Lecture 6' Meritocracy and Yardstick competition'

1
Lecture 6. Meritocracy and Yardstick competition.
2
Introduction.

• While fiscal decentralization gives instruments
  to pursue growth, the incentives work well if
  local authorities can get rents from higher
  growth via expanded revenues.
• However, better growth performance may also lead
  to faster promotion within the government
  bureaucracy.
• Meritocracy has always been an important
  component of Chinese administration (with
  fluctuations).

3
Yardstick competition

• Meritocracy makes use of yardstick competition
  (Lazear and Rosen, 1981 Holmström, 1982
  Nalebuff and Stiglitz, 1983)
• Yardstick competititon used a lot in China.
  Different provincial governments were ranked
  according to different measures of performance
  and this played a role for promotion of
  bureaucrats. Similar principle used for airline
  companies and divisions of different agencies.
• Maskin et al. (1999) show that yardstick
  competition works well in a M-form organizational
  structure (regional organization) compared to
  U-form (functional organization).

4
Yardstick competition

• Assume two regions A and B and two industries 1
  and 2. The performance of region r has the
  following form
• Rr er er
• er denotes unobservable effort by bureaucrat in
  region r
• er denotes shock. eA and eB are jointly normally
  distributed with E(er)0, variances s2A, s2B and
  covariance sAB.

5
Yardstick competition

• Similarly, define for industry
• Ri ei ei
• e1 and e2 are jointly normally distributed with
  E(ei)0, variances s21, s22 and covariance s12.
• The utility of a regional bureaucrat is
  Uw(RA,RB)-g(er), g gt 0, g gt 0, g gt 0
• Define similarly the utility of a bureaucrat
  heading a functional ministryUw(R1,R2)-g(ei)

6
Yardstick competition

• If Var(eA r eB) lt Var(e1 r e2) , then managerA
  can be given better incentives than manager 1.
• Managers A and B can be given better incentives
  than 1 and 2 if the following holdmin Var(eA r
  eB), Var(eB r eA) lt min Var(e1 r e2), Var(e2 r
  e1)max Var(eA r eB), Var(eB r eA) lt max
  Var(e1 r e2), Var(e2 r e1)
• Intuition is the same as moral hazard. Better
  incentives can be given if less noise (here
  conditional noise) since insurance motive less
  important.

7
Yardstick competition.

• Proposition more easily derived with specific
  functional form.
• Assume Uwr(RA,RB) -e exp- rwr(RA,RB)
• Assume linear incentive schemes in regions and
  industries
• wA(RA,RB)aA bARA cARB
• wB(RA,RB)aB bBRA cBRB
• w1(R1,R2)a1 b1R1 c1R2
• w2(R1,R2)a2 b2R1 c2R2

8
Using certainty equivalent expression for U(wr)
E(wr)-1/2 rvar(wr) where r is constant
absolute risk aversion derived from functional
form of U Uwr(RA,RB)-g(er) breAcreB-g(er)-1
/2 r(br2sA2cr2sB22brcrsAB) We can derive
similar expression for i. Assume the principal
is risk neutral and maximizes payoff of Managers
A and B net of incentive payments E(RA)
E(RB)-wA(RA,RB)-wB(RA,RB) (1-bA-bB)eA
(1-cA-cB)eB-aA-aB
9

• An efficient contract maximizes the joint payoff
  of principal and managers
• (er,br,cr)chosen to Max eA-g(eA)-1/2r(bA2sA2cA2s
  B22bAcAsAB) eB-g(eB)-1/2r(bB2sA2cB2sB22bBcBsA
  B)
• If effort were observable, first best would be
  g(er) 1 but it is not observable.
• Managers will maximize their utility, yielding
  bAg(eA), cBg(eB)
• Replacing bA in the objective function, we get
  the following f.o.c. for eA and cA
• 1-g(eA)rg(eA)sA2 cA2sAB g(eA)
• cA -g(eA) sAB/sB2
• Plugging the expression for cA in the f.o.c. for
  eA, we get
• 1-g(eA)rg(eA) g(eA)sA2 -sAB/sB2
• rg(eA) g(eA)Var(eA / eB) (since eA and eB
  are jointly normally distributed)

10

• Denoting VarAB Var(eA / eB) and
  differenciating expression for f.o.c. w.r.t. eA,
  we get

A higher conditional variance will thus lower
incentives for effort. Yardstick competition
works better when lower conditional variance.
11
Yardstick competition

• Maskin et al. (1999) give evidence that
  conditional variances of regional shocks are
  smaller than conditional variances of industrial
  shocks. Moreover, yardstick competition is
  effective for promotion.
• Sample of 520 SOEs from 1986 to 1991. Industrial
  and regional partition.
• Estimate industry-specific and region-specific
  shocks based on log-linear Cobb-Douglas
  production function.
• Compare conditional variance under regional and
  under industrial partition. Find on the basis of
  pairwise comparison that means of conditional
  variances higher on basis of industry rather than
  on basis of region.

