Parking slots in Europe: Results review and new model proposal. Phase 2: EPA s model improvement and design of a new survey



Similar documents
Scope of Parking in Europe

On the Degree of Openness of an Open Economy Carlos Alfredo Rodriguez, Universidad del CEMA Buenos Aires, Argentina

2. Linear regression with multiple regressors

Forecasting the US Dollar / Euro Exchange rate Using ARMA Models

Competition as an Effective Tool in Developing Social Marketing Programs: Driving Behavior Change through Online Activities

Determinants of Stock Market Performance in Pakistan

Air passenger departures forecast models A technical note

Report on impacts of raised thresholds defining SMEs

13 th Economic Trends Survey of the Architects Council of Europe

The Impact of Privatization in Insurance Industry on Insurance Efficiency in Iran

Linear Regression. Chapter 5. Prediction via Regression Line Number of new birds and Percent returning. Least Squares

ARE THE POINTS OF SINGLE CONTACT TRULY MAKING THINGS EASIER FOR EUROPEAN COMPANIES?

UK GDP is the best predictor of UK GDP, literally.

Applied Econometrics and International Development Vol (2012)

COMMUNICATION FROM THE COMMISSION

Determinants of demand for life insurance in European countries

An Analysis of the Effect of Income on Life Insurance. Justin Bryan Austin Proctor Kathryn Stoklosa

How To Understand Factoring

Broadband and i2010: The importance of dynamic competition to market growth

THE IMPACT OF DISTANCE IN RELIGION BETWEEN COUNTRIES ON FOREIGN DIRECT INVESTMENT FLOWS FROM THE EUROPEAN UNION (EU) DIRECTED TO TURKEY AND POLAND

IRG-Rail (13) 2. Independent Regulators Group Rail IRG Rail Annual Market Monitoring Report

REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL

The relationship between stock market parameters and interbank lending market: an empirical evidence

PUBLIC VS. PRIVATE HEALTH CARE IN CANADA. Norma Kozhaya, Ph.D Economist, Montreal economic Institute CPBI, Winnipeg June 15, 2007

The Tax Burden of Typical Workers in the EU Edition. James Rogers & Cécile Philippe May (Cover page) Data provided by

DOCTORAL (Ph.D) THESIS

Competition Strategies between Five Cell Phone Service Providers in Pakistan

IMPACT OF WORKING CAPITAL MANAGEMENT ON PROFITABILITY

IAB Adex Benchmark 2012 Daniel Knapp, IHS Electronics & Media

European Journal of Business and Management ISSN (Paper) ISSN (Online) Vol.5, No.30, 2013

THE EVOLUTION AND THE FUTURE ROLE OF THE BRANCH IN DISTRIBUTION OF THE BANKING PRODUCTS AND SERVICES

Statistical Data on Women Entrepreneurs in Europe

International ACH: Payment Gateway to Europe

How To Calculate Tax Burden In European Union

Do Currency Unions Affect Foreign Direct Investment? Evidence from US FDI Flows into the European Union

GDP per capita, consumption per capita and comparative price levels in Europe

FUSIONS Food waste data set for EU-28. New Estimates and Environmental Impact

Correlation of International Stock Markets Before and During the Subprime Crisis

RULES FOR THE REIMBURSEMENT OF TRAVEL AND SUBSISTENCE EXPENSES FOR EXCHANGE OF OFFICIALS

168/ November At risk of poverty or social exclusion 2 rate in the EU28, (% of total population)

IAB Europe AdEx Benchmark Daniel Knapp, IHS Eleni Marouli, IHS

PUBLIC & PRIVATE HEALTH CARE IN CANADA

Methodology For Illinois Electric Customers and Sales Forecasts:

EXECUTIVE SUMMARY. Measuring money laundering at continental level: The first steps towards a European ambition. January 2011 EUROPEAN COMMISSION

How to get your invoice paid on time?

Competitiveness of Travel Agencies in the European Tourism Market. Iris Mihajlović. University of Dubrovnik, Dubrovnik, Croatia

Current-Account Imbalances and Economic Growth During the Financial Crisis: an Empirical Analysis

EUF STATISTICS. 31 December 2013

Insurance corporations and pension funds in OECD countries

ERASMUS+ MASTER LOANS

European Hotel Distribution Study: The Rise of Online Intermediaries. Special Focus Switzerland

Electricity and natural gas price statistics 1

The Effect of Seasonality in the CPI on Indexed Bond Pricing and Inflation Expectations

Energy prices in the EU Household electricity prices in the EU rose by 2.9% in 2014 Gas prices up by 2.0% in the EU

187/ December EU28, euro area and United States GDP growth rates % change over the previous quarter

Expenditure on Health Care in the UK: A Review of the Issues

TPI: Traffic Psychology International on a common European curriculum for postgraduate education in traffic psychology

Time Series Analysis

99/ June EU28, euro area and United States GDP growth rates % change over the previous quarter

Immigration Reform, Economic Growth, and the Fiscal Challenge Douglas Holtz- Eakin l April 2013

INTERNATIONAL COMPARISONS OF PART-TIME WORK

BEST PRACTICES/ TRENDS/ TO-DOS

Labour Force Survey 2014 Almost 10 million part-time workers in the EU would have preferred to work more Two-thirds were women

Competition in Mobile Communications and the Allocation of Scarce Resources: The Case of UMTS. Jörn Kruse

