ECONOMETRIC ANALYSIS OF PUBLIC PARKING PRICE ELASTICITY IN EUGENE, OREGON
by Moshe Farber & Erin Weld University of Oregon June 13th, 2013 ABSTRACT
This paper explores the demand sensitivity to price changes of on campus municipally operated parking infrastructure in Eugene, Oregon ex ante and ex post of a 13% campus-‐wide price increase in 2012. We find that elasticity of demand is relatively inelastic at -‐0.3, and that a further 17% increase of price will net the city an additional $137,230.40 per year across all campus zones. This paper additionally looks at the effects of distance, time, and various other demand shifters on total revenue.
Presented to the Department of Economics, University of Oregon, in partial fulfillment of the requirements of honors in Economics
Farber & Weld
[This Page is Intentionally Left Blank]
1
Farber & Weld
2
Table of Contents:
I. Introduction..................................................................................................................................4 II. Literature Review.....................................................................................................................5 III. Methodology...........................................................................................................................10 A. Research Questions..................................................................................................10 B. Empirical Model.........................................................................................................11 C. Variable Description.................................................................................................12 D. Data Description........................................................................................................13 E. Model Description.....................................................................................................16 IV. Econometric Results............................................................................................................18 A. Price.................................................................................................................................18 B. Distance..........................................................................................................................19 C. Distance Interactions...............................................................................................21 D. Time.................................................................................................................................23 E. Student Enrollment...................................................................................................25 F. Further Analysis.........................................................................................................26 G. Conditional Forecast................................................................................................27 V. Conclusion.................................................................................................................................28 VI. Works Cited.............................................................................................................................30 Appendix A.....................................................................................................................................32 Appendix B.....................................................................................................................................35
Farber & Weld
[This Page is Intentionally Left Blank]
3
Farber & Weld
4
I. Introduction
University parking is an issue that affects college campuses and cities across the nation. Whether it is sufficient parking availability, enforcement of parking rules and regulations, or the nominal price of parking, many different factors influence the efficiency and accessibility of university parking. Because of the higher than normal concentration of students and faculty residing on and around campus, as well as other various commercial operations focused on serving the university population, parking on and around university campuses is typically in higher demand in comparison to many other locations.1 Although there are grounds to argue parking as either a public good or a marketable commodity, it is important not to overlook the immense value they hold.2 However, due to the many different factors that affect parking on a university campus, determining the specific value of an individual parking structure or space is often a difficult task. This paper will explore the value of on-‐campus public parking in Eugene, Oregon—more specifically the sensitivity of demand to price changes and its effect on total revenue. Providing the necessary resources for commuters to conveniently park at their desired destinations typically falls under the purview of the city, with the aim to stimulate local business and generate additional income. In the case of the University of Oregon, the city of Eugene and the University of Oregon through their Department of Parking and Transportation have implemented a network of parking structures in and around campus in an effort to provide these necessary resources. However, for the purposes of this paper
1 (Dagget & Gutkowski, 1) 2 (Epstein, 2002)
Farber & Weld
5
we will focus solely on the parking infrastructure under the domain of City of Eugene Parking. Eugene City Parking is faced with two primary tasks. Firstly, like most other government entities, they have the responsibility of serving the public, which in this case translates to affordable and accessible public parking. However, Eugene City Parking is also responsible for returning a certain amount of revenue earned back to the city each year. For this reason, Eugene City Parking is forced to balance two opposing tasks: maximize revenue, while minimizing extraneous costs to consumers. Focusing on revenue maximization, it is imperative for the city parking department to fully understand the willingness of its customers to pay for the service they provide. By analyzing the changes in total revenue ex ante and ex post of a 13.33% increase in price from $1.50/hr. to $1.70/hr. implemented in the fall of 2012, we are able to provide Eugene City Parking with a comprehensive view of on-‐campus parking’s sensitivity to price changes and its implications for total revenue. Additionally this paper will examine different aspects of elasticities as applied to parking on campus, such as distance, time, and population effects, etc. Finally, this paper will investigate the effect a simulated price increase would have on total revenue.
II. Literature Review
There are numerous studies available that look at the issue of city parking. Although
none of them specifically address price elasticity at a university campus, they do offer insight into the important features that contribute to public parking demand. Litman (2013) discusses transport demand elasticities as a whole, including the effects of parking fees on demand, by using case studies from around the world. The report analyzes various
Farber & Weld
6
studies of parking price elasticity and concludes that nominal increases in the price of parking generally decreases parking demand due to its direct cost to motorists. It particularly looks at the effects of price changes in parking on drivers behavior, i.e. switching to public transportation or forgoing solo-‐travel in favor of carpooling. However, the study is complied by an urban planner with no formal economic background, and therefore takes the issue on in a more public policy/urban planning setting with little emphasis on economic implications or methodology. Stepping away from the public policy view, Clinch & Kelly (2003) looked specifically at parking elasticities in downtown Dublin, Ireland ex ante and ex post of a citywide 50% increase in parking prices. It specifically looks at the price elasticity of demand based on average parking levels for specific periods of time during 5-‐ and 6-‐day intervals. The authors found an aggregate 4.18% drop in parking activity and 16.5% percent drop in duration of stay as a result of the price increase. Nevertheless, they also found that the price increase raised revenues by 21%, and that even after the increase, parking prices were still not maximizing revenue. They further showed that the price elasticity of demand averages -‐0.29, pointing to relative inelasticity. A similar study, Miller & Everett (1982), examined the effects of a price increase and parking-‐subsidy decrease in the Washington D.C. metro area in 1982. The paper uses a before and after control group structure to show that price increases lead to a sharp increase in higher-‐occupancy modes of transportation, though the results were not fully uniform. Though this and Clinch & Kelly (2003) provide important information about city parking, both were focused on larger cities, which may lead to results that vary drastically from a smaller urban region such as Eugene.
