Political Economy of Development

  Stockholm Doctoral Course Program in Economics Development Economics II

  Lecture 5

Political Economy of Development

  Masayuki Kudamatsu

  IIES, Stockholm University Big question today How does politics

affect development?

  More specific questions today

  

1. Why do welfare-enhancing reforms

  often get stalled?

  2. Is democracy good for

  development?

  3. What makes democracy work?

  4. What makes autocracy work? reforms often get stalled?

  Structural adjustment reforms in

  • 1980s Streamlining procedures of setting
  • up a firm (cf. Djankov et al. 2003) Formalizing property rights
  • >Anti-corruption meas
Basic idea for why welfare-enhancing reform gets stalled Winners may propose

  • compensation for losers

  As welfare (total surplus) increases,

  • winners still better off

  Once reform enacted, winners lose

  • incentive for compensation

  ⇒ Winners’ promise not credible Reform gets stalled because...

  • Losers constitute a majority
  • Even if losers are a minority:

  

(1) Small group size solves collective

action problem to lobby govt (Olson 1965)

  

(2) Individuals uncertain whether to win

or lose (Fernandez & Rodrik 1991)

  • Uncertainty over whether losers will have political power

  (Jain & Mukand 2003)

  • (Majumdar

  These papers look at only one

  • policy reform In reality, there are many policies to
  • reform, whose feasibility may depend on each other Caselli and Gennaioli (2008) tackle
  • this issue

  (2008)

  Research questions: Deregulation of entry

  • Improving contract enforcement in
  • credit market (financial reform) Both increase entry and thus
  • >welfare But politically feasi
Why original & important?

  Entry deregulation & financial

  • reform:

  Perhaps two of the most important

  • reforms for development

  Aghion et al. (2008) and Burgess &

  • Pande (2005) for impact evaluations

  Only analyzed separately before

  • Provide a 1st example of
  • endogenous (ie. market-driven) compensation of losers from a
Model: Demography &

Endowment

  A continuum (of measure 1) of

  • agents Each agent: endowed w/ 1 unit of
  • labor & managerial talent θ Fraction λ: θ = 1 (talented)
  • − λ: θ = θ(≤ 1)

  Fraction 1

  • (untalented)
Model: Demography & Endowment (cont.)

  Fraction η: also endowed w/ a firm

  • (incumbent)

  e.g. Family-owned firm

  Fraction 1 − η: (outsiders)

  • None own wealth
  • ⇒ Financial reform matters
Model: Technology

  If agent w/ talent θ owns a firm and

  • hires l labor (incl. his own):

  1 −α

  y = θl α: degree of decreasing return to

  • scale

  See sec. V.B. for constant return to

  • scale
Model: Market for control

  Outsiders can buy a firm from an

  • incumbent by paying price p Each agent can run at most 1 firm
  • ⇒ Talented incumbents cannot buy a

  firm from untalented incumbents

Main results would still hold if

  • otherwise (p. 1227)

  No market for managers

  Incumbent cannot hire talented agent as manager of their firm Model: Policy variable 1

  ε(≥ k ): cost of setting up a firm for outsiders

  • k (> 0): technologically determined fixed cost
  • If k = 0, no role of market for control (pp. 1206-7)
  • ε − k : degree of entry regulation
  • Bureaucratic setup costs (incl. bribes) cf. Djankov et al. 2003
Model: Policy variable 2

  φ: fraction of borrower’s profit that lenders can recover in case of default Quality of financial legislation /

  • contract enforcement

  As no agents has assets, they need • to borrow for setting up / buying a firm ⇒ φ ↑: financial reform

  Credit supply: perfectly elastic w/

Model: payoffs

  Workers: w

  • Incumbents who don’t sell: π + w
  • (profit plus his own wage) Incumbents who sell: p + w
  • Outsiders who set up a firm:
  • >π + w − ε Outsiders who buy a firm: π + w &minu
Model: Timing of events

  1. Outsiders: decide whether to

  become a firm owner (either by paying setup cost ε or paying p to an incumbent) by borrowing money

  2. Firm owners: hire workers by

  paying wage w

  3. Production takes place

  

4. Borrowers: decide whether or not to Assumptions 1. η < λ < 1

  

To focus on whether talented

  • outsiders can set up a firm

  α−1

  = αλ

  2. k ⇒ At first best, no untalented agent runs a firm

  Main results would still hold if α−1

  • αλ > k > 0 (p. 1206)

Analysis: How to proceed

  Characterize equilibrium choice of

  • outsiders by:

  f : total # of firms (entrepreneurship) s

  : share of firms run by talented agents (meritocracy )

  = η and s = λ Initially we have f

  • First best: f = λ and s = 1 (Lemma
  • 1)

  

Analysis: How to proceed (cont.)