The Tax Burden of Typical Workers in the EU Edition

Forecasting Using Eviews 2.0: An Overview

FUNCTIONAL EXPLORATORY DATA ANALYSIS OF UNEMPLOYMENT RATE FOR VARIOUS COUNTRIES

The adjustment of the Spanish Real Estate Sector. May 2011

Predicting The Outcome Of NASCAR Races: The Role Of Driver Experience Mary Allender, University of Portland

Expenditure and Outputs in the Irish Health System: A Cross Country Comparison

Introduction to Regression and Data Analysis

INNOBAROMETER THE INNOVATION TRENDS AT EU ENTERPRISES

Health Care Systems: Efficiency and Policy Settings

First estimate for 2014 Euro area international trade in goods surplus bn 24.2 bn surplus for EU28

ERASMUS+ MASTER LOANS

The role of the banking sector in enhancing extractive industries in Sudan

What can econometrics tell us about World Cup performance? May 2010

Size and Development of the Shadow Economy of 31 European and 5 other OECD Countries from 2003 to 2015: Different Developments

Health Care a Public or Private Good?

Guidance on Performance Attribution Presentation

Integrated Resource Plan

Flexicurity. U. Michael Bergman University of Copenhagen

SURVEY ON THE LONG-TERM PRESERVATION OF DIGITAL DOCUMENTS IN EUROPEAN LIBRARIES Monika Krimbacher Michael Neuhauser Martina Vogl

Fourth study of the Solvency II standard approach

Brochure More information from


The education system and lifelong learning in Finland. October 2015 Petri Haltia

The Architectural Profession in Europe

NERI Quarterly Economic Facts Summer Distribution of Income and Wealth

41 T Korea, Rep T Netherlands T Japan E Bulgaria T Argentina T Czech Republic T Greece 50.

Survey on the access to finance of enterprises (SAFE) Analytical Report 2014

- Assessment of the application by Member States of European Union VAT provisions with particular relevance to the Mini One Stop Shop (MOSS) -

Financial market integration and economic growth: Quantifying the effects, Brussels 19/02/2003

Hong Kong s Health Spending 1989 to 2033

The Nordic Tourism Investment Index 2012

How many students study abroad and where do they go?

Polish insurance market: growth and potential

ERGP (12) 33 ERGP report on data collection ERGP REPORT WITH DATA ON INDICATORS ON THE POSTAL MARKET

Transcription:

Parking slots in Europe: Results review and new model proposal Phase 2: EPA s model improvement and design of a new survey Research team: Jordi Suriñach (director) Manuela Alcañiz José Ramón García Montserrat Guillén Institut de Recerca en Economia Aplicada (UB-IREA) Dept. d Econometria, Estadística i Economia Espanyola Universitat de Barcelona November, 2011

TABLE OF CONTENTS: 1. Abstract... 3 2. Introduction and Phase 2 goals... 4 3. Methodology used to improve EPA s proposed model... 5 3.1. Generating the data file... 7 3.2. Estimate by raising the total number of slots... 8 3.3. Estimate by regression of the total number of slots... 10 4. Proposal of a new questionnaire... 16 5. Conclusions... 17 6. References... 18 Appendix 1: Preliminary models... 20 Appendix 2: Model for the total parking slots estimate... 24 Appendix 3: New questionnaire... 24 2

1.- Abstract The European Parking Association (EPA) wishes to have quantitative information about the number of parking slots in Europe. Nevertheless, it does not exist an official unified record about this type of information. That is why the EPA has set the statistical obtaining of the total number of regulated parking slots (on-street and off-street) for all cities above 20,000 inhabitants from 23 European countries 1 as one of its objectives. This will remark the importance of this sector, proving its strong effect on other areas, such as commercial activities or the public sector, among others. In order to obtain a first approximation to the number of parking spaces, the EPA conducted its own estimates based on a survey. Some national parking associations cooperated by helping to obtain data from the municipalities of their countries. The completion of this work was assessed as very positive because it was pioneering research, which serves as reference for further work in the field. Based on the results obtained by the EPA, the research team at the University of Barcelona revised the model used by the association to carry out the estimate. In addition, it discussed the method of calculation used, as well as the reliability of the results obtained (Suriñach et al., 2011). It was concluded that the model proposed by the EPA has been an important step towards increasing the industry knowledge. It allows a first overview, making it very useful to explore the situation and drive, at later stages, improvements in both data collection and in the estimates. As it was seen, the main limitation of the model proposed by the EPA is caused by the insufficient number of municipalities that responded to the survey, providing the information about the number of parking spaces available to them. This fact alone already indicates that the first approach towards improving the results must come from the increase in the number and the quality of the base information. However, the methodology can be improved as well. For example, one can relax the assumption that the population and the number of parking slots grow at the same rate as the GDP. 1 The countries for which information wants to be known are: Germany, Austria, Belgium, Croatia, Cyprus, Denmark, Slovakia, Spain, Finland, Greece, The Netherlands, Hungary, Ireland, Italy, Luxembourg, Norway, Poland, United Kingdom, Serbia, Sweden and Switzerland. 3