Farber & Weld
7
At the aggregate level, Concas & Nayak (2012) analyzed the variation in results across past literature regarding parking elasticities. They found that results vary across different regions, data collection methods, and techniques of estimation, also stating that elasticities are site-‐specific and vary based on policy. They finally, arrived at a baseline elasticity of -‐0.39, consistent with previous estimates. Specifically examining the University of Oregon, Barrar (1978) applies a multinomial logit model to explain an individual’s transportation choices when they commute to and from campus. The study focuses on the utility gained by the different modes of transportation, as individuals act to maximize their utility at the margin, stating that individuals, “derive utility from the attributes of each commodity rather than from the commodity itself.“3 The reason the majority of individuals choose to use a personal vehicle is due to the comfort and convenience they provide, therefore deriving the most utility from personal transportation. Barrar proposed a utility function that he believed explained an individuals decision making process when traveling, which includes such variables as purpose of travel, mode of travel, frequency of travel, time of day, destination of travel, route of travel, vehicle ownership, work and residential location. The multinomial logit model describes an individual’s choice probability for exclusive alternatives. However, “...because alternatives are mutually exclusive, the choice of one alternative precludes the selection of any other. Alternatives are ‘competitive’ in this sense…within the MNL framework there is no pattern of differential substitution among the various [competitive] alternatives; there is always equal proportional substitution.”3 It continues to show that this creates a bias because alternatives are viewed as ‘similar’ instead of opposite, and, “if 3 (Barrar, 30-‐31)
Farber & Weld
8
alternatives are perceived as ‘similar’, the definition of the number of available alternatives becomes obscured resulting in the observed bias.”4
Again looking at the university level, Lipscomb & Kofor (2011) directly explores
university parking from the perspective of a university operated parking program. The study clearly differentiates students from faculty and staff, finding that the two groups have different willingness’ to pay for on-‐campus parking passes. The authors were able to determine the mean willingness to pay and found that students have a higher willingness to pay than faculty/staff, and that students also have a higher price change acceptance rate, for reasons other than income. These findings are useful when analyzing the effects parking price increases have on the separate groups, faculty and students. Integrated Urban Traffic Management (1978) explores the different characteristics of urban traffic management, breaking them down into eight separate groups: time/place of trip, choice of mode of transportation, choice of route, street usage allocation, junction control (signals), parking control, safety, and environmental problems.5 The element most applicable to our research is parking control, consisting of parking supply, fee schedules, and time restrictions. The paper outlines that efficient parking is two-‐fold: It first implies uncongested roads and parking structures. Second, it means those who are willing to pay receive the parking. Furthermore, “The motivation for road pricing is rooted in the economic idea of efficiency. The transport system will be used efficiently only if the potential trip maker perceives the full cost of his journey”.6 The paper also discusses how
4 (Barrar, 39) 5 (Integrated Urban Traffic Management, 15) 6 (Integrated Urban Traffic Management, 31)
Farber & Weld
9
choices of mode of transportation also incorporate alternatives and attractiveness of particular methods, consistent with the findings in Barrar (1978). The Transit Cooperative Research Program conducted by Federal Transit Administration and Transportation Research Board (2005) also provided useful insights. Parking Prices and Fees section provides invaluable information on elasticity rates based on rate changes, location, and available substitutions. It analyzes the effect that price increases have on individuals’ parking frequency, duration, and willingness to seek alternative modes of transportation (carpool, public transportation, walk, etc.). The study also takes into consideration an individual’s income and value of their time, finding that as a person’s income rises, their sensitivity to price changes decreases, which may offer useful insight for University of Oregon parking. The general consensus of this study is that the elasticity rate falls between -‐.1 and -‐.6 depending on multiple factors, but the average rate tends to be approximately -‐.3, denoting inelastic demand. Though not specifically directed toward university parking, this is a useful source that provides a helpful starting point for our research. The key to determining whether these previous case studies can be compared to the University of Oregon accurately is through a comprehensive understanding of the unique aspects of the University of Oregon parking system. In order to present credible predictions for the future, we must understand the trends of this specific scenario. The Role of Economic Studies in Urban Transportation Planning by the U.S. Bureau of Public Roads differentiates between “projecting” and “forecasting” for future changes in transportation demand.7 Projecting implies simply plugging in known numbers into an equation. 7 Meck (1965)
Farber & Weld
10
Forecasting, on the other hand, is using outside information to create a more realistic look into the future. In order to correctly forecast, one must be knowledgeable of the trends of transportation, and separate correlation between variables and true cause and effect.8 Applying this method to our forecast, we can provide a more accurate view of the parking landscape
After surveying the existing literature four things are clearly evident: First, parking
demand is in general relatively price inelastic, and although increasing the price of parking may negatively affect quantity of demand, parking revenue tends to increase, though the extent of these depend greatly on the situation. Secondly, an individual’s utility function directly determines what mode of transportation they will choose to take. Third, there are multiple substitutes of driving, and increases in parking prices tend to lead to a higher rate of substitution. Lastly, university faculty/staff members and students respond differently to price changes on campus, which may in turn influence the migration to alternative parking options, in this case city maintained meter parking. Utilizing the information and findings provided in these studies as a building block and applying it to our own data, we hope to provide Eugene City Parking with a comprehensive model showing the price elasticity of University of Oregon campus area parking.