  

1. Derive payoffs as functions of (f , s)

  2. Analyze how (f , s) change with ε

  φ and

  3. Identify winners and losers from

  reforms ( ε ↓ and φ ↑)

  ⇒ Assess political feasibility of

  reforms (by assuming the more winners, the more feasible) Analysis 1: equilibrium payoffs

  given w

  clearing c. Obtain π as a function of exog. variables & (f , s) Analysis 1a: labor market

  • 1 −α

  Firm owner solves:

  max θl − wl

  l

  By FOC, the optimal labor demand

  • by firm owner of talent θ is: � (1 − α)θ �
  • 1

      ∗ α

      l (θ) = w

      

    Analysis 1b: labor market (cont.)

    • ∗ ∗

      Labor market clearing:

      = fsl (1) + f (1 − s)l (θ)

      1 � 1 − α � 1 α α 1 = f (s + (1 − s)θ ) w

      ⇒ Equilibrium wage: 1 α α α

      w (f , s) = (1 − α)f [s + (1 − s)θ ] Analysis 1c: profit function

      To obtain profit function

      1 ∗ −α ∗

      π(θ) = θl (θ) − w(f , s)l (θ)

      ∗

      we also need expression for l (θ) optimal labor input

      l

      ∗

      (θ) = [(1 − α)θ] w (f , s) 1 α

      = [(1 − α)θ] 1 α

      [(1 − α)f

      

    α

      [s + (1 − s)θ 1 α ]

      α

      ] 1 α =

      θ 1 α f [s + (1 − s)θ 1 α ] Analysis 1c: profit function (cont.)

      π(θ) = θl

      ∗

      (θ)

      1 −α

      − w(f , s)l

      ∗

      (θ) =

      θ ∗ θ 1 −α α [f [s + (1 − s)θ 1 α ]]

      1 −α

      − (1 − α)f

      

    α

      [s + (1 − s)θ 1 α ]

      α

      θ 1 α f [s + (1 − s)θ 1 α ] =

      αθ 1 α

      

    Analysis 1c: profit function (cont.)

      For talented firm owners

      H α−1 α−1

      π (f , s) ≡ αf [s + (1 − s) g ] For untalented firm owners

      L H

      π (f , s) ≡ π (f , s)

      

    g

      where α 1

      

    g ≡ θ (≤ 1)

      < 0:

      H

      ∂π L (f ,s) ∂s

      < 0,

      (f ,s) ∂s

      < 0: ⇐ More firms push up wage

      ∂π L (f ,s) ∂f

      < 0,

      (f ,s) ∂f

      (f , s)

      (f , s) ≡ gπ

      Analysis 1c: profit function (cont.)

      L

      π

      α−1

      [s + (1 − s)g]

      α−1

      (f , s) ≡ αf

      H

      π

    • ∂π H
    • ∂π H
    Talented incumbent’s payoff (See Proof of Proposition 3)

      As f or s goes up:

      H H α (1−α) 1−α

    • α

      π (f , s) + w(f , s) decreases

      π (f , s) = [ ]

    w (f ,s)

    • As a function of w , minimized at
    • w = 1 − α α {f ,s} max w (f , s) = (1 − α)λ < 1 − α

      Talented incumbents prefer lower f , s if they don’t sell Untalented incumbent’s payoff (See Proof of Proposition 3)

      As f or s goes up:

      L α π (f , s) + w(f , s) is...

