MEEDDTL-Slides_Montero-PSE

Document Sample
MEEDDTL-Slides_Montero-PSE Powered By Docstoc
					        The E¤ect of Transport Policies on Car Use:
       Theory and Evidence from Latin American Cities

         Francisco Gallego          Juan-Pablo Montero               Christian Salas

                                   Department of Economics
                                P. Universidad Católica de Chile


                          May 12, 2011: PSE & SEEIDD, Paris




Gallego, Montero, Salas (PUC)       Transport Policies and Car Use    PSE & SEEIDD, May 2011   1 / 36
Motivation


      Air pollution and congestion remains serious problems in many Latin
      American cities because of steady increase in car use (Economist
      Intellingence Unit, 2010)
      Two examples of policy intervention:
             Hoy-No-Circula (HNC) in Mexico-City:
                    driving restriction established in November 1989
                    most drivers cannot use their vehicles one day of the week
                    near universal compliance (Eskeland Feyzioglu, 1997; Davis, 2008)
             TranSantiago (TS) in Santiago-Chile:
                    major public transportation reform in February 2007 (Muñoz et al.,
                    2009)
                    good idea in theory, poor implementation (Yañez et al., 2010)
                    recognized by many as a major policy failure



 Gallego, Montero, Salas (PUC)      Transport Policies and Car Use   PSE & SEEIDD, May 2011   2 / 36
Transantiago: the day after...




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   3 / 36
Table 1.1. Transport policies in Latin America: Candidates for Evaluation
      Program              City       Start date Typea Scope in force?
  Restricción Vehicular                  Santiago               Apr 1986         DR     gradual             yes
  Hoy No Circula                       Mexico D.F.              Nov 1989         DR     drastic             yes
  Operação Rodizio                      Sao Paulo               Aug 1996         DR     gradual             yes
  Pico y Placa                            Bogotá                Aug 1998         DR     gradual             yes
  Transmilenio                            Bogotá                Dec 2000         PT     gradual             yes
  Pico y Placa                           Medellín               Feb 2005         DR     drastic             yes
  Metrobus                             México D.F.              Aug 2005         PT     gradual             yes
  Restricción Vehicular                 San José                Aug 2005         DR     gradual            yesb
  TranSantiago                           Santiago               Feb 2007         PT     drastic             yes
  Pico y Placa Quito                      Quito                 May 2010         DR     drastic             yes
a DR: driving restriction; PT: public transportation reform/investment.
b The program   su¤ered a temporary interruption in June-July of 2009.

Source: Lizana (2011)




  Gallego, Montero, Salas (PUC)                 Transport Policies and Car Use    PSE & SEEIDD, May 2011     4 / 36
this paper

      have these policies been e¤ective in moving people away from their
      cars? (and hence in reducing local pollution, congestion, and carbon
      emissions?)
      hard to construct a counterfactual for policy evaluation:
      transportation systems remarkably complex and dynamic (Small and
      Verhoef, 2007)
      interested in both:
             short-run responses (e.g., Davis, 2008, for HNC)
             long-run responses (here: how long it takes for households to adjust
             their stock of vehicles). See Duranton & Turner (2011) for longer run.
      we see this paper as a theoretical and empirical exploration to the
      extent to which we can evaluate the e¤ect of these policies on car use
      HNC and TS for empirical exploration: drastic interventions like any
      other

 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   5 / 36
Quick look at the data and results




      natural candidate por policy evaluation: hourly records of vehicle
      tra¢ c from tra¢ c-control stations scattered across the city
      but some problems....besides we have this data only for Santiago




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   6 / 36
Quick look at the data and results, cont.


      alternative candidate: hourly records of carbon monoxide (CO) from
      monitoring stations across the cities
      vehicles are responsabile for 90% or more of CO emissions (Molina &
      Molina, 2002; DICTUC, 2009)
      Other papers using this high frequency pollution data: Davis (2008)
      for HNC; Chen & Walley (2010) for Taipei; Gomez-Lobo (2011) for
      TS
      we compare CO levels before and after policy implementation for
      di¤erent hours of the day and di¤erent days of the week
      empirical results




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   7 / 36
CO observations for Mexico-City




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   8 / 36
CO observations for Santiago




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   9 / 36
Summary of CO empirical results


                                         Mexico-City (HNC)
                                 short-run long-run T (months)
 peak hours (8-9 am)               -11%      +13%          11
 o¤-peak (12-2 pm)                  -6%      +11%          10
 sunday (8-10 am)                   0%       +20%          10


                                              Santiago (TS)
                                 short-run     long-run T (months)
 peak hours (7-8 am)                0%          +33%        10
 o¤-peak                             ?             ?         ?
 sunday                              ?             ?         ?