III. Methodology A. Research Questions Through this paper, we wish to examine three main questions regarding the price elasticity of public parking on campus:
8 (Meck, 20)
Farber & Weld •
11
Determine the price elasticity of demand for public parking on and around the University of Oregon campus. Focus on the application to total revenue, specifying effects price changes have on total revenue while controlling for exogenous variables.
•
Examine marginal effects of distance, time, and population effects, as well as other possible demand shifters.
•
Perform conditional forecast of price increase from current price of $1.70/hr. to $2.00/hr. B. Empirical Model
The value of a parking zone i at time t is defined by the relationship:
𝑇𝑜𝑡𝑎𝑙𝑅𝑒𝑣!" = 𝛽! + 𝛽! 𝑃𝑟𝑖𝑐𝑒!" + 𝛽! 𝑀𝑎𝑗𝐸𝑣𝑒𝑛𝑡! + 𝛽! 𝑍𝑜𝑛𝑒! + 𝛽! 𝑇𝑖𝑚𝑒! + 𝛽! 𝑆𝑒𝑎𝑠𝑜𝑛! + 𝛽! 𝑆𝑡𝑢𝑑𝑒𝑛𝑡𝐸𝑛𝑟𝑜𝑙𝑙𝑚𝑒𝑛𝑡! + 𝛽! 𝐷𝑎𝑦𝑜𝑓𝑊𝑒𝑒𝑘! + 𝛽! 𝐹𝑖𝑛𝑎𝑙𝑠! + 𝛽!" 𝐵𝑟𝑒𝑎𝑘! + 𝜀!
where 𝜀! is our error term, i represents a particular zone, and t represents a one-‐hour period. The following table gives a description for all variables: Table 1. Variable Priceit MajEventi Zonei Timet StudentEnrollmentt Seasont DayofWeekt Finalst
Description of Variables Explanation Price of parking per hour at time t Dummy variable indicating occurrence of major event on campus (=1 if yes, =0 if otherwise) Zone observation took place in Variable denoting time of day revenue was recorded University population at time t Dummy variable indicating observation was recorded during fall (=1 if yes, =0 if otherwise) Dummy variable indicating if observation occurred on weekday, Saturday, or Sunday (=1 if yes, =0 otherwise) Dummy variable indicating observation occurred during finals week (=1 if yes, =0 otherwise)
Farber & Weld Breakt
12
Dummy variable indicating observation occurred during school break (=1 if yes, =0 otherwise)
C. Variable Description Based on our hypothesis and previous research, we predict certain outcomes for
each individual explanatory variable, controlling for distance, time and population factors; demand shifters; and monetary costs associated with parking described below: Price: As Litman (2005) and Concas & Nyak (2012) found, we expect price to have an inverse relationship with the demand for a given parking spot, with most literature pointing to an average price elasticity hovering around -‐0.3. Major Event: We expect that the observation of a major event, such as a track meet, football game, or basketball game on a given day will share a direct relationship with demand for on campus parking, increasing the demand for parking. Zone: Representing the zone which an observation took place, we are able to control for distance effects. We expect that as the distance increases, the demand for that zone will decrease with it. Time of Day: Based on the results of Kelly & Clinch (2005), we expect a similar result where parking demand is highest in the morning, as students and faculty arrive for morning classes or to take advantage of the flat fee for all-‐day parking, tapering off throughout the afternoon and evening.
Farber & Weld
13
University Population: We expect demand for parking to be highly correlated with university population, as increases in student populations brings increases in cars, therefore increasing the demand for parking. Season (Fall, Winter, Spring, Summer): We expect demand to be at its highest during fall, winter and spring, before tapering off in summer as many students return home or engage in other activates outside of the Eugene area. Day of Week: We expect demand to be highest during the week, when class is in session, and substantially lower on Saturday and Sundays as the traffic to and from campus decreases. Finals: We expect demand to decrease during finals week, as there is no class, and students gradually leave the Eugene area to return home for breaks throughout the week, causing the majority of demand for on-‐campus parking to temporarily dissipate. Break:
Again, we expect that due to the significant decrease in local population and need to
be on campus, we expect total revenue to fall during breaks from school (spring, summer, and winter breaks).