    • α

      Minimized at w (f , s) = (1 − α)g

    • Max w (f , s) = (1 − α)λ could be
    • larger than this λ is large Which happens if
    • We’ll see in such a case no one will
    • set up a firm and thus wage does not reach its maximum
    Analysis 2: Derive f and s

    • a. Set up a firm by paying ε

      An outsider’s decision to

      paying p

      Either way, an outsider needs to

    • borrow How much can they bor
    How much to borrow

      Let L be amount of loan, π profit Repay: π − L

    • Default: π − φπ
    • ⇒ ≤ φπ

      Repay iff L Lenders’ profit

    • − L = 0 if L ≤ φπ

      L
    • φπ − L < 0 if L > φπ •

      ⇒ Maximum loan: φπ See fn 7 for what if wage can be • Analysis 2a: entry

    • want to set up a firm if π(f , s) ≥ ε

      Outsiders:

    • can set up a firm if φπ(f , s) ≥ ε

      If ⇒ φ < 1, credit constraint (want to own a firm, but cannot borrow for it)

      Deregulation & financial reform both

    • encourage entry
    Analysis 2a: entry (cont.)

      Lemma 2: No untalented outsiders set up a firm under assumption 2 (ie.

      α−1

      k = αλ ) This simplifies analysis below

    • All the results would still hold
    • otherwise
    Proof of Lemma 2 H L

      π > π

    • ⇒ If some untalented enter, all

      talented should have already entered

      L

      π : largest when 1st untalented

    • enter after all talented entered (w/ no firm sale)

      L ⇒ If such π is smaller than ε/φ, Proof of Lemma 2 (cont.)

      λ

      L

      φπ (η + λ(1 − η), ) η + λ(1 − η)

      λ

      H

      ≤ π (η + λ(1 − η), ) η + λ(1 − η) � � α−1

      λ + η(1 − λ)g

      

    α−1

      = α(η + λ(1 − η)) η + λ(1 − η)

      α−1

      = α[λ + η(1 − λ)g]

      α−1

      < αλ = k (by Assumption 2) Proposition 1

      Assume no market for control. The unique equilibrium is:

      H

      (η, λ) ≥ ε

    • Entry iff φπ
    • At least some talented outsiders set up new firms

      φπ

      

    H

      (η, λ) < ε

    • No Entry iff
    • No outsider sets up a firm

    Proposition 1 (cont.)

      φπ

      H

      (η, λ): Amount of money talented agents can borrow if no one sets up a firm in the equilibrium

    • If larger than the set-up cost ε, talented agents enter the market ( ⇒ f ↑, s ↑) until π

      H

      (f , s) goes down to ε. Corollary 1

      Without market for control, ε ↓ or φ ↑ ⇒ Both f & s: ↑

    • Untalented outsiders won’t set up a
    • firm (lemma 2)

      ⇒ # of talented outsiders setting up a firm ↑ ⇒ f ↑ & s ↑

    • L

      1st best cannot be achieved

    • L

      1st best: no untalented owns a firm

      

    π : lowest at ε = k , φ = 1

    • But π > 0, ∀f , s (by α > 0)

    Analysis 2b: market for control L

    • ≥ π (f , s) want to sell if p

      Untalented incumbents:

    • H

      Talented outsiders:

    • want to buy if π (f , s) ≥ p H • can buy if p ≤ φπ (f , s)

      ⇒ Transaction takes place if L H

      π (f , s) ≤ p ≤ φπ (f , s) ⇐⇒ φ ≥ g Talented incumbents won’t sell

    Analysis 2b: market for control (cont.)

    • Financial reform encourages the sale of a firm from untalented to talented
    • Consistent w/ evidence by Rossi & Volpin (2004)
    • p: determined by bargaining power btw. seller & buyer
    • For much of analysis, the exact p does not matter (fn. 17)
    Proposition 2

      If market for control exists, the unique equilibrium is

    • No Sales if φ < g
    • NO untalented incumbents sell

      (η, λ) < ε

    • No Entry if φπ H
    • Entry otherwise (by Proposition 1)

      φ ≥ g

    • All Sell if
    • ALL untalented incumbents sell

      (η, 1) < ε

    • No Entry if

      φπ

      H
    • Entry otherwise

      φ ε g

      1 k φ

      = ε/π

      H

      (η, 1) φ

      = ε/π

      H

      (η, λ)

      Entry No sales Entry All sell

      No entry All sell

    No entry

    No sales Proposition 2 (cont.)

