Coments on Praja Trivedi, “Performance Contracts”
1. Very interesting presentation, learned a lot from reading the book Praja edited on MOU in India.
2. Basic premises: what get measured, got done.
3. Praja seems to be optimistic about the use of PCs in reforming bureaucracy and enterprises.
4. I’m more pessimistic about what PCs can accomplish:
a. For PCs to work, PCs must solve three problems well: information advantage of agents, incentive designs for both the principal and the agent, and commitment.
b. Information advantage of the agent would have serious consequences for the design of performance criterion.
i. often agents have to propose criteria themselves, and clearly it has no incentive effects when agents propose easily achievable goals.
ii. “ratchet effects”: lower down y[t] to make y[t+1] low. Again, agent manipulation. My friend as the chief of a bank in a district: made mistakes in having very high y[t] and then later judgment of lack of progress; once learned how to “pace yourself”, fine.
iii. often there is no measurements for particularly important aspects of agency performance. For instance, service quality, … without measurements, won’t be done.
c. Incentives:
i. public sector cannot pay well due to political constraints, so CEO incentives tend to be weak—Joskow and his co-authors’ study of US utility companies; Mengistae and Xu (2004) on CEOs of Chinese SOEs.
ii. political patronage: tough to fire workers if non-achieving of performance goals, again lower incentives.
iii. no strong incentives for gov’t agencies to monitor and design contract well: they don’t stand to gain much from good contract design.
d. Commitment:
i. if non-attainment of PC goals, cannot commit to punish firms, cannot bankrupt state enterprises, or fire managers (political appointees).
ii. firms cannot commit to act in the interest of the principal, especially in those hard to measured dimensions.
iii. subject to renegotiation problem.
Based on both case studies and regressions, Shirley and Xu (1998, 2001) find evidence that PCs have been shown not to work on average.
Shirley and Xu (1998, JLEO)
1. analyzes the experience with performance contracts between developing country governments and the managers of their state owned enterprises.
2. It identifies how problems of information asymmetry, incentives, and commitment can lead to shirking.
3. It applies this conceptual framework to a sample of 12 contracts with monopoly state enterprises in 6 developing countries
4. finds that all suffer from serious contracting problems and there is no pattern of improved performance that can be attributed to the contracts.
Table 4. Summary of the features of Performance Contracts
traits: / Ghana Electricity / Ghana Water / Ghana Telecom / India Electricity / India Oil / Korea Electricity / Korea Telecom / Mexico Electricity / Philippines Water / Philippines Electricity / Senegal Electricity / Senegal Telecom
Information:
quality of annual reports / poor / adequate / adequate / adequate / adequate / poor
Information:
Managerial bargaining power / * low pay and status
* reduction in powers / * used outside
experts
* competent
professionals
* stable staff / * used outside experts
* competent
professionals
* stable staff / * low pay
and status
* stable staff / *low pay and status
* high staff turnover / *low pay and status
* frequent shifts in
hierarchical position
Information:
Targets manipulation by managers / * targets set by managers
* targets changed frequently / * delay caused
targets based on
ex post
performance
* weights changed
frequently / * the weight of shadow
priced profit
decreased over time
* more than 40 targets / * targets not
weighted
* targets
changed
frequently / * most important targets
had decreasing weights / * averaged
23 targets
* targets set
by
managers / * targets set
by
managers
* averaged 37
targets
Incentives: rewards and punishments / * had bonus, but not paid
* managers political appointees / managers could receive award, not viewed as a strong incentives / manager and staff receive 3 month pay if got “A” grade / No / No / No
Commitment:
Costly action as signal / No / No / yes, allowed managerial autonomy / No / No / No
Commitment:
Enforcement mechanisms / * loans from the World Bank,
backed government contractual
obligation
* no neutral third party / * used outside
experts to
evaluate
performance
* no neutral third
party / * used outside experts
to evaluate
* no neutral third party / * loans from
the World
Bank,
backed
government
contractual
obligation
* no neutral
third party / no neutral third party / * loans from the World Bank
backed government
contractual obligation; but
not credible because the
government reneged
* no neutral third party
Commitment:
Government reneging on key promises / reneged on bonus, but met some early promises (increased tariff) / yes, e.g. not requiring public sector to pay bills / no / met its obligations (with some delays) / yes, e.g. not requiring the public sector to pay bills / yes, e.g. not requiring public sector to pay bills
(1)
with Yit = / TFP, ROA, or LP,,= / firm-specific, pre-contract productivity level and growth rate in Yit.
GDPGRit = / the GDP growth rate for the country of firm i; is its coefficient.
PCit= / a dummy variable whose value is 1 if a performance contract was adopted at year t. is the effects of contracts on productivity level.