D. Data Description: The main source of data is the city of Eugene’s IPS system, from which total revenue is aggregated into hourly revenue summaries containing 317,856 individual observations across a 32-‐month period from September 2010 to April 2013. It is important to note that
Farber & Weld
14 the meters only operate for 13 hours a
Figure 1 Total Revenue Across Time 30
day from 07:00 to 20:00, and therefore
Mean of Total Revenue 10 20
observations outside the hours of operation take on zero-‐values, which will be discussed further in next section. It is
0: 0 1: 0 0 2: 0 0 3: 0 0 4: 0 0 5: 0 0 6: 0 0 7: 0 0 8: 0 0 9: 0 10 00 : 11 00 : 12 0 0 : 13 00 : 14 00 : 15 00 : 16 00 : 17 00 : 18 00 : 19 00 : 20 00 : 21 00 : 22 00 : 23 00 :0 0
0
also of note that all meters on-‐campus Hour of Day
Figure 2
operated by the city accept both coins and credit cards, therefore mitigating a
40
Total Revenue Across Zones
Mean of Total Revenue 10 20 30
reliance on those who carry change for revenue, and making parking much more accessible. Figure 1 shows the
th
ily
ar Ki d nc a O id rc ha Pa rd tte rs on Vi W llar d al nu tS tr
Zone
H
th
E. 18
th
E. 17
th
E. 15
th
E. 14
th
E. 13
r
E. 12
de Al
Ag
at e
0
distribution of total revenue across the 24 hours and Table 2 further details total revenue over time: Table 2. Summary Statistics (Time) Time 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00
Obs. 13244 13244 13244 13244 13244 13244 13244 13244 13244 13244 13244 13244 13244 13244 13244 13244
Mean 0.0028 0.0011 0.0004 0.0006 0.0814 1.5829 16.5569 27.5157 26.5339 20.0357 24.0352 21.9127 23.3585 16.8509 17.1826 12.5706
Std. Dev. Min. Revenue/Hour Max. Revenue/Hour 0 2.25 0.0482 0 1.35 0.0269 0 1.25 0.0189 0 1.25 0.0212 0 35 0.9303 0 94.1 4.8545 0 317.95 32.5260 0 283.4 46.0031 0 405.35 41.3788 0 287.35 29.4135 0 293.9 34.6601 0 267.35 29.5912 0 241.4 33.2191 0 283.7 24.0312 0 264.1 26.6566 0 245.05 19.1682
Farber & Weld 17:00 18:00 19:00 20:00 21:00 22:00 23:00
13244 12.2999 19.3306 13244 7.8215 12.8225 13244 2.3970 3.9319 13244 0.0406 0.1732 13244 0.0182 0.1129 13244 0.0101 0.0892 13244 0.0069 0.0716
15 0 0 0 0 0 0 0
177.85 132.7 48.35 4.1 2.8 3.25 1.8
Observations are additionally separated into districts and zones. This paper will focus specifically on the campus district, which is further partitioned into 14 individual zones: Agate, Alder, E. 12th, E. 13th, E. 14th, E. 15th, E. 17th, E. 18th, Hilyard, Kincaid, Orchard, Patterson, Villard, and Walnut St. Figure 2 above shows the distribution of revenue over zones and Table 3 below summarizes the data by zone: Table 3. Summary Statistics (Zones) Zone Mean Std. Dev. Min. Max. Num. of Revenue/Hour Revenue/Hour Meters Agate St. 13.00745 24.0705 0.00 293.40 75 Alder St. 4.95858 12.00307 0.00 104.80 10 E. 12th Ave. 8.391354 12.377443 0.00 71.40 22 E. 13th Ave. 16.18966 24.29274 0.00 272.30 52 E. 14th Ave. 14.06947 21.58082 0.00 165.40 37 E. 15th Ave. 23.00794 40.05337 0.00 405.35 109 E. 17th Ave. 2.174943 7.237046 0.00 161.35 25 E. 18th Ave. 10.96401 22.88102 0.00 303.65 71 Hilyard St. 1.521644 2.560688 0.00 15.80 9 Kincaid St. 33.2851 49.3523 0.00 317.95 102 Orchard St. 1.952621 7.49574 0.00 163.40 29 Patterson St. 0.318 0.8893352 0.00 9.40 4 Walnut St. 4.346738 10.57665 0.00 120.90 37 Villard St. 0.457622 4.260945 0.00 263.35 21 For the purposes of this paper, we assume that meters within the same zone are valued equally, and therefore total revenue data is aggregated by zone instead of examining meters at the individual level. The remaining exogenous variables were sourced externally. Student population and events for major events at the university were taken from the publically available data posted on the university and athletic department websites. Finally,
Farber & Weld
16
distance values were calculated by hand using midpoints of each zone from maps and data provided by the city of Eugene. The data exhibits some minor problems that required correction. During the summer of 2011, a major construction overhaul altered the parking infrastructure of 13th and Alder, a frequented parking destination on the west edge of campus. This included changing one block of the E. 13th zone from a typical individual coin meter system and installing a multi-‐space pay station. For this reason, we have opted to exclude the block from our data, but due to this area being only a small portion of total revenue for the campus district as a whole, we do not expect the exclusion to alter our results. With respects to distance, three zones, E. 13th, E. 15th, and E. 18th streets operate on both east and west sides of campus, leading to a zone-‐midpoint that is in a deceptively central location, and generating distance values that are downward biased. Because of the unreliability of these distance values, the three zones have been combined together and utilized as the reference category not included in the regression. This phenomenon will be further discussed in later sections. Finally, within the date range of our study, only one price increase was implemented, amounting to only a $0.20 increase from $1.50/hr. to $1.70/hr. Therefore, the specific demand curve for parking on campus and its associated elasticity will be difficult to determine. E. Model Description Building on the empirical model described in section B above, we found it necessary to alter the structure and implementation of our model to better fit our data. Noting that of the 317,856 observations for total revenue, 62.25% are zero-‐value observations, we found that this high of a zero-‐value proportion was unsuitable for OLS regression. Figure 3 shows
Farber & Weld
17 the distribution of actual total revenue
Figure 3 Distribtuion of Total Revenue
.08
.1
values, and zero-‐values are again seen to
Density .06
occupy a large proportion of the data.