      Why all sell, not some sell? One more untalented incumbent

    • sells

      

    ⇒ Labor demand ↑ (talented demand

      more)

      H ⇒ ↑, π ↓

      Wage But this does not affect the

    • condition for sale as long as φ ≥ g:
    Corollary 2

      If φ does not cross the φ = g line, ε ↓ or φ ↑ ⇒ Both f & s: ↑

    • Transition from φ < g to φ ≥ g: s jumps up to 1
    • This change would be smooth if
    • continuous θ (p.1233)

      H ⇒ π (f , s) goes down

    ⇒ Fewer talented agents set up a firm

      

    Consistent w/ empirical findings

    (& new empirical implications)

    • φ ↑ or ε ↓ ⇒ s ↑
    • Djankov & Murrell 2002
    • ε ↑⇒ f ↓
    • Klapper, Laeven, & Rajan 2004
    • Fisman & Sarria-Allende 2004
    • >φ ↑ may redu
    • ?
    Analysis 3: Political feasibility

      Who are the winners and losers

    • from the two reforms? Assumption 3:
    • In the status quo ( ε , φ ), the
    • equilibrium is No Entry No Sale
    Winners & Losers from ε ↓ / φ ↑

    • Outsiders: always win (Prop 4
    • • Their payoff: at least w (f , s)

    • Talented incumbents: always lose (Prop 4(ii))

      more than they increase wage

    • Both φ ↑ & ε ↓ erode π H
    • Maximum wage (w

      (λ, 1)) at 1st-best is not large enough to compensate profit loss

      Can untalented incumbents be winners

      Proposition 4 (iii)

      Untalented incumbents’ payoff: Always ↓ by deregulation

    • For φ < g, wage cannot be large
    • enough to compensate profit loss

      For φ ≥ g, price of firm ↓

    • H
    • >

      where ε ∈ [k , π ˜ (η, 1))

      ↑ by financial reform if ε > ˜ ε Why untalented incumbents’ payoff goes up with φ ↑?

      Once All Sell equilibrium achieved, their

      

    H

      payoff: w (f , 1) + φπ (f , 1) No Entry equilibrium: wage

    • constant at w (η, 1) while the firm price increases Entry equilibrium: wage goes up
    • with φ via f ↑ while the firm price is
    But this is conditional on large ε For small ε : financial reform

    • induces entry before All Sell equilibrium achieved In this case, untalented incumbents’
    • payoff goes down with φ ↑ due to profit erosion
    Intuition

      Untalented incumbents:

    • endogenously compensated by market for control

      Welfare gain by financial reform: firms

    • run by more talented By selling firms to talented,
    • untalented incumbents share the benefit of financial reform
    Implications

      Untalented incumbents: more likely

    • to support financial reform w/ higher ε ⇐ Because higher entry barrier kills the profit-reducing effect of financial reform

      ⇒ To enact financial reform, raising ε

      is a good idea! Optimal sequence of reform can then be

      (section IV.C)

      First, financial reform without entry

    • deregulation

      so untalented incumbents support the

    • reform
    • Now untalented incumbents have

      Then entry deregulation

    • become workers supporting the reform
    A very nice policy implication that

    • incorporates political-economy But perhaps naive to assume that
    • political feasibility goes up with # of supports What’s the role of political
    • institutions?
    development?

      Democracy: perhaps most

    • important political institutions Development assistance
    • >practitioners: often advocate democracy as a means to good governance # of democracies: ↑ recently (Fi

      Figure 1 (Besley 2006, p. 5)

    What is democracy, BTW?

      Definition in this lecture: A political system in which

    • policy-makers are chosen in a competitive popular election

      

    Digression: Democracy datasets

    • treats different aspects of democracy as

      Polity IV or Freedom House

      substitutes

    • (recently updated by Cheibub et al. 2010)

      Przeworski et al. (2000)

      treats different aspects of democracy as complements cf. Munck and Verkuilen (2002) for

      2-1. Evidence

      Large social science literature on

    • which political regime can achieve higher growth cf. Przeworski & Limongi (1993) for an early literature survey Acemoglu et al. (2008) provide
    • three stylized facts

      2-1 Evidence (cont.)

      democratic (Fig.2)

      Figure 2 (Acemoglu et al. 2008, p. 809)

      2-1 Evidence (cont.)

