An experimental analysis of the Tiebout

1
An experimental analysis of the Tiebouts model
in a decentralized system of public goods
provision  

• Behavioural and Experimental Economics
  WorkshopI Workshop degli EconomistiSperimentali
  e Comportamentali Italiani, Florence 14-15 may
  2009.
•  
• Alessandro Innocenti, Siena University  Chiara
  Rapallini, Firenze University

2
Summary

• Aim- is to provide experimental evidence on the
  Tiebout efficiency-enhancing property in a
  decentralized system of public goods provision
• Object - a test of the Tiebout model in order to
  assess if decentralization and local sorting
  produces efficiency gains
• Key finding we show that the model fails
  because residents are different, not only in
  their preferences on local public goods, but also
  because voting and moving are not equally shared.
  

3
Talk outline

1. The Tiebout model (1956) and the recent
   contribution by Rodhe and Strumpf (American
   Economic Review, 2003)
2. The experimental design
3. Results
4. Preliminary conclusions

4
The Tiebout Model

• Tiebouts (1956) model shows that if a sufficient
  number of local communities exists to accommodate
  different types of preferences, individuals could
  move to that community whose local governments
  best satisfies his set of preferences (pattern of
  expenditures and taxes).
• Quoting Oates (1969) Tiebouts world is one in
  which the consumer shops among different
  communities offering varying packages of local
  public services and selects as a residence the
  community which offers the tax-expenditure
  program best suited to his tastes.

5
The Tiebout Model

• This is the so called voting by feet if
  households are free to move, people effectively
  sort themselves into groups that are homogeneous
  with respect to their demands for local services.

6
More recently

• Since 1956, there is a wealth of literature
  surrounding the Tiebout hypothesis, both from an
  empirical and a theoretical point of view.
• Rodhe and Strumpf (American Economic Review,
  2003) argue that longrun trends in geographical
  segregation are inconsistent with choice model
  where residential choices depends solely on local
  public goods
• More precisely they demonstrate that the secular
  decline in the mobility costs is not matched with
  an increase in the heterogeneity in local public
  policies

7
Rhode and Stumpf (AER, 2003)

• Some example of mobility costs historical data
• CAR in 1903 to drive a car cost 143,8 cents/mile
  in 1998 dollars, while in 1998 54,9 cents
• TRAIN a passenger mile cost 37,4 cents in 1895
  and 13,4 cents in 1995
• AIR a passenger mile cost 108 cents in 1929 vs.
  13,7 cents in 1995
• 3 minute of transcontinental call in January 1915
  cost 20,70 in current dollars, which was almost
  314 in 1998 dollars.

8
Rhode and Stumpf (AER, 2003)

• Data on a sample of US municipalities
  (1870-1990), all Boston-area municipalities
  (1870-1990) and all US counties (1850-1990) show
  that municipal per capita taxes, school district
  per capita taxes, total spending, protection
  spending etc. incomes and racial composition are
  now less differentiate across local communities
  than in the past.

9
Rhode and Stumpf (AER, 2003)

• They explain their result with the following
  reasons
• growing federal role in providing public services
• zoning policies are more and more popular after
  the war, avoiding the poor-chasing-the-rich
  phenomenon
• growing local government competition (to have the
  same category of residents the richer !)

10
Our explanation

• In our paper we test a fourth reason of the
  phenomenon observed. We show that the Tiebout
  model fails because residents are different, not
  only in their preferences on local public goods,
  but also because voting and moving are not
  equally shared.

11
The experimental desing

• 3 sessions
• 10 rounds for each session
• Initially 15 undergraduate students randomly
  allocated in 5 communities
• playing cards randomly assigned to each student
  for the whole experiment
• the 4 different suits of cards (hearts, diamonds,
  clubs, spades) represent types of public goods
• the number over the cards determine the
  preferences over public goods
• Each community is allowed to provide only one
  public good

12
The experimental design

• The individual welfare related to the public good
  provision is the difference between individual
  benefit and individual cost.
• For each student, the benefit of the public good
  provided in his/her community, is the minimum
  between the provision level and his/her sum of
  cards of that type.
• The community production cost is equal to the
  quantity produced (we assume constant and unitary
  costs). The total cost is shared equally among
  the members of the community, i.e. individual
  cost is equal to the community cost divided by
  the members.

13
The experimental design

• The dictator design

• The democracy design

• In each community there is a person ( the
  dictator) who decides the type of public good and
  the amount to produce
• The others can only move. They decide knowing
  the good produced by the proper and the other
  communities

• Each community decides in each of the 10 rounds
  of the session, both the type of public good and
  the amount to produce
• The decision is taken by the majority of votes

14
Total welfare
15
Cost per capita (in average)
16
Quantity produced (in average)
17
Preliminary conclusions

• total welfare is always higher in the democracy
  design, i.e. voting by feet increases efficiency
  if voting by ballot is equally shared
• in democracy cost per capita increases in the
  first part of the session, but it decreases after
  the 5th-6th rounds at the end of all the three
  sessions, the average cost per capita in the
  dictator design is higher
• the comparison between the quantity produced in
  the two designs doesnt show any trend

18
More in deep
19
More in deep

• The number of individual moving is positively
  related with the increase in individual welfare
  between the 1st and the 10th rounds in the
  democracy design in the dictator design in two
  of three session
• Under democracy, the welfare increase is
  negatively related with the number of rounds in
  which each individual met a person with a CS of
  the opposite sign, while in the dictator design
  this variable is not significant or positively
  related with the welfare. That means that voting
  by feet may enhance welfare, by reducing
  heterogeneity among members of the same community
  if all that people has the right to vote.

20
More in deep

• The previous result is confirmed by the
  coefficient of the dummy indicative of at least
  two people with a different CS
• In the dictator design the increase in welfare is
  positively related with a right decision on
  moving/non moving only if the community is
  correctly chosen, while in the democracy design
  there is a positive coefficient if the moving/non
  moving decision is rightly taken.

21
More in depth into the moving decision
22
More in depth into the moving decision

1. an individual loss in the previous round is
   always significant of the decision to move both
   in democracy and in the dictator design
2. in 4 over 6 sessions most of the moving decisions
   are taken in the first five rounds
3. in the democracy design the decision to move is
   influenced by the presence of different people in
   ones own community (session 1) or if the
   collective decision on the type and quantity of
   good is different from ones own vote (session 2)

23
Main conclusions

• Total welfare is always higher in democracy, i.e.
  voting by feet increases efficiency if voting by
  ballot is equally shared among residents
• under democracy increasing in individual welfare
  is positively related with the number of moving
  decisions taken
• the democracy process succeed in decreasing
  average cost per capita for public goods better
  than in the dictator framework

24
Main conclusions

• In democracy, the welfare increase is negatively
  related with the heterogeneity of people in each
  community, while in the dictator design this
  variable is not significant
• In the dictator design the increase in welfare is
  positively related with a right decision on
  moving/non moving only if the community is
  correctly chosen, while in the democracy design
  there is a positive coefficient if the moving/non
  moving decision is rightly taken.