2.- Introduction and Phase 2 goals This second phase intends to study and analyze the viability of alternative methodologies for estimating the number of parking spaces, based on currently available statistical information (primarily on the previous survey conducted by the EPA). The aim is to improve the reliability of the estimate, although the lack of representativeness of the sample and the absence of more complete information on the municipalities will continue posing a limiting circumstance to take into account. To obtain a statistically reliable estimate, it will be necessary to obtain new data. This data may be obtained from a second survey aimed at European municipalities. This survey will not only collect information on the number of parking spaces, but also on some characteristics of the municipalities that may lead to explain why they could have a deviation on the offer of parking slots. These variables were partly explored through the literature review conducted in the first phase (Anderson and de Palma, 2007, Davis et al., 2010). They have been chosen and discussed in more detail to draw up a new questionnaire, which is shown in Appendix 3. Also, the National Parking Association (NPA) method has been reviewed. This association carries out a study for the United States similar to the one desired here (NPA, 2011) 2. Thus, the second survey will provide more complete data, both for the inclusion of new variables of use (suggested by the literature), as for the participation of a more extensive sample of municipalities, which should be encouraged. Once that information is accessible in the third phase, the application of specific statistical and econometric techniques will allow the design and implementation of a new model 3. This model will provide more precise values for municipalities or for entire countries for which no parking data is available, leading necessarily to an improvement of the final estimate quality. Also, it will be possible to know the reliability of the result, endowing it with lower and upper confidence limits and giving a high probability of success. 2 The review of that study has shown that the objectives of the NPA are not equivalent to those of this study. Furthermore, neither the city nor the type of methodology used are comparable to those presented here, which is why the NPA s methodology could not be transferred or adapted. 3 The proposal and implementation of a new model, which will use additional data from a second survey, is contemplated for Phase 3 of the project. 4

Specifically, the double objective of the second phase is: 1. Suggest changes or improvements to the model developed by the EPA. Using existing data, which comes from the survey designed by the EPA, it is intended to obtain an improved estimate of the parameter of interest: the total number of parking spaces in Europe. 2. Designing a new questionnaire to obtain useful additional information to better estimate the number of parking spaces. Once Phase 2 is finished and the proposed survey has been conducted, the treatment of the data obtained will be carried out in Phase 3, which is the last one. A new statisical model will be suggested and implemented in order to estimate the total number of parking spaces by identifying significant variables, validating the resulting model and analyzing its reliability. 3.- Methodology used to improve EPA s proposed model First of all, remember that in the EPA s questionnaire municipalities are asked to count two types of parking slots 4 : on-street (regulated slots at street level in which some control takes places and where payment is not always necessary) and off-street (parking above or below ground level, parking in outdoor enclosures with controlled access and parking in hospitals, malls, universities and park & rides). The steps to be taken at this stage, in order to obtain a better estimate of the total parking spaces in Europe, and starting from the data available from the EPA survey, are: 1) Generation of a single file with all the data from the first survey, distinguishing between real data and estimates. The file will contain all municipalities with over 20,000 inhabitants of the countries studied, accounting for its population and the data of on-street and off-street parking, specifying whether they are real or estimated. 2) Using information from the survey, an additional homogenization of the data will be carried out (based on the growth of the population) to that provided by the EPA (based on per capita GDP growth), taking 2009 as the reference year. 4 Further detail on the type of slots that come into the count can be found in the document EPA s Data Collection: representing the parking sector (A. Roig). 5

3) A first methodology used, which is done at a basic level, starts from the assumption (restrictive) that municipalities for which data is available are a representative sample of all municipalities with over 20,000 inhabitants. In this case, an initial estimate is done by raising the number of on-street, off-street and total parking slots. This estimate will be based on three different situations: without a homogenization of the data, homogenizing the data from GDP growth and, finally, homogenizing data from population growth countries. 4) Also, with the same previous methodology, a first approximation to the hypothetical number of parking spaces that all municipalities as a whole could have will be obtained. These municipalities will also include those with less than 20,000 inhabitants. The result should be used with extreme caution, as for its obtention an additional assumption is made; it assumes the total number of parking spaces per capita is the same regardless of the number of inhabitants of the municipality. In this sense, it is reasonable to think that very small municipalities are not as regulated as other places, so the estimate may be considered biased upwards. 5) A second approach is based on the use of multiple regression models, using the number of on-street, off-street and total parking slots as endogenous variables. There will not be a segmentation by size of municipality when carrying out the estimates. Nevertheless, the data from the number of inhabitants of the municipalities will be incorporated as an explanatory variable. These regression models will be raised following several alternatives: separately using the data without an update, homogenized data from GDP growth and updated data regarding the population growth. If required, a simultaneous equations model will be used. 6) In addition to the methodology based on the regression model, the possibility of introducing other explanatory variables in the model is explored later on. Eurostat databases are used to take into account additional GDP per capita variables. 7) Next, the possibility of atypical behavior in some municipalities is analyzed. These could substantially change the results of the estimate. Also, an analysis into the regression model is taken to consider if it is worth to take into account different municipality behaviors by population ranges. 6

8) Finally, from the regression models considered, a prediction into the total number of slots for all municipalities with over 20,000 inhabitants will be obtained. 3.1. Generating the data file A single file has been generated from the spreadsheet with the actual data or th estimates for each of the 23 countries included in the study. This file includes 3,805 municipalities with more than 20,000 inhabitants, each with its own population and its own on-street and off-street number of parking spaces, given that the estimate is available. It has been distinguished whether data is real or an estimate for the parking spaces 5. At this stage of the study, the estimated data will be excluded, since a different approach to that used by the EPA for unknown values is used. This will be made through a linear regression model. Subsequently, such model will be improved through the contrast of its underlying assumptions (linearity, homoscedasticity, etc.) and the presence of outliers (items with values far from the average), which will be treated properly so as not to distort the set of results. Thus, there will be a more accurate estimate of the number of parking slots. The initial database does not refer to all variables within a same year. There are municipalities for which data is referred to a year that is not 2009 (base year considered in the study). On the other hand, for some other municipalities there is no existing match between the year parking slots data slots and the year of population numbers. A first task has been to estimate these variables so that all municipalities have the parking and the population data changed into 2009 values. Then, two possible estimates of the total parking spaces in Europe is proposed. One is based on raising the available data for all Europe. A second one results from applying a simple linear regression model to predict the unknown data based on the population of the municipalities and the gross domestic product of the country. 5 EPA s initial estimate 7