      democratic (Fig.2) This could be either due to

    • 1. Democracy ⇒ Development

      2. Development ⇒ Democracy 3. 3rd Factor ⇒ Development & Democracy

      2-1 Evidence (cont.)

      income and changes in democracy in 20th century

      1975-1990 (Fig.3)

    • 1900-2000 (Fig.4)
    Figure 3, 1970-1995 (Acemoglu et al. 2008, p. 811)

      Figure 4, 1900-2000 (Acemoglu et al. 2008, p. 812)

      2-1 Evidence (cont.)

      changes in income and changes in democracy in the very long run: 1500-2000 (Fig.5) Figure 5, 1500-2000 (Acemoglu et al. 2008, p. 832)

      2-1 Evidence (cont.)

      run: 1500-2000 (Fig.5) This could be either due to

    • 1. Very long-run effect of democracy on

      development

      on democracy

    3. 3rd factor driving both changes

      2-1 Evidence (cont.)

    • Positive cross-sectionally

      Correlation with health

    • Zero for changes after WWII (Besley
    • & Kudamatsu 2006, Ross 2006)
    • Positive cross-sectionally • Zero for changes after WWII, if years • of schooling used (Acemoglu et al. 2005) Positive for changes after WWII or

      Correlation with education

    • 1865, if enrollment ratio used

      2-1 Evidence (cont.)

      Very difficult to identify causal effect of democracy Democracy: endogenous

    • education, emergence of middle
    • class, etc. (Lipset 1959)

      transitory negative income shock
    • (Brucker & Ciccone 2011)

      Kudamatsu (2012): use

    • cross-counry micro panel data on survival of a woman’s babies to

      2-2 Beyond simple comparison

    4 reasons for why simple comparison is fruitless

    a. Representation theory of politics

      ⇒ Heterogenous treatment effect

      b. Accountability theory of politics

    ⇒ Complementary institutions to make

    democracy work c. One more stylized fact

      ⇒ Autocracy: highly heterogenous Digression: 2 views on politics

    • Conflict of interest

      Representation

    • among citizens

      cf. Persson & Tabellini (2000, ch. 3 & 5)

    • Conflict of interest

      Accountability

    • between govt & citizens

      

    cf. Persson & Tabellini (2000, ch. 4),

    Besley (2006, ch. 3)

      

    2-2a Representation perspective

    Acemoglu (2008)’s model nicely summarizes this perspective for the impact on growth

    • Firm managers and Workers • Firm managers prefer
    • Low tax rate
    • Entry barrier to keep wages low
    • Workers prefer
    • >High tax rate for redistribution
    Standard growth theory suggests: ↑ ⇒ Investment ↓ • Tax ⇒ Short-run growth ↓ Entry barrier ↑ ⇒ Innovation ↓ • ⇒ Long-run growth ↓

    • Tax ↑ ⇒ Investment ↓ ⇒ Short-run growth ↓ ↑ ⇒ Innovation ↓ • Entry barrier ⇒ Long-run growth ↓

      Autocracy: Firm managers decide policies •

      1. Tax: low ⇒ Short-run growth: high

      2. Entry barrier: high ⇒ Long-run growth: low

    • Tax ↑ ⇒ Investment ↓ ⇒ Short-run growth ↓ ↑ ⇒ Innovation ↓ • Entry barrier ⇒ Long-run growth ↓

      Democracy: Workers decide policies •

      1. Tax: high

    ⇒ Short-run growth: low

      2. Entry barrier: low

    ⇒ Long-run growth: high Impact of democracy on growth has

      two channels

      Investment

    • Innovation
    • Democracy: good for growth in
    • >sectors where innovation is more important than capital accumulation ⇒ Aghion et al. (2008) provide evidence consistent with this Theory uncovers heteroge

      2-2b Accountability perspective

      Democracy: regularized leadership

    • contest Citizens can replace incompetent
    • >leaders w/ (potentially) better ones ⇒ Incentives for leaders to behave well So democracy is always g

      This argument makes (at least) 3

      implicit assumptions

      1. Voters know what leaders did ⇒ w/o free media, this is unlikely

      2. Voters behave as one agent

    ⇒ w/ conflict of interest among

    voters, this is unlikely

      3. Challenger is better than bad-behaving incumbent

      ⇒ w/o complementary institutions, democracy won’t work

      2-2c One more stylized fact

      More volatile economic growth in autocracy than in democracy Performance varies a lot more for