 Gallego, Montero, Salas (PUC)      Transport Policies and Car Use   PSE & SEEIDD, May 2011   10 / 36
Adaptation under HNC




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   11 / 36
Adaptation under TS (only for peak hours)




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   12 / 36
Theory: making sense of these empirical results



                 s
      household’ problem: allocation of existing vehicle capacity to
      competing uses (peak vs o¤-peak hours) and decision to adjust
      capacity in response to a policy shock
      short and long-run e¤ects on car travel for peak and o¤-peak hours
      can be di¤erent
      same model can accomodate for both HNC and TS policies
      simple model: we (partially) borrow from the non-linear pricing (or
      bundling) literature (e.g., McAfee et al., 1989; Armstrong and
      Vickers, 2010)




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   13 / 36
Table 2.1. Numerical results for a HNC-type policy
 Case variable     SR      LR LR(1) LR(2)
 1        peak    -41% 8%          8%    43%
        o¤-peak -28% 0%            0%    19%
         ∆ stock   0%      6% 21%        21%

      hard to match empirical results
                             t
      LR(1): households don’ return their cars because of
      transaction/lemon costs (Eberly, 1994)
      LR(2): LR(1) + additional cars 70% dirtier than existing stock (used
      cars); consistent with Eskeland & Feyzioglu (1997) for HNC and
      Cantillo & Ortuzar (2011) for Colombia




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   14 / 36
Numerical example 1: HNC (driving restriction)



Table 2.1. Numerical results for a HNC-type policy, cont.
 Case variable     SR       LR     LR(1) LR(1)
 2        peak    -10%      1%      3%     12%
        o¤-peak -4%         3%      4%     10%
         ∆ stock   0%       3%      7%      7%
 3        peak    -29% -20% -13%           -3%
        o¤-peak -9%         4%      9%     21%
         ∆ stock   0%      -2%      9%      9%




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   15 / 36
Numerical example 2: TS (public transportation)

Table 2.2. Numerical results for a TS-type policy
 Case variable change in relative prices SR                         LR       LR(1)         LR(2)
 1        peak    40% (price pub trans ") 2%                       31%          -             -
        o¤-peak                0%             -2%                  19%          -             -
         ∆ stock                               0%                  21%          -             -
 2        peak                0%              -5%                 30%           -             -
        o¤-peak             230%               8%                  85%          -             -
         ∆ stock                               0%                  40%          -             -
 3        peak               30%              1%                  31%           -             -
        o¤-peak               30%              0%                  35%          -             -
         ∆ stock                               0%                  25%          -             -
 4        peak               40%              4%                  20%         21%           30%
        o¤-peak              -40%             -5%                 -10%         0%            3%
         ∆ stock                               0%                   8%        12%           12%


 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011     16 / 36
Pollution Data


      Mexico-City (HNC):
             5 - 15 monitoring stations from Atmospheric Monitoring Network of
             Mexico-City
             hourly concentration measures of 5 criteria pollutants and 4 weather
             variables
      Santiago (TS):
             7 monitoring stations from CONAMA
             hourly concentration measures of 7 criteria pollutants and 5 weather
             variables
      We use (simple) hourly average of all stations in both cities and focus
      on a 4-year window around policy implementation



 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   17 / 36
Data for HNC: Nov 1987 to Nov 1991
      Series     Obs  Frequency Mean                                               Std.Dev.            Min          Max
 Carbon Monox. 33704     hour    5.102                                              2.110             0.644         20.78
 Sulfur Dioxide 33794     "      0.052                                              0.019             0.012         0.254
 Temperature    33378     "      15.94                                              4.786             0.467         30.77
 Real Humidity  32773     "      47.92                                              20.20             2.300         99.60
 Wind Speed     33671     "      4.597                                              2.032             0.400         17.60
 Wind Direction 33677     "      173.3                                              56.03             1.000          420
 Precipitation  35088     "      2.232                                              4.381             0.000         53.52
 Real Ex. Rate    48    month     7.30                                               0.65              6.28          9.41
Footnote: Pollutant levels are reported in parts per million, Temperature in celsius degrees, Humidity in percentage, Wind Speed

in kilometers per hour,

Wind Direction in azimut degrees, and Real Exchange Rate in Mexican Pesos.