3.2. Estimate by raising the total number of slots The following table summarizes the baseline data. Two approaches are used to update the population and the parking spaces data are shown. The first approach is the one already used by the EPA, which is to update the population and the parking slots from the growth of GDP per capita of each country. The second approach is the one that uses the growth of the population of each country, which has been obtained from Eurostat. This growth is used to update the parking slots and the population at the required municipalities. The second approach results from picking up two ideas. On the one hand, NPA s idea (National Parking Association, USA) of correlating the population density to the number of parking spaces. On the other hand, implying that the weight of the municipality within the country has not changed over the period that goes from the collection of data to the reference year (2009). This approach also assumes that the population of a municipality increases or decreases only if the population of its country has done so, regardless of whether GPD per capita has increased or not. Note that this second approach is not intended as a substitute but as a complement to that counducted by the EPA from GDP growth, as both have their own strengths. In fact, the study will use the 3 approaches (also including the one in which the database will not be homogenized to a base year) in order to assess which one of these approaches fits best. 8

Real values of parking slots Updated from No update GDP Updated from population Off-street slots 2.128.959 2.137.447 2.158.493 On-street slots 2.009.434 2.008.410 2.044.529 Total 4.138.393 4.145.857 4.203.022 Population from municipalities with slots information 69.109.553 68.047.712 69.593.577 Total population from municipalities 258.226.096 260.315.198 260.886.189 Population from the 23 countries. Year 2009 474.991.399 Now will come the estimate of the number of parking spaces from the initial data, shown above. This is the so-called raising of the total number of slots. Assuming that municipalities for which there is information about the number of slots are a representative sample of all municipalities in Europe, a first way to estimate the total number of parking spaces would consist on increasing the number of parking spaces proportionally to the population. The following table shows the results: Estimate (increase) of the number of off-street and on-street slots No update Updated from GDP Updated from population Off-street slots 7.954.801 8.176.762 8.091.566 On-street slots 7.508.199 7.683.133 7.664.348 Total slots for municipalities with population over 20,000 inhabitants 15.463.001 15.859.895 15.755.914 Total slots for all municipalities 28.443.261 28.939.201 28.686.545 9

Similar results are observed, regardless of the criteria used for the data update (GDP growth or population growth) Also noteworthy is the fact that the estimate for municipalities with over 20,000 inhabitants does not differ substantially from that obtained by the EPA, which was 15,041,608 slots. The crucial feature of the observed differences from the different updates is the fact that the estimate considered by the EPA is leading to a lower number of parking spaces. This result may be closely linked to the separation of the municipalities in 7 groups according to intervals of population. In this sense, one aspect to analyze is the relevance of this consideration by introducing the population of the municipalities variable into the regression model, as well as introducing dummy variables for municipalities with a diferenciated behavior. In the last table provided, for the first time, it is also shown an estimate of the total number of municipalities in Europe, whatever their number of inhabitants. This estimate is only an approximation and not error-free, since it is implicitly assuming that the number of slots per capita is the same in municipalities with a population over 20,000 than in those of smaller dimensions. This assumption implies an upward bias of the result, as it seems reasonable to think that small municipalities do not have as many regulated slots as bigger municipalities. 3.3. Estimate by regression of the total number of slots In order to make a prediction through a linear regression model, various strategies have been developed. A. First estimates First, some simple linear regressions have been carried out to separetely estimate the number of on-street, off-street and total slots based on the population variable. The estimate was made for the non-updated data (of1, on1, tot1, pob1), the updated data by GDP (of2, on2, tot2, pob2) and the updated data by population (of3, on3, tot3, pob3). The endogenous variables used were offstreet, on-street and total slots, while the population variable has been used as explanatory. 10

These simple linear regressions express the idea raised by the EPA to estimate the number of parking slots from the population and that the EPA conducted a different methodology from the one discussed here 6. In our case, the aim is to find a stable relationship between the parking and the population of all municipalities for which this information is available. This relationship is then intended to be inferred to the rest of municipalities for which we know (or can know) its population, without knowing its number of parking slots. This inference will be conducted by using the estimated parameters from the initial regression model into the rest of municipalities, for which an estimate of the total number of parking slots is wanted. Additionally, it should be noted that data from municipalities for which information exists (based on the survey conducted by the EPA) has been used to estimate the parameters. This has been done by jointly considering the total number of parking spaces and the off-street and on-street slots. It should therefore be noted that the number of observations included in the regression model depends on whether the endogenous variable is the offstreet, the on-street or the total slots. Appendix 1 shows the results obtained from the regressions. While the models may have limitations and econometric problems to be studied further, the following initial conclusions can be made: - The models are highly significant in all cases. Therefore, it can be argued that the population of the municipalities has a correlation with the number of slots, being these on-street, off-street or an overall measure. Thus, the idea raised by the EPA to estimate the number of parking spaces from the population has also a statistical/econometric proof. - The adjustment of the models is acceptable, ranging between 45% (mainly in the regresions analyzing off-street slots) and 60% (for models analyzing both the total and the on-street slots) - For every 100 additional inhabitants, the expected number of parking spaces is increased by about 3 at off-street and 3 at on-street. Thus, the overall increase is about 6 slots per 100 inhabitants. Once the models have been estimated, a prediction of the number of slots for the rest of the municipalities has been carried out. A simple sum allows to have an improved estimate of the total slots: 6 The review of the model used by the EPA can be checked at the research by Suriñach et al., 2011. 11