    • autocracy (Fig.7)
    Figure 7 (Besley & Kudamatsu 2008, p. 453)

      2-2c One more stylized fact

      More volatile economic growth in autocracy than in democracy Performance varies a lot more for

    • autocracy (Fig.7) Jones & Olken (2005): Natural
    • death of a leader ⇒ Changes in growth in autocracy, but not in democracy

      ⇒ Autocracy appears to be VERY heterogeneous ⇒ Not an ideal control group to estimate the impact of democracy

      2-2d de facto political power

      Acemoglu & Robinson (2008): Rich

    • may intensify de facto political power to nullify de jure political power of poor ⇒ Regime change may not be followed by policy change

    More fruitful questions to ask:

    • What makes democracy work?
    • What makes autocracy work?
    work?

    • Competitive election per se may not bring benefits to citizens
    • What institutions are complementary to elections?
    • Literature has so far identified:
    • Free media (or information provision)
    • Reservation of political office

    • Enfranchisement

      3-1 Free Media

      Political agency model: citizens

    • punish incumbents who chose wrong policies Unless citizens observe what policy
    • is chosen, however, they won’t be able to punish Role of media crucial to make
    • politicians accountable in

      3-1 Free Media (cont.)

      Quite a few papers by now

    • empirically show this complementary role of media (or information provision in general)

      Besley & Burgess (2002) for

    • newspaper circulation in India

      Ferraz & Finan (2008) for corruption
    • audit in Brazil Enikolopov, Petrova, & Zhuravskaya
    • (2011) for independent media in Russia

      3-2 Political Reservation

      Citizen-candidate model:

    • policy-makers cannot commit to electoral platforms Disadvantaged groups (women,
    • minority ethnic groups, etc.) tend to be underrepresented in democratic politics

      ⇒ Their preferred policies won’t be

      3-2 Political Reservation (cont.)

      India has attempted to correct this

    • by reserving political office to disadvantaged groups

      In 1992, 1/3 of rural municipalities

    • randomly chosen in a state: only women can become municipality prime minister (Pradhan) By Constitution, some seats in state • legislature reserved for scheduled castes (SC) and scheduled tribes (ST), and seat share should equal to

      3-2 Political Reservation (cont.)

      Chattopadhyay & Duflo (2004)

    • empirically show that reservation for women increases adoption of policies cared by women Pande (2003) empirically shows
    • reservation for SC and ST increase redistribution to them

      3-3 Franchise extension

      Median voter theorem: suffrage

    • extension leads to policy change Before WWII, suffrage often first
    • given to men, then to women Women more concerned on child
    • health than men Miller (2008) shows enfranchising
    • women reduced child mortality in

      3-3 Franchise extension (cont.)

      Today, all democratic countries

    • have universal suffrage In reality, poor people are effectively
    • barred from voting

      

    Illiterate people cannot write a

    • candidate’s name on the ballot

      Poll taxes, literacy test in US South
    • until 1960s (cf. Besley, Persson, & Sturm 2010)
    In this context, electronic voting can

    • be “de facto” enfranchisement Fujiwara (2010) estimates its
    • impact in Brazil

      

    The following pictures are taken from Figures 1 and 2 of

    Fujiwara (2010)

      Before

      After

      If you enter correct candidate # ...

      If you enter wrong candidate # ...

    • Using RD design, Fujiwara (2010) finds that due to electronic vot># of invalid votes ↓
    • public health care spending ↑
    • LBW for uneducated mothers ↓

      3-4 Term limit

      Some democracies impose term

    • limit

      Often due to the nasty dictatorship in

    • the past (e.g. Latin America)

      Ferraz & Finan (2011) show those

    • Brazilian mayors who cannot run for re-election due to term limit are more corrupt

      3-5 Secret Ballot

      In agrarian society, landlords supply

    • the votes of their workers in exchange for money, favors, or policies from politicians

      Britain, Germany, France in 19th

    • century Chile until 1958
    • No secret ballot in these cases
    • Ballots have to be obtained from
    • >candidates Voter’s decision can be induced from
    Secret ballot makes tenants’ voting