  Gallego, Montero, Salas (PUC)                 Transport Policies and Car Use           PSE & SEEIDD, May 2011           18 / 36
Data for TS: Feb 2005 to Feb 2009
      Series      Obs   Frequency                                   Mean           St.Dev.            Min           Max
 Carbon Monox. 34994       hour                                     0.919           1.151           0.000          9.649
 Sulfur Dioxide  34944      "                                        9.258          5.873           0.852          102.7
 Temperature     35064      "                                        14.30           5.18            0.18          31.60
 Real Humidity   35064      "                                        66.44          16.01           13.99           98.01
 Wind Speed      35064      "                                         2.68           1.40            0.20           9.02
 Wind Direction 35064       "                                       187.08          49.98           38.62          302.14
 Precipitation   34752      "                                         0.01           0.09            0.00           4.87
 Atm. Pressure   34719      "                                       970.63          14.14           718.53          1021
 Real Ex. Rate     48     month                                       95.5            6.3            81.4           108.8
 Gasoline Price    48       "                                        517.9          517.9            368.4          721.7
Footnote: Pollutants concentration is measured in micrograms per cubic meter with the exception of Carbon Monoxide which is

measured in parts per million (ppm);

Temperature in celsius degrees, Humidity in percentage, Wind Speed in kilometers per hour, Wind Direction in azimut degrees,

Precipitation in milimeters,

Atmospheric Pressure in milibars, Real Exchange Rate and Gasoline Price in Chilean Pesos.




  Gallego, Montero, Salas (PUC)                Transport Policies and Car Use          PSE & SEEIDD, May 2011          19 / 36
why CO?


      mobile sources are responsible for 97% of it in HNC (Molina &
      Molina, 2002) and 89% in TS (DICTUC, 2009)
      contribution of light vehicles (passanger cars and commercial vehicles
      other than buses and trucks): 72% in HNC and 85% in TS
      only pollutant non-reactive on a time scale of 1 day (Schmitz, 2005)
      it has a relatively short residence time (more so under windier and
      warmer conditions)
      all these important for identi…cation because concentration can act as
      a good proxy for emissions
      unlike hourly tra¢ c observations, CO records are better at capturing
      e¤ects at the entire city level



 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   20 / 36
but....
emissions vs concentration: weekday in January 2002 (Schmitz, 2005)




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   21 / 36
Empirical Strategy



      Imposing adaptation process:
      yt = α + φytb + [a + b (t tT )]At + cTt (1                  At ) + θt + γxt +         t

             ytb : background pollution
             xt : includes …xed e¤ects (day of week, month),        weather variables
             dt : dummies for transition months
             Tt = 1 if t > tT (time of policy adoption) andzero otherwise
             At = 1 if tT < t    tA (end of adjustment phase, endogenous) and
             zero otherwise
      why linear trend θ?




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use    PSE & SEEIDD, May 2011       22 / 36
HNC Results




Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   23 / 36
HNC Estimation with more structure (all hours)
                                      Peak             O¤-Peak         Sunday
     Inmediate Impact               -.114**             -.064**          .034
                                     (.053)              (.030)         (.038)
     Adaptation Trend            3.03e-05***         2.29e-05***      2.43e-05
                                  (1.04e-05)          (7.07e-06)     (7.45e-06)
 Impact after Adaptation              .132               .106**        .196***
                                     (.083)              (.041)         (.048)
             Trend                 -9.13e-6*            7.13e-7        1.80e-6
                                   (4.60e-6)           (3.30e-6)      (4.17e-6)
             ynight                0.339***            0.153***       0.526***
                                    (0.052)             (0.044)        (0.033)
              SO2                  0.258***            0.334***        0.100**
                                    (0.046)             (0.030)        (0.038)




 Gallego, Montero, Salas (PUC)     Transport Policies and Car Use   PSE & SEEIDD, May 2011   24 / 36
Adaptation under HNC




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   25 / 36
TS Results




Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   26 / 36
TS Estimation with more structure (peak hours only)
                                      (1)               (2)
     Inmediate Impact                .045              .059
                                    (.084)            (.076)
     Adaptation Trend            3.87e-05*         4.14e-05**
                                 (2.08e-05)         (1.7e-05)
 Impact after Adaptation           .310***           .339***
                                    (.067)            (.080)
             ynight               0.414***          0.395***
                                   (0.026)           (0.024)
           ln(Pgas)              -0.637***         -0.671***
                                   (0.204)           (0.218)
              SO2                 0.503***          0.497***
                                   (0.085)           (0.086)
               Obs                   2004              2004
               R2                   0.771             0.773


 Gallego, Montero, Salas (PUC)     Transport Policies and Car Use   PSE & SEEIDD, May 2011   27 / 36
Adaptation under TS (only for peak hours)




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   28 / 36
Alternative short-run impact estimation: RDD



      RDD estimations using Imbens and Kalyanaraman (2009) with
      optimal bandwith
                    monthly      weakly           daily
 HNC peak           -0.057***    -0.25***         -0.64***
                    (0.012)      (0.017)          (0.170)
 TS peak            -0.0002      -0.55***         -0.27
                    (0.031)      (0.10)           (0.45)
      daily or weekly pollution data too noisy (lots of idiosyncratic volatitily)