12
(No Transcript)
13
Yardstick competition

• Look at provincial representation in CCP Central
  Committee normalized by provincial population in
  1987 compared to 1977 (before reforms started).
• Find that there was a positive correlation
  between growth ranking and political
  representation.

14
(No Transcript)
15
Yardstick competition

• Hongbin Li and Li-An Zhou (2005) look at better
  measure effect of individual growth performance
  of provincial leader on speed of promotion but
  also demotion.
• Promotion within the government administration
  seen as very desirable and also government career
  highly desirable among elite.
• China has 31 provincial units in Central
  Committee representation 22 provinces, 4 cities
  (Beijing, Shanghai, Tianjing, Chongqing and 5
  autonomous units.
• Since 1978, evaluation of provincial cadres not
  only based on political loyalty but also on
  expertise, education and youth.
• Since 1983, retirement of provincial leaders at
  65 if not promoted, but not strictly enforced.

16
Yardstick competition.

• Provincial party secretaries can become members
  of the State Council, vice-premier or premier,
  member of Politburo or its standing committee.
• Provincial governor one step below provincial
  party secretary.
• People can also be promoted to honorary
  positions without power.
• At that level, positions in private sector not as
  attractive.

17
Yardstick competition

• Ordered probit model.
• Turnover variable y takes values 0 (retirement),
  1 (same), 2 promotion. Assume evaluation y
  unobserved, y xbe where b are coefficients, x
  economic performance measure and a1, a2 are
  cutoff points for evaluation. Call F cdf of
  normal distribution.

18
Yardstick competition

• Data on 254 provincial leaders between 1979 and
  1995. Information on career moves, age,
  education, past work experience. More than 70 of
  leaders had turnover during that period.
• Average annual turnover rate of 20 (like in US
  corporations).

19
(No Transcript)
20

• Estimates show that when annual growth rate
  increases by one standard deviation (0.06) from
  the mean (.10), the probability of promotion
  increases by 15 of the average probability of
  promotion (.75).
• Regressions robust in
• Change of weights in average growth rates,
• Non linearity in tenure effect
• Measurement error in age 65 (year of birth
  available, not month) using age 64 and 66.
• Differenciating between normal and forced
  retirement by only looking at leaders below 65 or
  sample only before 1983

21
Decentralization and yardstick competition
compared.

• The sole incentive effect of decentralization is
  that a higher tax base can be in the interest of
  local government. Implicit argument that local
  leaders get direct and indirect rents from more
  tax revenues associated to more economic growth.
• It seems the effect of yardstick competition is
  very strong as long as a career in the
  bureaucracy is seen as very attractive for the
  most brilliant people.
• Decentralization gives local government the
  policy instruments to respond well to the
  yardstick competition. So their effects must be
  seen jointly.

22
Implications.

• This analysis gives a view of how the specific
  institutions in China contributed to economic
  growth.
• It also has implications for further reform.
• Picture is one where growth and reform happen not
  by simple shrinking of the role of government,
  leaving room for free markets. Image of a system
  where the bureaucracy and the organization of
  government are geared towards growth objectives
  and market promotion (bird in a very large cage,
  paraphrasing Chen Yun).
• Meritocracy and decentralization act as
  substitutes for rule of law and separation of
  powers.
• The meritocracy that was able to achieve
  spectacular growth should be able to achieve
  other objectives such as climate change if a
  strong political will existed.

23
Implications

• System with mainly accountability from above.
  Accountability from below (local elections) only
  used as instrument of personnel control. Weak
  accountability at the highest level necessitated
  mandatory retirement of leaders, good and bad.
  Substitute for absence of democracy but
  represents a specific solution to the succession
  problem .
• Breaking the meritocracy and accountability from
  above (centralized personnel control) are likely
  to break the engine of growth in the absence of
  (unlikely) compensating changes.
• The system has however scope for gradual reforms
  towards separation of party and state, getting
  gradually rid of the dual power structure.
  Parallel careers possible within the
  meritocratic system. Also scope for more media
  freedom and judicial independence.

24
Dangers in the long run

• Strong development of large private firms may
  reduce the attraction of a career in the
  bureaucracy (see however France).
• More leniency towards corruption over time is a
  nearly inevitable development. The meritocratic
  system will gradually stop functioning once
  corruption becomes too strong and the lack of
  accountability at the highest level cannot over
  time create enough resistance to counter the
  expansion of corruption.
• This is not a short term prospect but Chinese
  history shows that brilliant dynasties could be
  brought down by corruption.
• A bold counterweighing reform could be
  election of party chief like within party
  primary.
• Copying Singapore and Hong Kong not so easy.
  Meritocracy is therefore key.
• External environment will also be a danger (North
  East Asia, South East Asia, Taiwan, US).