    • choice unobservable to landlords Landlords then cannot fire tenants
    • who voted against their will Baland & Robinson (2008) provide
    • such evidence in Chile: landlord party’s vote share ↓ after secret ballot introduced in 1958

      Some developing countries: still

    • autocracy

      China, Cuba, North Korea, many Gulf

    • states, etc.
    • Thailand, Fiji, Guinea, Honduras, • Mali, Guinea-Bissau, Egypt

      Military coups: still relevant

      See Besley & Kudamatsu (2007,

    • section 2) for a literature review Best paper in this literature so
    Stylized facts on autocrats

    • Extracting enormous rents from power
    • Personal wealth equivalent of

      country’s foreign debt
    • Extensive & very inefficient redistribution
    • Buy crops from farmers for way below world price
    • Bloated bureaucracy
    • Active support from sizable share of

    Research question

      What characteristics of autocracy &

    • economy explain these three stylized facts?
    Model: demographics

      A continuum of infinitely-lived

    • citizens of mass 1 Proportion π
    • A : group A; the rest

      group B

      

    Group identity: observable &

    • impossible to change (e.g. skin color)
    • t

      ∈ {A, B} in Leader of group S

      power (government) Model: technology & citizens’ actions a A

    • ω = 0) a A A t if engage in activity a (z

      Group A:

    • ω − θ if activity b (z = 1) • t b B
    • ω if engage in activity b (z = 0) • b B B t t

      ω − θ if activity a (z = 1)

      Group B:

    • i θ (i ∈ {A, B}): comparative adv.
    • e.g. tree crops (cocoa etc.), ethnic

      network (Banerjee & Munshi 2004, Model: leader S

    t

      ’s actions (policies)

      S t a t

      , τ

      S t b t

    ⇐ No state capacity to tax income

    • Tax on each activity: τ

      S t A t

      , η

      S t B t ⇐ Group identity observable

    • Patronage to each group: η
    Model: preference

    • S a A a t

      Group A citizens

      (1 − z )(ω − τ )

      

    t t

    S b A a A t

    • z (ω − θ − τ )

      t t S A t

    • R(η )
    • � �� � t

        R > 0, R < 0, R(0) = 0, R (0) > 1 • S A t

        

      ⇒ Social optimal η > 0, satisfying

      t

      • Group B citizens

        (ω

        S t B t

        )

        S t a t

        − τ

        B

        − θ

        b

        

      B

      t

        )

        S t b t

        − τ

        b

        )(ω

        B t

        (1 − z

        Model: preference (cont.)

      • z
      • R(η

        ) Model: preference (cont.)

      • t S a t

        Leader of group S

        

      A A A B

        τ [π (1 − z ) + (1 − π )z ]

        t t t S b t A A A B

      • τ [π z + (1 − π )(1 − z )]

        t t t S A S B

      A A

      t t

        − π η − (1 − π )η

        t t If ousted, leader’s payoff is 0 from t on

      • δ: discount factor for both group
      Model: political institutions

      • S t

        Group S t : decide whether to

        support leader S t (s = 1) or to

        t S t

        oust him (s = 0)

        t S t S t

      • t = S t w/ prob. γ

        If s = 1, S

      • 1 t S t

        S t

      • t t

        If s = 0, S = S w/ prob. γ

      • 1 t S S t t
        • Uncertainty of leadership succession

        γ > γ

      • in autocracy
      Model: Timing of events at t

        

      1 Leader of S chooses policy vector

      t

      S a S b S A S B

      t t t t

        P ≡ {τ , τ , η , η }

        t t t t t

        2 Group S decides whether to t S t

        support leader, s ∈ {0, 1}

        t

        3 Both groups decide whether to A B

        switch activity, z , z ∈ {0, 1}

        t t

        

      Model: Timing of events (cont.)