.04
For this reason, we opted to use the
0
.02
Tobit model, designed to mitigate large 0
100
200 Total Revenue
300
400
Actual Values
zero-‐biases outlined in Tobin (1958), in lieu of the unreliable OLS model. The
Figure 4
Tobit model assumes that both a latent
.008
Distribution of Total Revenue
.006
and observed variable exist, which are
Density .004
equal for positive values, but the
.002
observed variable takes a zero for all
0
negative values, referred to as censoring -300
-200
-100 Total Revenue
0
100
from the left. After applying the Tobit
Fitted Values, Tobit Model
model to our original regression model, we find that the fitted values are distributed much more normally, as seen above in Figure 4. Further examination shows that, according to the findings in Amemiya (1984), if the Tobit model is subject to heteroskedasticity or has a non-‐normally distributed residual, it yields inconsistent estimates. Using the tobcm command in Stata as detailed in Drukker (2002), a test for normally distributed residuals in Tobit regression models, we find that 𝑃 > 𝜒 ! = 0.00, unfortunately rejecting the null hypothesis of normally distributed residuals. However, using the comparison model shown in Greene (1981), where OLS models converge to 𝜌𝛽!"#$% , where 𝜌 is the percentage of uncensored data within the sample, we find that the results from our Tobit model are consistent with the OLS model,
Farber & Weld
18 and further confirming the normality of our residuals with the Q-‐Q residual plot shown in Figure 5 above. Finally, to correct for heteroskedasticity, all regressions were run using White’s robust standard errors detailed in White (1980).
IV. Econometric Results A. Price First applying the Tobit model to our regression we find that price increases of $1.00 would bring in a total $11.71/hr./zone, or a total of $163.94/hr. across all zones. This is a significant increase in light of current total revenue per hour is only $134.68 on average across all zones. Comparing results from both Tobit and OLS models seen in Table 4 below, we find that OLS results in only a mere $5.33/hr./zone, totaling in only $74.62/hr. across all zones.
Variable Price Constant Observations
Table 4. Regression Results Tobit Coef. 11.706 (11.71)** -‐120.831 317,856
OLS Coef. 5.330 (10.46)** -‐4.257 317,856
Robust Standard Errors in Parentheses p |t| [95% Conf. Interval] DistW 0.0031569 0.000201 15.71 0 0.002763 0.0035508 DistC -‐0.019082 0.0003169 -‐60.21 0 -‐0.019703 -‐0.01846 DistE 0.0069074 0.0001398 49.41 0 0.0066334 0.0071814 Standard Errors are Robust
Farber & Weld
21
It is crucial to note that distance is not the only determining factor of a zone’s revenue. Each zone possesses a varying quantity of meters; those with a higher quantity of meters are in turn able to accommodate more vehicles, and therefore generate more revenue per hour than other zones with fewer spots [See Table 2]. While this is controlled for by zone fixed effects in the above regressions using Zonenumi, it is not accounted for in the distance specific regression. For this reason, we felt it necessary to include a variable to account for the quantity of meters in a given zone, Metersi, therefore controlling for the variation caused. The coefficients presented in the above table represent the change in hourly revenue per zone as they increase one foot away from the points in east, west, and central campus. These results curiously show that as a zone increases in distance from east and west campus, their hourly revenue increases. Conversely, increasing the distance from central campus causes revenue to decrease. These results also suggest that the smaller the distance between a zone and central campus, the higher its hourly revenue. In terms of daily parking on days there is no major event on campus, the results align with the notion that commuters prefer to park closer to central campus, which captures the majority of educational buildings on campus.
C. Distance-‐Event Interactions Distance from particular campus locations is not solely a concern for the daily commuter. It is also a major factor for people driving to the university district for particular events, more specifically sporting events. To test the effect distance had on hourly revenue during days with sporting events, we created interactive variables for basketball games, football games, and track meets with the three distance variables—east, west, and central. The effects are shown below in Table 7:
Farber & Weld
Total Rev. DistW DistC DistE DistEFootball DistWFootball DistCFootball DistEBasketball DistWBasketball DistCBasketball DistCTrackmeet DistWTrackmeet DistETrackmeet
Coef. 0.0030361 -‐0.0195583 0.0072427 0.0000452 0.005648 -‐0.0064784 -‐0.0011992 0.0033146 0.002385 0.0149529 -‐0.0049228 -‐0.0079638
Table 7. Regression Results Std. Error t 0.0002096 14.49 0.0003313 -‐59.04 0.0001448 50 0.0009323 0.05 0.0015064 3.75 0.0023418 -‐2.77 0.0006656 -‐1.8 0.0010284 3.22 0.0016583 1.44 0.0017761 8.42 0.001119 -‐4.4 0.0006902 -‐11.54
P> |t| 0.000 0.000 0.000 0.961 0.000 0.006 0.072 0.001 0.150 0.000 0.000 0.000 Standard Errors are Robust
22
[95% Conf. Interval] 0.0026254 0.0034469 -‐0.0202076 -‐0.018909 0.0069588 0.0075265 -‐0.0017822 0.0018725 0.0026954 0.0086005 -‐0.0110683 -‐0.0018885 -‐0.0025037 0.0001054 0.001299 0.0053302 -‐0.0008652 0.0056352 0.0114717 0.018434 -‐0.007116 -‐0.0027296 -‐0.0093166 -‐0.0066111
The results show that all interactive variables were statistically significant with the exception of those representing the eastern campus variable interacted with football games, central campus interacted with basketball games, and central campus interacted with basketball games. This shows that distance from east campus is not a deciding factor for football and basketball game commuters, and distance from central campus is not a factor for basketball game commuters. This can be attributed to drivers’ higher willingness to accept larger distances when parking for such events as sporting events, especially since on-‐campus parking is usually experiencing busier than average use at those times. The most curious result from this is the variable denoting the interaction between east campus and basketball games. With Matthew Knight Arena, the location of the basketball games, acting as the point of interest for east campus, the expectation would be that the distance of a zone to where the basketball game is held would have a significant influence on demand and hourly revenue, but the results suggest otherwise.
