011 xI xx           (3) A comparison of systems (2) and (3) enables system matrices to be identified. New elements other than the previous ones appear in equation (3): t F is the trend ‘slope’ or trend rate of change; it S ( 2/,,2,1 Pi  ) are the seasonal harmonics of frequencies Pi i2 , whereby it P it SS 2/ , with P being the fundamental frequency (12 observations per year in the case of monthly data with annual seasonality); * it S ( 6,,2,1 i ) are additional blocks of states necessary for the definition of seasonal terms. TF effect specification deserves further comment. All TF considered here are of order one, since only outlier corrections are considered. The general formulation of a single TF is given in equation (4), where B is the lag operator, so that mtt mxxB . pt p p pt I Ba b f1 (4) For Transitory Change outliers (TC), 0 p a , i.e., the effect disappears after some time. Additive outliers (AO) implies 0 p a , i.e., the effect is observed in just one observation. Finally, Level Shifts (LS) means that 1 p a , i.e., the effect is permanent. The extension of system (3) to accommodate additional TF terms or linear regression terms is straightforward. This methodology is implemented in the MATLABTM platform, in the SSpace toolbox (Pedregal and Taylor, 2012), which will subsequently be used for model estimation. 15 Lave, C. (1985). Speeding, coordination, and the 55 mph limit. American Economic Review, 75, 1159-1164. Lave, C., Elias, P. (1994). Did the 65 mph speed limit save lives?. Accident Analysis and Prevention, 26, 49-62. Litman, T.A., Doherty, E. (2009). Transportation Cost and Benefit Analysis: Techniques, Estimates and Implications. 2nd Edition. Victoria Transport Policy Institute. McCarthy, P. (2001). Effect of speed limits on speed distributions and highway safety: a survey of recent literature. Transport Reviews, 21, 31-50. Metz, D. (2008). The myth of travel time saving. Transport Reviews, 28, 321-336. Nilsson, G. (1982). The effects of speed limits on traffic accidents in Sweden. In: Proceedings of the international symposium on the effects of speed limits on traffic accidents and transport energy use, 6-10 October 1981, Dublin, OECD. Paris, 1-8. Nilsson, G. (2004). Traffic safety dimensions and the power model to describe the effect of speed on safety. Lund Bulletin, 221. Lund Institute of Technology, Lund. Sweden. Peden, M., Scurfield, R., Sleet, D., Mohan, D., Hyder, A.A., Jarawan, E., Mathers, C. (2004). World Report on Road Traffic Injury Prevention. World Health Organization (WHO). Geneva, Switzerland. Pedregal, D. J., Young, P. C. (2002). Statistical approaches to modelling and forecasting time series. In M. P. Clements, & D. F. Hendry (Eds.). A companion to economic forecasting. 69– 104. Oxford Blackwell. Pedregal, D.J. and Taylor, C.J. (2012). SSpace: A Flexible and General State Space Toolbox for MATLAB, in L. Wang and Garnier, H. (eds.), System Identification, Environmental Modelling, and Control System Design, Springer Verlag, London. Peltola, H. (2000). Seasonally changing speed limits: effects on speeds and accidents. Transportation Research Record: Journal of the Transportation Research Board, 1734, 46- 51. Retting, R.A., Teoh, E.R. (2008). Traffic speeds on interstates and freeways 10 years after repeal of national maximum speed limit. Traffic Injury Prevention, 9, 119-124. Richter, E. D., Barach, P., Friedman, L., Krikler, S., Israeli, A. (2004). Raised speed limits, speed spillover, case-fatality rates, and road deaths in Israel: a 5-year follow-up. American Journal of Public Health, 94, 568-574. Ritchey, M., Nicholson-Crotty, S. (2011). Deterrence theory and the implementation of speed limits in the American States. Policy Studies Journal, 39, 329-346. 16 Shafi, S., Gentilello, L. (2007). A nationwide speed limit ≤ 65 miles per hour will save thousands of lives. The American Journal of Surgery, 193, 719-722. Sliogeris, J. (1992). 110 Kilometre Per Hour Speed Limit - Evaluation of Road Safety Effects. Report No: GR 92 – 8. VIC ROADS, Road and Environment Safety, Road Safety Division, Carlton. Australia. Solomon, D. (1964). Accidents on main rural highways related to speed, driver and vehicle. Bureau of Public Roads, U.S. Department of Commerce, Washington D.C. U.S. Spanish Economic Ministry. Website: http://serviciosweb.meh.es/apps/dgpe/default.aspx. Accessed November 2012. SWOV (2010). Towards credible speed limits. Fact Sheet of September. Leidschendam, The Netherlands. SWOV (2012). The relationship between speed and crashes. Fact Sheet of April, Leidschendam, The Netherlands. Wegman, F. & Aarts, L. (eds.) (2006). Advancing sustainable safety; National Road Safety Outlook for 2005-2020. SWOV, Leidschendam. The Netherlands. Wilmot, Ch. G., Khanal, M. (1999). Effect of speed limits on speed and safety: a review. Transport Reviews, 19, 315-329. Wilson, C., Willis, C., Hendrikz, J.K., Bellamy, N. (2006). Speed enforcement detection devices for preventing road traffic injuries. Cochrane Database of Systematic Reviews, 2, 1-49. Wong, S.C., Sze, N.N., Lo, H.K., Hung, W.T., Loo, B.P.Y. (2005). Would relaxing speed limits aggravate safety? A case study of Hong Kong. Accident Analysis and Prevention, 37, 377-388.