= / the number of years since the contract’s adoption. captures the effects of the contract on the productivity growth rate.
Table 5. The effects of PC on ROA
(1) / (2) / (3) / (4) / (5) / (6)
Constant / 0.046 / -0.055 / -0.064 / 0.046 / 0.292* / 0.268
(0.668) / (-0.492) / (-0.564) / (0.665) / (1.752) / (1.396)
PC / -0.122 / -0.062 / -0.115 / -0.179 / -0.084 / -0.102
(1.235) / (-0.579) / (-0.894) / (-1.071) / (-0.767) / (-0.781)
/ 0.019 / 0.043 / 0.038
(0.425) / (1.092) / (0.829)
Inherent growth rate / -0.009 / -0.002 / -0.031 / -0.025
(-0.471) / (-0.093) / (-1.122) / (-0.721)
GDP growth rate / 0.027** / 0.029** / 0.032** / 0.032**
(2.234) / (2.331) / (2.481) / (2.468)
The year before PC adoption / -0.086 / -0.034
(0.747) / (0.261)
Control for firm dummies? / No / Yes / Yes / No / Yes / Yes
No. Obs. / 75 / 75 / 75 / 75 / 75 / 75
R. Square / 0.020 / 0.712 / 0.714 / 0.023 / 0.717 / 0.718
Table 6. The effects of PC on labor productivity
(1) / (2) / (3) / (4) / (5) / (6)
Constant / 1.095** / 0.253 / 0.241 / 1.095** / 0.382 / 0.383
(3.799) / (0.556) / (0.518) / (3.781) / (0.794) / (0.792)
PC / 0.767** / -0.016 / 0.035 / 0.516 / -0.072 / 0.170
(1.916) / (-0.032) / (0.058) / (0.838) / (-0.142) / (0.275)
/ 0.074 / 0.158 / 0.235
(0.5 40) / (0.849) / (1.078)
Inherent growth rate / 0.167* / 0.160 / 0.080 / 0.002
(1.862) / (1.593) / (0.587) / (0.013)
GDP growth rate / -0.056 / -0.058 / -0.037 / -0.036
(1.014) / (-1.017) / (0.617) / (-0.607)
The year before PC adoption / 0.080 / 0.421
(0.153) / (0.686)
Control for firm dummies? / No / Yes / Yes / No / Yes / Yes
No. Obs. / 77 / 74 / 74 / 77 / 74 / 74
R. Square / 0.047 / 0.668 / 0.668 / 0.050 / 0.672 / 0.675
Table 7. The effects of PC on TFP
(1) / (2) / (3) / (4) / (5) / (6)
Constant / 0.748** / 0.684** / 0.684** / 0.748** / 0.635** / 0.634**
(9.798) / (9.463) / (9.421) / (9.738) / (8.329) / (8.151)
PC / 0.032 / -0.093 / -0.063 / 0.081 / -0.062 / -0.065
(0.109) / (-1.304) / (-0.744) / (0.434) / (-0.864) / (-0.778)
/ -0.016 / -0.044* / -0.045*
(0.323) / (-1.806) / (-1.656)
Inherent growth rate / 0.032** / 0.028** / 0.052** / 0.052**
(2.731) / (2.181) / (3.260) / (2.705)
GDP growth rate / 0.020** / 0.019** / 0.016* / 0.016*
(2.533) / (2.311) / (1.89) / (1.873)
The year before PC adoption / 0.054 / -0.005
(0.677) / (0.061)
Control for firm dummies? / No / Yes / Yes / No / Yes / Yes
No. Obs. / 76 / 76 / 76 / 76 / 76 / 76
R. Square / 0.001 / 0.894 / 0.895 / 0.003 / 0.899 / 0.899
Shirley and Xu (2001, JLEO):
1. China’s experience with such contracts was one of the largest experiments with contracting in the public sector, affecting hundreds of thousands of state firms, and offered a rare opportunity to explore how PCs work.
a. The World Bank (1995) found 565 PCs in 32 developing countries as of June 1994, where they were principally used for large utilities and other monopolies,
b. and another 103,000 in China, where they were also used for manufacturing SOEs.