Predictions from simple regression models No update Updated from GDP Updated from population Off-street slots 10.378.710 10.650.549 10.678.718 On-street slots 7.540.751 7.708.940 7.844.514 Total as sum 17.919.461 18.359.489 18.523.232 Total (joint analysis) 18.052.614 18.471.328 18.726.536 As it can be seen, the provisional estimate of the number of parking spaces in Europe is just over 18.5 million. From those, around 10.7 million are offstreet, and the rest (7.8 million) are on-street. Also, it is found that predictions calculated from the regressions models considered are significantly higher than those obtained by raising the number of slots proportionally to the population. This method gave a lower estimate at 16 million. In contrast, the regression models give over 18.5 million. The predictions carried out do not differ substantially with different updates or homogenizations of the parking slots, making it advisable to use the update by the population. Nevertheless, regressions carried out until now (shown in Appendix 1) cannot be considered as definitive, since they only gather the idea that the population is a variable that can only partly explain the number of slots. Thus, it is necessary an improvement process of these. B. Additional considerations This improvement process of the equations has lead to consider new modelling strategies, which has caused a new definition of the regression model. Next, the tests that have been carried out are shown: I. In the first place, there has been an attempt to introduce other explanatory variables into the model in order to capture the effect of the country. These variables are GDP, registration of vehicles, number of vehicles or the overall length of roads. Analysis with these variables have provided results that show a weak relationship with the number of parking spaces, being the only exception GDP, which is a relevant variable when explaining the number of slots. 12

II. Also, dummy variables related to each country have been introduced as an alternative to account for the country-effect. The results obtained suggest that different country charasteristics do not sufficiently explain the different number of parking slots of the municipalities. It is important to remark that, through the analysis of the database, it has been detected that some municipalities show data which is far away from the behavior of the rest of municipalities. Specifically, the municipalities not matching the model are: Barcelona, Berlin, Madrid, Munich and Rome 7. The reasons lying behind their deviation from other municipalities are: o Berlin has a low number of parking slots in relation to its population. It particularly has very few on-street slots. o Barcelona has many off-street slots. In fact, it is ranked as the second municipality with this type of parking spaces, right below Munich. Also, it relatively has a small number of on-street slots. o Munich has a high number of parking slots in relation to its population. It is the municipality with the highest number of offstreet slots and the second one in on-street slots. o Madrid has many on-street slots. It is the municipality with a higher number of this type of slots. o Rome has very few parking spaces in relation to its population, and especially regarding its off-street slots. 7 In econometric terms, we could say they are outliers. That is, observations for which the data generation process is different to the rest of observations (municipalities) analyzed in the model 13

Population off-street slots (1) On-street slots (2) Total slots (1+2) Barcelona 1.621.500 144.000 40.000 184.000 Berlin 3.431.675 51.209 11.892 63.101 Madrid 3.255.900 115.600 165.000 280.600 Munich 1.326.807 194.555 103.370 297.925 Rome 2.743.796 20.815 83.734 104.549 The analysis carried out for these municipalities advised the introduction of dummy variables into the model which would account for their specific characteristics. This would mean an improvement in the predictive ability. Additionally, and checking that municipalities with a special behavior in the model correspond to large municipalities, the strategy consisting in introducing dummy variables accounting for the influence of big cities has been carried out. However, the results have not proven to be consistent with this argument. This shows that only dummy variables are suitable to account for the special behavior of municipalities mentioned above. To conclude this section about atypical observations, it can be concluded that: (a) there are reasonable doubts about some of the numbers of parking slots of these municipalities. Therefore, having the third phase in mind, a thorough analysis of outliers and in some cases a verification of the goodness of the figures obtained in the questionnaires should be made; (b) dummy variables associated to these atypical municipalities must be included. This is made in order to preserve estimated coefficients in the regression models, which should be as close as possible to the true population values. The objective is that these coefficients allow the use of better estimates in the predictive phase (for the rest of municipalities of the model); (c) for the third phase it is advised to try to group or find similar atypical observations through multivariate analysis which allow to define dummy variables associated to this specific type of cities. III. The estimate of the regression model with the improvements previously explained (additional explanatory variables and the introduction of outliers) has lead to important advances on the explanation of the number of parking spaces. However, the econometric validation carried out has shown that there is an existing 14

quadratic relationship between population and the total number of slots. This fact has been included in the regression model by the inclusion of the squared population variable. IV. Finally, it has been studied whether the specification of the model had to be: a. Single-equation model associated with each of the two types of slots available: off-street and on-street. Thus, the specification of two equations whose aggregate provides the total number of slots. b. Single-equation model associated to the total number of existing slots: off-street and on-street. Thus, the specification of a single equation. c. Multiequational model associated with each one of the two types of slots available: off-street and on-street. Thus, the specification of two equations, jointly estimated through a SURE model (Seemed Unrelated Regression Equations) of two equations explaining on-street and off-street slots separately, and imposing that the sum of both types of slots to be the total. In principle, the joint estimate could be improve the single-equation estimate. However, the econometric analysis carried out has shown that the estimate through a SURE model does not improve the efficiency of the single-equation estimate. That is why it has been rejected. C. Final model After much consideration and testing, and implementing the usual strategy for selecting the best model, the estimated regression model, which has been validated and finally selected is: TOTi = β1 + β2pobi + β3munichi + β4barnai + β5madridi + + β + β + β + β 2 6BERLINi 7ROMAi 8GDPi 9POB + u i i In Appendix 2 the estimate results of this regression model are shown. The model is highly significant. Thus, it can be stated that population of municipalities, squared population and gross domestic product (also dummy variables introduced for the analysis of outliers) allow the explanation of the number of slots. The adjustment of the model is good, explaining 92% of the variability in the number of parking spaces. 15