      S t

        = 1, P

        

      4(a) If s t implemented, payoffs

      t S t

        = S realize, S t t w/ prob. γ

      • 1 S t

        4(b) If s = 0, leader ousted, t

        P ≡ {0, 0, 0, 0} implemented,

        r payoffs realize, S = S w/ prob. t t

      • 1 S t

        γ Model: discussion on assumptions

        (1) No collective action problem (2) No repression

      (3) Group j �= S never supports leader

      t

        of S t (1) & (2): fine as leader would

      • otherwise steal more (3): Why not leader seeks sup
      Definition of equilibrium

        Focus on Markov Perfect Equilibrium (MPE)

        Strategies: contingent only on the

      • payoff-relevant state variable S &

        t

        prior actions within the same period

        Exclude history-dependent strategies

      • such as trigger strategy (see sec 3.1)

        ⇒ Below time subscript t dropped

        State transitions: as described in

        

      Digression: Equilibrium concepts

      for repeated games

        A repeated game has many SPEs ⇒ Economists focus on a subset of them that are most plausible

        1. Stationary SPEs

        

      2. SPEs on the Pareto frontier (& take

        a stance on relative bargaining power of players)

        ⇐ Baland & Robinson (2008) Definition of equilibrium (cont.)

        Pure-strategy MPE: set of the following 4 strategies that are best responses to each other

        

      A

      • B

        = P Leader of A: P

        Leader of B: P = P

      • Group A citizens:
      • A S A A

        σ (S, P ) = {s , z } Group B citizens:

      Analysis: value functions Assume S = A w.l.o.g.

      • Case of S = B can all be analyzed
      • symmetrically

        j

        V (S): value function for group j

      • citizens in state S

        j

        (S): value function for leader of j W

      • in state S A B

        (B) = 0, W (A) = 0 W

      Analysis: switching B

        Choice of z : no impact on

      • continuation value

        B b Ab b B Aa

        < ω

        ⇒ z = 1 iff ω − τ − θ − τ t A

      • A a Aa a A Ab

        Likewise, choice of z :

        < ω z = 1 iff ω − τ − θ − τ

        t Analysis: switching (cont.)

        Tax revenue from a citizen of group B

        Ab B Ab B Aa

      • Aa B Ab B Aa

        τ if z = 0 ⇐⇒ τ ≤ θ + τ

      • Ab B Aa

        τ if z = 1 ⇐⇒ τ > θ + τ

        ⇒ Maximized at τ = θ + τ

        Tax Revenue from a citizen of group B τ Aa 45 o B Aa Ab

        θ

      • τ τ
      Analysis: switching (cont.)

        Same is true for tax revenue from

      • group A

        Aa Ab ⇒ Leader sets (τ , τ ) so that

        A B

        z = z = 0

        ⇒ Upper limit on tax rates: Aa A Ab

        τ ≤ θ + τ

        Ab B Aa

        τ ≤ θ + τ

      • If s
        • R(η
        • δγ

        )V

        A

        )V

        A

        V A (A) + δ(1 − γ

        A

        a

        ω

        (B)

        A

        A

        V A (A) + δ(1 − γ

        A

        )

        AA

        Aa

        − τ

        a

        = 1 (support) ω

        Analysis: support for incumbent

      • If s = 0 (oust)
        • δγ

        (B)

        

      Analysis: support for incumbent

      (cont.)

        ⇒

        s = 1 iff

        Aa AA

        τ − R(η )

        A A A A

        ≤ δ(γ − γ )(V (A) − V (B))

        A

        ≡ Φ

        LHS: Net tax payment

      • (1 − π

        Ab

        (A) subject to

        Aa

        ≤ θ

        A

        

      Ab

        , τ

        ≤ θ

        W

        B

        Aa

        Aa

        − R(η

        AA

        ) ≤ Φ

        A

        A

        A

        Aa

        Analysis: leader’s problem

        max

        τ AaAbAAAB

        π

        A

        (τ

        − η

        )

        AA

        )

        A

        )(τ

        Ab

        − η

        AB

      • δγ
        • τ

      • τ
      • τ
        • τ
        Lemma 1 AB

        η

        1. = 0: no transfer to excluded

        group

        ⇐ Excluded group cannot affect leader’s survival

        Ab B Aa 2. τ = θ + τ : tax revenue from

        excluded group maximized

        Excluded group cannot affect leader’s • survival ⇒ Leader can tax them until they are

      indifferent btw. activities a & b Aa A Ab Lemma 1 (cont.) Aa A AA

        

      3. τ = Φ + R(η ): tax on leader’s

        group constrained

        ⇐ Leader’s group would kick leader out Aa if higher τ than this In equilibrium, leader is always

      • supported

        ⇒ Leader’s objective function:

      A A AA AA

        π max (Φ + R(η ) − η ) AA

        η A B A AA