Farber & Weld
23
Other noteworthy results from this regression include the track meet variables. Hayward Field, where University of Oregon track meets are held, is located roughly a block away from both the central and east campus point. Our results show that as a zone’s distance from east and west campus increases on track meet days, its hourly revenue decreases. On the other hand, distance from central campus has a significantly positive effect on hourly revenue. This can be interpreted as commuters not wanting to park near central campus, opting instead to park on the outskirts of University District closer to Hayward Field.
D. Time Considering hourly revenue is determined by the demand for parking at a given time, it is important to understand how time relates to demand, and in turn how total revenue changes over time. Because commuters to the campus district are primarily students and faculty, we expect there is a reliable pattern to traffic patterns to and from campus that affect parking use. We anticipate that hourly revenue would be at its highest in the early to mid-‐morning period, as people arrive early to take advantage either the $8.50/5 hour or $12.00/10 hour deals offered, before tapering off as the day goes on. The installation of meters that accept both credit cards and coins makes these discounted deals much more accessible to the general population, as large quantities of coins are no longer necessary to feed the meter. By generating dummy variables for each hour, we see that hourly revenue is at its highest between 8:00am to 1:00pm, consistent with our hypothesis. Furthermore, as the day progresses, revenue per hour/zone decreases, with a distinctive decrease at 8:00pm when meters end operation. Table 8 below displays the regression results:
Farber & Weld
Time 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00
Table 8. Regression results Coef. Std. Error t -‐4.648737 2.019054 -‐2.3 -‐15.78419 2.520904 -‐6.26 -‐28.67553 3.627735 -‐7.9 -‐24.73217 3.122558 -‐7.92 6.89041 1.631757 4.22 48.46292 1.373204 35.29 89.4661 1.411544 63.38 106.4348 1.446544 73.58 106.707 1.421794 75.05 100.0869 1.385921 72.22 105.8886 1.397073 75.79 103.8784 1.383962 75.06 105.7321 1.392731 75.92 97.83315 1.376169 71.09 97.96724 1.382226 70.88 91.75185 1.370575 66.94 90.77356 1.372107 66.16 82.89803 1.366684 60.66 73.15335 1.362034 53.71 32.36169 1.41678 22.84 19.26443 1.489658 12.93 10.63926 1.581744 6.73 4.611357 1.692632 2.72 Standard Errors are Robust
24
P>t 0.021 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.006
Of course, we expect that commuters travel to campus for non-‐class reasons as well. To test whether time influenced hourly revenue differently during days when class is in session and days when class is not in session, we used variables to interact time with weekday, time with Saturday, and time with Sunday. As the regression results in Table 9 below show, hourly revenue decreases at a faster rate as time passes on a weekday rather than during weekends, suggesting that hourly revenue is higher earlier in the day on weekdays than on weekends: Table 9. Regression Results Total Rev. Coef. Std. Error t P> |t| [95% Conf. Interval] Timeweekday -‐2.988949 0.1580454 -‐18.91 0.000 -‐3.298714 -‐2.679185 Timesaturday -‐1.118112 0.1621213 -‐6.9 0.000 -‐1.435865 -‐0.8003589 Standard Errors are Robust
Farber & Weld
25
We also expect that the time of day affects hourly revenue differently based on a zone’s location. Using the regression containing distance variables, and controlling for the number of meters per zone, Time*Distance variables were generated to test this hypothesis. The results show that distance does have a significant impact on how a zone’s hourly revenue changes with the time of day, and that the hourly revenue of zones closer to central campus are influenced more by time than zones further away shown in Table 10: Table 10. Regression Results Total Rev. Coef. Std. Error t P> |t| [95% Conf. Interval] TimeDistW 0.0002181 0.0000558 3.91 0 0.0001088 0.0003274 TimeDistC 0.0008421 0.0000895 9.41 0 0.0006666 0.0010175 TimeDistE 0.0001568 0.0000319 4.92 0 0.0000943 0.0002193 Standard Errors are Robust
E. Student Enrollment Another influential factor in accurately forecasting future revenue is the changes in student population. Originally hypothesizing that since University of Oregon students make up a large percentage of the total demand for university district parking services, a rise in the number of students may cause an initial increase for demand, the data reveals the opposite. In fact, it appears that each additional student attending the University results in a $0.005 decrease in total parking revenue per hour per zone. While this effect may be small, we expect the it to increase when magnified over time by the forecasted trends student population increases, shown in Figure 6 & Table 12 in Appendix B. Additionally, this inverse relationship can be explained by the traditional “Tragedy of the Commons” dilemma; as the number of students commuting to campus rises, the supply of parking, held constant, must be allocated amongst more people. Commuters may then become discouraged by the increased congestion and inability to find parking, and resort to taking
Farber & Weld
26 substitute modes of transportation, i.e.
Figure 6 Student Enrollment
biking, walking, public transportation.9 It Student Enrollment 20000 22000 24000 26000
is imperative to control for this effect, as we cannot credibly predict the affect of future price changes on revenue without 2000
2002
2004
2006
2008
2010 Year
2012
2014
2016
2018
2020
Actual & Forecast
also accounting for the predicted increase in student population.
According to Oregon University System (OUS) 2012 Factbook, the number of students, and thusly the necessary faculty and support staff, is set to steadily increase for the next nine years as shown in Figure 6.