Once the model has been estimated and validated, the prediction of the total number of slots for the rest of the municipalities has been carried out. A simple sum of the prediction for each one of them allows a better prediction for the total number of slots to be obtained. This prediction indicates that the total number of parking spaces for all municipalities of over 20.000 inhabitants for all the 23 countries considered can be estimated to be 19.593.660. We do not want to finish this section without remarking again the highly temporary nature of the predicted data, given the existing uncertainties about the base data used. As a way of improving the predictions, the detection of possible measuring errors in the survey must be considered. Also, obtaining new data from the new survey which improves the results previously obtained. 4.- Proposal of a new questionnaire As it has been already mentioned, the information that has lead to the estimates done by the EPA and later on checked by the Universitat de Barcelona was obtained through a survey directed to european municipalities with over 20.000 inhabitants. At the time of bringing up this investigation, the opportunity of carrying out a second survey was seen. It tries to improve the data gathering and increase the amount of information obtained through the incorporation of new questions. In Appendix 3 the new questionnaire suggested by the Universitat de Barcelona is shown. Such questionnaire consists of an initial page in which instructions are given, urging municipal officals to fill it and defining in a detailed fashion the on-street and off-street slots that are considered in the study. The design of the questionnaire has been carried out giving priority to its clearness, with relatively few questions (and easy to answer them if possible) so that there was an incentive to answer them. The structure of the questionnaire consists of 3 sections: mobility, parking slots and fares. In the section dedicated to mobility the charasterics of the municipalities which can be related to parking demand are gathered. These charasteristics 16

are, for example, the existence of metro or tramway, whether the municipality is coastal or not and whether it is densely populated. It also tries to clarify if the city attracts trips due to its position as center of administration, commerce, tourism, etc. In the section dedicated to parking slots a careful design to incentivate the answering of the number of on-street and off-street slots has been carried out. This data is of high interest regarding the future estimate of the total number of slots which will be carried out in phase 3 of this project. Questions about the percentage of charged parking spaces, average occupancy of these spaces and the percentage of off-street slots with a rent contract associated have been asked. Finally, the section dedicated to fares includes questions about the average fares of on-street and off-street slots, discerning among drivers with a parking pass, foreigners or residents. In Appendix 3 the quetionnaire is included. 5.- Conclusions Results shown in this report are based on data gathered through the first EPA questionnaire, a fact from which many limitations already mentioned are derived. Anyway, the estimates presented from the multiple linear regression model have been validated econometrically. With these precautions, one can state that the total number of slots in Europe is over 19,5 millions (the precise prediction is 19.593.660). From the literature and from the results that the statistical study of this phase has given, a second questionnaire for municipalities over 20.000 inhabitants has been suggested. This would be through national parking associations which are part of the EPA. This second phase will allow useful new data for the third phase to be obtained. In this phase, once the second questionnaire is carried out, it will be possible to include new variables to the model, improving the reliability and the precision of the results. 17

6.- References Anderson, S.P.; de Palma, A. (2007) Parking in the city, Papers in Regional Science, 86, 4, 621-632. Cuddy, M.R. (2007) A practical method for developing context-sensitive parking standards. PhD Dissertation. Rutgers, The State University of New Jersey. Davis, A.Y.; Pijanowski, B.C.; Robinson, K.D.; Kidwell, P.B. (2010) Estimating parking lot footprints in the Upper Great Lakes Region of the USA, Landscape and Urban Planning, 96, 68-77. Ferguson, E. (2004) Zoning for parking as policy process: a historical review, Transport Reviews, 24, 2, 177-194. Kish, L. (1995) Survey sampling. Wiley Classics Library, New York. Lohr, S.L. (1999) Sampling: design and analysis. Duxbury Press, Pacific Grove. Marshall, W.E.; Garrick, N.W. (2006) Parking at mixed-use centers in small cities, Journal of the Transportation Research Board, 1977, 164-171. McDonnell, S.; Madar, J.; Been, V. (2011) Minimum parking requirements and housing affordability in New York City, Housing Policy Debate, 21, 1, 45-68. Mukhija, V.; Shoup, D. (2006) Quantity versus quality in off-street parking requirements, Journal of the American Planning Association, 72, 3, 296-308. NPA (2011) Parking in perspective: The size and scope of parkint in America. National Parking Association, United States of America. Shoup, D. (2005) The high cost of free parking, Chicago: Planners Press. Suriñach, J.; Alcañiz, M.; García, J.R.; Guillén, M. (2011) Parking slots in Europe. Results checking and new model proposal. Phase 1: EPA s model cheking. Universitat de Barcelona. 18

Verhoef, E.; Nijkamp, P.; Rietveld, P. (1995) The economics of regulatory parking policies: the (im)possibilities of parking policies in traffic regulation, Transportation Research A, 29, 2, 141-156. 19