2. On average PCs did not improve performance and may have made it worse.
Table 3. The Effects of PC: Basic Results
ln(K/L) / 0.133 / 0.132
(5.10)*** / (5.28)***
ln(L) / 0.137 / 0.100
(2.06)** / (1.41)
markup / -0.512 / -0.496
(7.09)*** / (6.68)***
marginal rate / 0.308 / 0.086
(2.33)** / (1.43)
Autonomy / 0.082 / 0.063
(1.80)* / (2.52)**
W_discretion´ year / 0.030 / 0.026
(2.30)** / (2.25)**
New CEO ´ year / 0.041 / 0.045
(4.31)*** / (5.25)***
PC / -0.196 / 0.013
(2.41)** / (0.38)
N / 5094 / 5660
R-squared / 0.46 / 0.47
3. But China’s PCs were not uniformly bad;
a. in fact, PCs improved productivity in slightly more than half of the participants.
b. PC effects were on average negative because of the large losses associated with poorly-designed PCs.
c. Successful PCs were those that featured
i. sensible targets,
ii. stronger incentives,
iii. longer terms,
iv. managerial bonds,
v. and were in more competitive industries.
d. Selecting managers through bidding was not associated with performance improvement.
e. Good PC features were more often observed in SOEs
i. under the oversight of local governments,
ii. that faced more competition,
iii. that were smaller in size,
iv. and that had better previous performance.
Table 4. The Effects of PC provisions: dependent variable = log(value added/L)
controlling gfor ln(K/L), ln(L), markup, profit retention rate, autonomy, W_discretion * year, New CEO * year
PC / -1.599 / -0.125
(3.77)*** / (2.18)**
PC´W_elasticity / 0.897 / 0.203
(4.31)*** / (3.28)***
PC´bid_incumbent / -0.009 / -0.089
(0.08) / (1.93)*
PC´bid_newCEO / -0.397 / -0.393
(1.94)* / (4.38)***
PC´Length / 0.151 / 0.012
(3.40)*** / (1.05)
PC´Profit target / 0.384 / 0.072
(3.32)*** / (1.96)**
PC´1(bond) ´ year / 0.056 / -0.006
(1.46) / (0.27)
PC ´ markup / -2.049 / -0.087
(3.13)*** / (2.47)**
PC´profit target ´ markup / 1.179 / 0.106
(1.65)* / (2.97)***
Observations / 5094 / 5660
R-squared / 0.020 / 0.477
Table 5. Distribution of imputed PC effects by contract type (total PC participants=499 a)
LENGTH: lowW.ELAS.: low
markup: high / LENGTH: low
W.ELAS.: low
markup: low / LENGTH: low
W.ELAS.: high
markup: high / LENGTH: low
W.ELAS.: high
markup: low / LENGTH: high
W.ELAS.: low
markup: high / LENGTH: high
W.ELAS.: low
Markup: low / LENGTH: high
W.ELAS.: high
markup: high / LENGTH: high
W.ELAS.: high
markup: low
BID: yes
PROFIT: no
BOND: no / -2.28 / 0.52 / -1.81 / 0.99 / -1.95 / 0.85 / -1.47 / 1.33
-3.53 / 1.52 / -3.16 / 2.52 / -3.34 / 2.28 / -2.87 / 3.03
BID: yes
PROFIT: no
BOND: yes / -2.19 / 0.61 / -1.72 / 1.08 / -1.80 / 1.00 / -1.32 / 1.47
-3.55 / 1.73 / -3.15 / 2.64 / -3.31 / 2.48 / -2.78 / 3.11
BID: yes
PROFIT: yes
BOND: no / -1.25 / -0.06 / -0.77 / 0.42 / -0.91 / 0.28 / -0.44 / 0.75
-1.65 / -0.12 / -1.09 / 0.85 / -1.29 / 0.56 / -0.66 / 1.40
BID: yes
PROFIT: yes
BOND: yes / -1.16 / 0.36 / -0.69 / 0.50 / -0.76 / 0.43 / -0.29 / 0.90
-1.59 / 0.71 / -1.01 / 0.98 / -1.15 / 0.80 / -0.46 / 1.54
BID: no
PROFIT: no
BOND: no / -2.27 / 0.53 / -1.80 / 1.00 / -1.94 / 0.86 / -1.46 / 1.33
-3.51 / 1.55 / -3.15 / 2.56 / -3.32 / 2.33 / -2.88 / 3.10
BID: no
PROFIT: no
BOND: yes / -2.18 / 0.61 / -1.71 / 1.09 / -1.79 / 1.01 / -1.31 / 1.48
-3.51 / 1.75 / -3.13 / 2.67 / -3.29 / 2.51 / -2.77 / 3.16
BID: no
PROFIT: yes
BOND: no / -1.24 / -0.05 / -0.76 / 0.42 / -0.90 / 0.28 / -0.43 / 0.76
-1.62 / -0.11 / -1.07 / 0.90 / -1.27 / 0.60 / -0.64 / 1.47
BID: no
PROFIT: yes
BOND: yes / -1.15 / 0.04 / -0.68 / 0.51 / -0.75 / 0.43 / -0.28 / 0.91
-1.56 / 0.08 / -0.98 / 1.03 / -1.12 / 0.83 / -0.44 / 1.60