F. Further Analysis It is clear from our regression that demand is not constant across zones; those directly near campus, particularly Kincaid St. and 15th Ave, have a higher concentration of people wanting to park in comparison to zones further away from campus. The large percentage of observations that have zero hourly revenue, denoting no earned revenue for that period, even during hours of operation data, also shows that parking is not being utilized to its maximum capacity. Therefore it can be inferred that revenue is also not being maximized. Wilson (1992) researches the utilization adjusted break-‐even fee; the concept that each parking space has a certain price that, when aggregated across all parking spaces, would result in break-‐even profit. However, this only functions under the assumption that all spaces are 100% occupied. If this is not the case, as is in Eugene, the break-‐even price
9 (Integrated Urban Traffic Management, 61)
Farber & Weld
27
must increase for spaces that are being used to compensate for those that are left empty.10 If this is not done, vital revenue is being wasted. The key to handling increasing population effects is either increasing parking availability in areas of highest demand, or developing a way to incentivize commuters to park in the zones further away from campus. The latter option would allow for the utilization of a greater quantity of meters varied across all zones, and also has the potential to offer drivers a lower overall cost to park in select zones. One possible method is to implement a varied fee schedule, decreasing hourly fee rates for the underutilized areas, thus incentivizing drivers to park in underutilized zones. In our opinion, with regards to those zones that are being underutilized, creating even a small amount of hourly revenue is still better than making zero hourly revenue. Doing so would also open up space in higher-‐ demand zones for those with a larger willingness-‐to-‐pay, simultaneously alleviating congestion issues and increasing parking efficiency and effectiveness.
G. Conditional Forecast Utilizing the results found through our research, we are able to construct a rudimentary forecast for the city of Eugene in the event of a planning increase in parking prices from the current $1.70 to $2.00 at the end of 2013. Taking into account this potential 17.64% increase in the price of parking, we were able to simulate multiple scenarios based on the price coefficient and its respective confidence interval. Beginning with a simple simulation, we found that, ceteris paribus, a $0.30 increase in parking price per hour will result in a baseline $2.06/hour increase in total revenue for each respective zone. This results in an aggregate increase of $137,230.40 annually across 10 (Wilson, 25)
Farber & Weld
28
all zones. Compared with total revenue for the campus district in 2012: $1,198,186.70, the potential increase in price results in an 11.45% increase in total revenue. Further examining the effects of a $0.30 price increase, we are able to utilize the confidence interval associated with our price coefficient to construct a “best/worst case scenario” simulation. We found that along with the previously stated price increase, the city of Eugene can expect increases in total revenue ranging from $1.72 to $2.41 per hour for a given zone, or a total of $114,216.95 to $160,196.97 annually across all zones. From these results, we can calculate both the price elasticity of demand, as well as the elasticity of total revenue. We find that the elasticity of demand for public parking on campus is approximately -‐0.3, pointing to relative inelasticity. This leads to a revenue elasticity of 0.65, again confirming our original hypothesis. These results are consistent with previous studies such as Clinch & Kelly (2003) and Concas & Nayak (2012), with the general consensus centering on a price elasticity of -‐0.3.
V. Conclusion From our results we have explained many valuable aspects to public parking in Eugene, Oregon. Firstly, we found that increases in price and the subsequent increase in total revenue point to relatively inelastic demand averaging -‐0.3. Furthermore, when looking at distance in relation to changes in total revenue, we found that increases in distance per feet from key locations on campus for a given zone result in a reduction of total revenue for the central campus hotspot, but showed slight increases for the east and western campus hotspots. We next looked the distribution of total revenue across times of day, and found revenue at its peak during the period from 8:00am to 1:00pm. We additionally looked at the interactions between distance and major events on campus, as
Farber & Weld
29
well as time of day and day of week, with mixed results. Finally, we looked at the effect of a future price change of $0.30 on total revenue, and found that revenue increased $2.06 for a given zone every hour, totaling in $137,230.40 annually. This paper outlines a comprehensive framework for the study of price elasticity of parking demand in Eugene, Oregon. Our hope is that the City of Eugene and other organizations can utilize the results of this paper to improve the effectiveness of their parking policies. Additionally, we hope that future inquiries into the nature of demand elasticity for campus-‐area parking can apply the basic techniques and methodologies used in this paper as a jumping off point.
Farber & Weld
30
VI. Works Cited Amemiya, Takeshi, 1984. "Tobit models: A survey," Journal of Econometrics, Elsevier, vol. 24(1-‐2), pages 3-‐61. Barrar, Jack E. A Multinomial Logit Model of the Journey to School: A Study of Individual Travel Demand. , 1978. Print. Clinch, Peter J., Kelly, Andrew (2003), Temporal Variance Of Revealed Preference On-‐Street Parking Price Elasticity, Department of Environmental Studies, University College Dublin (www.environmentaleconomics.net). Concas, SisinnioNagesh Nayak (2012), A Meta-‐Analysis of Parking Price Elasticity, presented at the Transportation Research Board Annual Meeting (www.trb.org); http://amonline.trb.org/1spbip/1spbip/1. Dagget, j. and R. Gutkowski. University Transportation Survey: Transportation in University Communities. City of Fort Collins, CO.: Mount-‐Plains Consortium, Colorado State University, 2002. Drukker, David M. "Bootstrapping a Conditional Moments Test for Normality After Tobit Estimation." The Stata Journal 2.2 (2002): 125-‐39.