Appendix 1: Preliminary models The following tables show the preliminary estimates of the simple linear regression models for the non-updated data (of1, on1, tot1, pob1), for the updated data by GDP (of2, on2, tot2, pob2) and for updated data by population (of3, on3, tot3, pob3). In all cases, the explanatory variable is the population. Dependent Variable: ON1 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 481 Excluded observations: 2824 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C -147.0209 390.1637-0.376819 0.7065 POB1 0.031369 0.001173 26.74146 0.0000 R-squared 0.598863 Mean dependent var 4177.618 Adjusted R-squared 0.598025 S.D. dependent var 12282.49 S.E. of regression 7787.279 Akaike info criterion 20.76252 Sum squared resid 2.90E+10 Schwarz criterion 20.77988 Log likelihood -4991.386 F-statistic 715.1054 Durbin-Watson stat 1.376261 Prob(F-statistic) 0.000000 Dependent Variable: ON2 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 481 Excluded observations: 2824 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C -157.7057 381.6474-0.413224 0.6796 POB2 0.031879 0.001174 27.15092 0.0000 R-squared 0.606141 Mean dependent var 4170.290 Adjusted R-squared 0.605319 S.D. dependent var 12105.49 S.E. of regression 7605.110 Akaike info criterion 20.71518 Sum squared resid 2.77E+10 Schwarz criterion 20.73254 Log likelihood -4980.000 F-statistic 737.1725 Durbin-Watson stat 1.403479 Prob(F-statistic) 0.000000 20

Dependent Variable: ON3 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 480 Excluded observations: 2825 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C -70.13015 387.9612-0.180766 0.8566 POB3 0.031092 0.001161 26.79141 0.0000 R-squared 0.600261 Mean dependent var 4253.195 Adjusted R-squared 0.599425 S.D. dependent var 12212.86 S.E. of regression 7729.642 Akaike info criterion 20.74767 Sum squared resid 2.86E+10 Schwarz criterion 20.76506 Log likelihood -4977.441 F-statistic 717.7797 Durbin-Watson stat 1.387069 Prob(F-statistic) 0.000000 Dependent Variable: OF1 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 358 Excluded observations: 2947 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C 612.9411 690.2139 0.888045 0.3751 POB1 0.031161 0.001805 17.26791 0.0000 R-squared 0.455808 Mean dependent var 5946.813 Adjusted R-squared 0.454279 S.D. dependent var 15809.16 S.E. of regression 11678.69 Akaike info criterion 21.57449 Sum squared resid 4.86E+10 Schwarz criterion 21.59617 Log likelihood -3859.834 F-statistic 298.1807 Durbin-Watson stat 1.825532 Prob(F-statistic) 0.000000 Dependent Variable: OF2 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 359 Excluded observations: 2946 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C 644.5343 693.2080 0.929785 0.3531 POB2 0.031656 0.001858 17.03338 0.0000 R-squared 0.448338 Mean dependent var 5953.890 Adjusted R-squared 0.446793 S.D. dependent var 15773.11 S.E. of regression 11731.70 Akaike info criterion 21.58353 Sum squared resid 4.91E+10 Schwarz criterion 21.60517 Log likelihood -3872.244 F-statistic 290.1359 Durbin-Watson stat 1.793935 Prob(F-statistic) 0.000000 21

Dependent Variable: OF3 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 358 Excluded observations: 2947 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C 677.0512 689.8527 0.981443 0.3270 POB3 0.031058 0.001796 17.28983 0.0000 R-squared 0.456437 Mean dependent var 6029.309 Adjusted R-squared 0.454910 S.D. dependent var 15799.33 S.E. of regression 11664.67 Akaike info criterion 21.57209 Sum squared resid 4.84E+10 Schwarz criterion 21.59377 Log likelihood -3859.404 F-statistic 298.9382 Durbin-Watson stat 1.818783 Prob(F-statistic) 0.000000 Dependent Variable: TOT1 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 341 Excluded observations: 2964 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C 506.8989 1086.329 0.466616 0.6411 POB1 0.062441 0.002786 22.41524 0.0000 R-squared 0.597121 Mean dependent var 11260.74 Adjusted R-squared 0.595932 S.D. dependent var 28313.88 S.E. of regression 17998.10 Akaike info criterion 22.43977 Sum squared resid 1.10E+11 Schwarz criterion 22.46224 Log likelihood -3823.980 F-statistic 502.4429 Durbin-Watson stat 1.441183 Prob(F-statistic) 0.000000 Dependent Variable: TOT2 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 342 Excluded observations: 2963 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C 522.5303 1083.047 0.482463 0.6298 POB2 0.063452 0.002848 22.28219 0.0000 R-squared 0.593543 Mean dependent var 11240.09 Adjusted R-squared 0.592347 S.D. dependent var 28106.70 S.E. of regression 17945.48 Akaike info criterion 22.43390 Sum squared resid 1.09E+11 Schwarz criterion 22.45632 Log likelihood -3834.196 F-statistic 496.4961 Durbin-Watson stat 1.410375 Prob(F-statistic) 0.000000 22

Dependent Variable: TOT3 Method: Least Squares Sample(adjusted): 430 3734 Included observations: 341 Excluded observations: 2964 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C 667.9406 1083.809 0.616290 0.5381 POB3 0.062039 0.002768 22.41453 0.0000 R-squared 0.597105 Mean dependent var 11428.23 Adjusted R-squared 0.595917 S.D. dependent var 28227.37 S.E. of regression 17943.44 Akaike info criterion 22.43368 Sum squared resid 1.09E+11 Schwarz criterion 22.45616 Log likelihood -3822.943 F-statistic 502.4112 Durbin-Watson stat 1.437261 Prob(F-statistic) 0.000000 23