 Gallego, Montero, Salas (PUC)    Transport Policies and Car Use   PSE & SEEIDD, May 2011   29 / 36
Additional Empirical Evidence for TranSantiago




      Gasoline sales
      Car registrations (new cars) and transfers (used cars)
      Price of taxi permits (medallions)
      NOTE: Eskeland & Feyzioglu (1997) and Davis (2008) discuss
      additional similar data for HNC




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   30 / 36
Gasoline Sales
      panel of monthly data of gasoline sales at the region level; include
      region and time …xed e¤ects and adjust for seasonal variation:
                                                               (1)             (2)

                                      TranSantiago           .058***         .048**

                                                             (0.018)         (0.016)

                                      GDP growth              0.054          -0.013

                                                             (0.054)         (0.200)

                                 Log of Pgas/Pdiesel

                                        Region 2            -0.663***       -0.457***

                                           ...               (0.007)         (0.004)

                                       Region 13            -0.680***       -0.370***

                                                             (0.042)         (0.039)

                                          Obs                  936            611

                                           R2                 0.945           0.957


      Consistent with CO results: from pre-TS data we get that 1%
      increase in gasoline leads to a 0.25% increase of CO at peak
 Gallego, Montero, Salas (PUC)             Transport Policies and Car Use              PSE & SEEIDD, May 2011   31 / 36
Transfers and registrations
                                       Transfers                               Registration

                             (1)          (2)          (3)            (4)          (5)            (6)

     TranSantiago         2,406.6***   1,028.5       -272.6       3,078.6***   2,201.2**      2,421.2**

                           (503.1)     (1,035.5)    (1,230.7)       (474.0)     (969.5)        (1,037.2)

       Month 1                                     1,889.8***                                 2,989.1***

                                                     (674.6)                                    (568.4)

       Month 2                                     2,594.7***                                   -316.3

                                                     (647.6)                                    (540.9)

       Month 3                                       1,032.5                                  -1,644.3***

                                                     (622.3)                                    (515.8)

       Month 4                                     2,778.7***                                   -560.5

                                                     (598.8)                                    (493.3)

       Month 5                                      1,438.1**                                 -1,212.0**

                                                     (577.4)                                    (474.0)

       Month 6                                       -702.1                                   -1,876.9***

                                                     (558.3)                                    (458.1)

 Regional Linear Trends      No          Yes           Yes            No          Yes            Yes
 Gallego, Montero, Salas (PUC)            Transport Policies and Car Use        PSE & SEEIDD, May 2011      32 / 36
Price of taxi permits (medallions), direct observations




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   33 / 36
all prices "collected"




 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   34 / 36
                                          Dependent Variable: Log (Price of taxi medallion)

                   (1)            (2)       (3)          (4)          (5)           (6)         (7)         (8)

    TS          0.709***    0.613***      0.612***    0.521***     0.623***      0.604***     0.517**    0.846***

                 (0.041)        (0.057)   (0.166)      (0.065)      (0.065)       (0.088)     (0.248)     (0.085)

Ln(#Permits)                -0.832**       0.197        -0.285       -0.541       -0.745      -0.745    -2.951***

                                (0.332)   (0.651)      (0.357)      (0.449)       (0.592)     (0.592)     (0.452)



   Trends         Yes             No        Yes          No           No            No          Yes         No

  Year F-E         No             No        No           Yes          No           Yes          Yes         No

 Model F-E         No             No        No           No           Yes          Yes          Yes         No

  Sample          430            430        430          430          430          430         430      permit-adds



    Obs           430            430        430          370          370          370         370          60

    R2            0.422         0.431      0.466        0.486        0.533        0.740        0.740       0.709




Gallego, Montero, Salas (PUC)             Transport Policies and Car Use          PSE & SEEIDD, May 2011           35 / 36
Conclusions & Some Policy Implications

      empirical results show (i) non-trivial adaptation period and (ii)
      di¤erentiated e¤ects over peak and o¤-peak hours
      CO good proxy for car use particularly at peak hours
      people adjust rather quickly: little time to react
                                                   t
      Mexico-City driving restriction (HNC) didn’ include, unlike others,
      incentives for a faster and cleaner ‡eet turnover
      welfare evaluation for TS: how much of the increase in the stock of
      vehicles was going to take place later anyway?
      policy design: more market-based oriented either working as
      substitute or complements (see London Congestion Charge and
              s
      Sigapur’ vehicle quota market).
      more on policy instruments for vehicle pollution: Feng et al., (2005),
      Fullerton and Gan (2005)

 Gallego, Montero, Salas (PUC)   Transport Policies and Car Use   PSE & SEEIDD, May 2011   36 / 36

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:3
posted:6/29/2011
language:English
pages:36