Epstein, Richard A. "The Allocation of the Commons: Parking on Public Roads." The Journal of Legal Studies 31.S2 (2002): S515-‐544. JSTOR. Web. 10 June 2013. . Federal Transit Administration. Transportation Research Board of the National Academies. (2005) Parking Pricing and Fees: Traveler Response to Transportation System Changes. Retrieved from http://onlinepubs.trb.org/onlinepubs/tcrp/tcrp_rpt_95c13.pdf Greene, William H, 1981. "On the Asymptotic Bias of the Ordinary Least Squares Estimator of the Tobit Model," Econometrica, Econometric Society, vol. 49(2), pages 505-‐13, March. Integrated Urban Traffic Management: A Report. Paris: Organisation for Economic Co-‐ operation and Development, 1978. Print. Johansson, B, and L.-‐G Mattsson. Road Pricing: Theory, Empirical Assessment, and Policy. Boston: Kluwer Academic Publishers, 1995. Print. Lipscomb, Clifford, and Brandon Kofor. "Conservative Dichotomous Choice Responses in the Active Policy Setting: The Case of University Parking." Applied Economic Letters. (2010): 1-‐4. Web. March. 2013.
Farber & Weld
31
Litman, Todd. "Understanding Transport Demands and Elasticities." Http://www.vtpi.org. Victoria Transport Policy Institute, 12 Mar. 2013. Web. 19 Mar. 2013. Meck, J. P., & United States. (1965). The role of economic studies in urban transportation planning. Washington: U.S. Dept. of Commerce, Bureau of Public Roads. Miller, G, & Everett, C, (1982). Raising commuter parking prices—an empirical study. Transportation 11: 105-‐-‐ 129 White, Halbert, 1980. "A Heteroskedasticity-‐Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity," Econometrica, Econometric Society, vol. 48(4), pages 817-‐38, May. Willson, Richard W. Suburban Parking Economics and Policy: Case Studies of Office Worksites in Southern California. Washington, D.C: Federal Transit Administration, University Research and Training Program, 1992. Print.
Farber & Weld
32
Appendix A: Regression Results
Table 11. Regression Results Variable
Tobit
Price 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00
11.706 (11.71)** -‐3.992 (1.99)* -‐15.237 (6.09)** -‐27.658 (7.71)** -‐23.998 (7.74)** 6.819 (4.25)** 46.487 (34.56)** 87.667 (63.96)** 104.892 (74.72)** 105.048 (75.97)** 98.169 (72.86)** 104.003 (76.71)** 101.938 (75.75)** 103.854 (76.75)** 95.774 (71.43)** 95.880 (71.32)** 89.429 (66.91)** 88.403 (66.01)** 80.333 (60.07)** 70.149 (52.58)**
OLS 5.330 (10.46)** -‐0.001 (0.01) -‐0.003 (0.02) -‐0.004 (0.03) -‐0.004 (0.03) 0.077 (0.60) 1.579 (12.70)** 16.553 (66.41)** 27.511 (77.78)** 26.530 (83.53)** 20.031 (90.91)** 24.031 (93.37)** 21.908 (100.11)** 23.354 (94.97)** 16.847 (93.20)** 17.178 (86.45)** 12.566 (82.45)** 12.296 (77.08)** 7.817 (58.30)** 2.393 (20.22)**
Farber & Weld
20:00 21:00 22:00 23:00 Spring Fall Winter Student Enrollment Alder E. 12th E. 14th E 17th Hilyard Kincaid Orchard Patterson Villard Walnut Agate Weekday Saturday Finals Winter Break
33 31.820 (22.80)** 19.123 (13.00)** 10.426 (6.69)** 4.663 (2.79)** 6.548 (13.23)** 4.103 (7.22)** 4.657 (8.49)** -‐0.001 (4.63)** -‐30.680 (103.78)** -‐16.270 (75.51)** -‐4.637 (20.24)** -‐44.211 (138.45)** -‐32.495 (123.18)** 27.912 (79.23)** -‐36.721 (122.57)** -‐52.911 (168.71)** -‐28.517 (100.34)** -‐72.371 (124.05)** -‐7.527 (27.53)** 56.227 (150.90)** 42.060 (103.26)** -‐2.006 (6.38)** -‐22.332 (81.09)**
0.036 (0.28) 0.014 (0.11) 0.006 (0.05) 0.003 (0.02) 1.591 (7.76)** 0.819 (3.37)** 1.081 (4.56)** -‐0.000 (5.17)** -‐11.762 (98.07)** -‐8.329 (76.51)** -‐2.651 (19.63)** -‐14.546 (122.94)** -‐15.199 (132.08)** 16.565 (57.73)** -‐14.768 (119.80)** -‐16.403 (138.29)** -‐12.374 (100.59)** -‐16.263 (133.55)** -‐3.713 (23.91)** 11.786 (150.82)** 6.108 (57.84)** -‐1.037 (5.54)** -‐7.425 (74.24)**
Farber & Weld
Spring Break Summer Break Basketball Game Football Game Tack Meet Constant Sigma Observations R2
-‐17.276 -‐6.555 (39.89)** (42.69)** -‐13.637 -‐6.834 (9.59)** (11.43)** 6.357 1.985 (15.36)** (10.00)** 10.475 4.161 (17.44)** (14.26)** 7.545 2.507 (17.98)** (12.69)** -‐120.831 – (41.36)** – 26.271 -‐4.257 (218.07)** (3.72)** 317,856 317,856 0.37
Robust Standard Errors in Parentheses * p