Appendix 2: Model for the total parking slots estimate Next table shows the estimate of the multiple linear regression model, which uses the number of parking slots as an endogenous variable and the population of municipalities, the squared population, the GDP of countries and many dummy variables as an explanatory variable, in order to treat the outliers of some municipalities. This model has given an estimate of 19.593.660 parking spaces for Europe. Dependent Variable: TOT Method: Least Squares Sample(adjusted): 430 3734 Included observations: 341 Excluded observations: 2964 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C -254.1313 1098.658-0.231311 0.8172 POB3 0.045573 0.005353 8.513252 0.0000 MUNICH 199069.7 8826.201 22.55441 0.0000 BARNA 53126.24 10048.62 5.286918 0.0000 MADRID -92441.21 35780.12-2.583591 0.0102 BERLIN -334962.8 39046.21-8.578625 0.0000 ROMA -178457.6 23836.06-7.486874 0.0000 GDP 0.001263 0.000659 1.916710 0.0561 POB3*POB3 2.04E-08 4.55E-09 4.490989 0.0000 R-squared 0.920747 Mean dependent var 11428.23 Adjusted R-squared 0.918837 S.D. dependent var 28227.37 S.E. of regression 8041.708 Akaike info criterion 20.84871 Sum squared resid 2.15E+10 Schwarz criterion 20.94984 Log likelihood -3545.705 F-statistic 482.1405 Durbin-Watson stat 2.795397 Prob(F-statistic) 0.000000 Appendix 3: New questionnaire Next, the new questionnaire with which the new data gathering will be carried out is shown. 24

The scope of parking in Europe Instructions to fill in the questionnaire The aim of this study of the European Parking Association is to assess the magnitude and scope of the parking industry in Europe. To do this, please help answer the following brief questionnaire. The greater the precision of the answers, the more reliable the results obtained will be. If you have no data, but you can give an approximation, please do it. Otherwise, leave the question blank. If there are no parking spaces of a given type, indicate it with a 0. Do not take into account the size of the parking spaces. Fill in the number of parking spaces using your own standard. OFF-STREET parking spaces Parking of public use with access control (barrier, guard, etc.). No matter whether the manager is public or private. No matter whether payment is required or not. Excluded: Private use parking (housing, offices, etc.). Off-street type of spaces: In a structure: Multi-story or underground parking. On the surface: As long as there is an access control. Park&Ride: Dissuasive parking, normally next to train stations. In sport, cultural or leisure facilities: stadiums, museums, theaters, cinemas, etc. In shopping malls or markets. Others (in hospitals, universities, airports, etc.) ON-STREET parking spaces Regulated parking spaces on the road. No matter whether payment is required or not. Excluded: Spaces on private land, non-regulated spaces. On-street type of spaces: Regulated spaces for general public use. Resident-only spaces Loading and unloading spaces. Motorbike spaces. Other regulated spaces: Handicapped, embassies, police, etc. Important note: Non-regulated spaces are taken into account in another separate question. Please, do not add them to the rest of on-street parking spaces. Non-regulated spaces are considered to be those without signaling, where no special surveillance is carried out.

The scope of parking in Europe The aim of this study from the EPA is to conduct a solid description of the parking industry in Europe, allowing the assessment of its true magnitude and scope. Please, read the indications on the attached sheet before filling in the questionnaire. If you don t know an answer but you can give an approximation, please do it. Name of the municipality: Country: MOBILITY 1. First, we raise some questions regarding mobility in your municipality. Please, answer yes or no, checking the box with a cross. Yes No The municipality has metro The municipality has tram On-street parking has undergone an extensive regulating plan There are usually traffic congestion problems at peak hours It is a coastal municipality It is a municipality densely populated 2. The city attracts trips because: Yes No It is the administrative / sanitary / universitary center of the area It is a business / work / industrial center It has lots of commercial activity It is a cultural or touristic municipality Others: (specify)...... PARKING SPACES 3. How many off-street and on-street parking spaces are there in your municipality? Give the total number of parking spaces of each kind. If you wish, you can use the itemization shown in order to approximate the total. OFF-STREET total spaces Please, fill in this In a structure On the surface Park&Ride (dissuasive) In sport, cultural or leisure facilities In shopping malls or markets Others (Hospitals, Universities, Airports, etc.) data even if it is approximated ON-STREET total regulated spaces Please, fill in this data even if it is approximated Regulated for general public use Exclusive for residents Loading and unloading Motorbike spaces Other regulated spaces (handicapped, police, etc.) 26

4. Approximately, which percentage of the total spaces indicated before are charged parking spaces? Spaces off-street on-street No charged spaces Less than 20% 20%-50% 50%-80% More than 80% 5. If you had to approximate the average occupation rate of the charged parking spaces, which one would it be? Spaces off-street on-street Less than 20% 20%-50% 50%-80% More than 80% 6. Approximately, which percentage of the off-street spaces requires a pass? Pass Less 20% 20%-50% 50%-80% More than 80% 7. How many non-regulated on-street spaces are there in your municipality? Please, try to give an approximate value. NON-REGULATED on-street spaces RATES 8. Finally we will ask you some questions about Rates. If you do not have the exact figure or there are different rates, try to approximate the average rate. If possible, we would appreciate you to show the figures in Euros. Otherwise, use the currency of your country. Please, include the VAT. OFF-STREET parking Average rate per hour Average monthly rate (pass) Euros ON-STREET parking Average rate per hour (foreign) Average monthly rate (resident) Euros Thank you very much for your collaboration. 27