Push or pull? Performance‐pay, incentives, and information

• Published date: 01 March 2020
• Author: Yu Chen,David Rietzke
• Date: 01 March 2020
• DOI: http://doi.org/10.1111/1756-2171.12314

RAND Journal of Economics

Vol.51, No. 1, Spring 2020

pp. 301–317

Push or pull? Performance-pay, incentives,

and information

David Rietzke∗

and

Yu Chen∗∗,†

We study a principal-agent model wherein the agent is better informed of the prospects of the

project, and the project requires both an observable and unobservable input. We characterize the

optimal contracts, and explore the trade-offs between high- and low-powered incentive schemes.

We discuss the implications for push and pull programs used to encourage Research and Devel-

opment (R&D) activity, but our results are relevant in other contexts.

1. Introduction

To what extent should incentives be tied to performance? This question is relevant in many

areas, including labor markets—where it relates to the debate on salaries versus piece rates (see,

e.g., Lazear, 1986, 2000)—and innovation incentives, where it pertains to the efficacy of “push”

and “pull” programs (see, e.g., Kremer, 2002). Push programs, such as research grants or R&D

tax credits, subsidize inputs; payments are not contingent on results. Pull programs, such as

inducement prizes, or patent buyouts, tie rewards to output.

Adverse selection (AS) and moral hazard (MH) are inherent challenges in incentive provi-

sion. Given these problems, Kremer (2002) raises the concerns that push programs may finance

projects unlikely to succeed, and provide weak incentives for unobservable inputs. Indeed, the

MH literature stresses the importance of performance-pay; in the canonical model,1compensation

∗Lancaster University; d.rietzke@lancaster.ac.uk.

∗∗University of Graz.

A previous Working Paper was circulated under the title, “Push or pull? Grants, prizes, and information.” We thank

the Editor, David Martimort, and three anonymous referees for substantially improving the article. We also thank Stan

Reynolds, Andreas Blume, Asaf Plan, John Wooders, Martin Dufwenberg,Rabah Amir, Derek Lemoine, Dakshina De

Silva, Brian Roberson, Junichiro Ishida, Matthew Mitchell, Tim Flannery, Dominik Grafenhofer, Dawen Meng, and

Bo Chen. Finally, we thank participants at the Lancaster University Conference on Auctions, Competition, Regulation,

and Public Policy (May 2015, Lancaster, UK), EARIE Conference (August 2015, Munich, Germany), and the TILEC

Conference on Competition, Standardization, and Innovation (December 2015, Amsterdam, The Netherlands). Chen

gratefully acknowledges financial support from the National Natural Science Foundation of China (grant no. 71673133)

and the National Social Science Fund of China (grant no. 16BJL035).

†YuChen unexpectedly passed away in early 2019. He was a great friend and coauthor, and will be sincerely missed.

1See, for example, Grossman and Hart (1983) or Bolton and Dewatripont (2005).

C2020, The RAND Corporation. 301

302 / THE RAND JOURNAL OF ECONOMICS

must be tied to output to provide an incentive for effort. Yet, low-powered incentives in which

compensation is weakly, or not at all, tied to performance, are commonly used. In this article,

we explore trade-offs between high- and low-powered incentives in a model with AS and MH.

We show that performance-pay may not be optimal for all types, but is always optimal for the

highest types.

We consider a principal-agent model wherein a risk-neutral funder (he; the principal) moti-

vates a risk-neutral researcher (she; the agent) to undertake an R&D project. The outcome depends

on the researcher’s private type, and two essential and complementary inputs—“investment” and

“effort.” Investment is contractible; effort is not. 2If she succeeds, the researcher profits by mar-

keting the technology,but this incentive is insufficient from the funder’s perspective.To motivate

greater R&D activity, the funder specifies a transfer independent of performance—a “grant”–and

a payment for success—a “prize.”

Our results reveal that the virtues of performance-pay depend on the relative strengths of

AS and MH. In our model, a prize creates a strong incentive for effort, but generates costly rent

for the researcher due to AS. A grant effectively limits rent, but creates only an indirect incentive

for effort (by motivating investment). The virtue of the prize depends on the balance of these

trade-offs. In some circumstances, the optimal prize is zero for a range of types. For high enough

types, however, the prize is alwaysstrictly positive; moreover, when MH is more severe,the prize

is strictly positive for all types.

We contribute to the contracting literature under AS/MH. In many models, output is the

only verifiable signal available to the principal.3This renders performance-pay indispensable, as

output-independent rewards will not affect marginal incentives. Although it may be infeasible to

monitor research effort, some inputs, such as large-scale capital investments, may be easier to

verify. If so, then investment can be encouraged with rewards tied only to these expenditures.

However, a researcher’s effort may be more productive when she has better equipment with

which to work. Then, as long as there is some benefit to success, greater investment increases the

marginal returns to effort. A similar intuition obtains in multitasking models (e.g., H¨

olmstrom

and Milgrom, 1991; Meng and Tian, 2013). Given multiple complementary tasks, a stronger

incentive on one task induces greater effort on others.

We also contribute to the literature on innovation incentives under MH, which has largely

focused on pull programs. For instance, there is a large literature on optimal patent design

(e.g., Gilbert and Shapiro, 1990; Klemperer, 1990; O’Donoghue, Scotchmer, and Thisse, 1998;

Cornelli and Schankerman, 1999; Hopenhayn and Mitchell, 2001; Hall, 2007), and a literature on

alternatives to intellectual property such as prizes or contracts; (e.g., Wright, 1983; Kremer, 1998;

Shavell and Van Ypersele, 2001; Hopenhayn, Llobet, and Mitchell, 2006; Weyl and Tirole, 2012;

Che, Iossa, and Rey, 2015). However, few studies have examined the interactions between push

and pull programs taking MH into account. Maurer and Scotchmer (2003) argue that repeated

interaction between grantees and grantors resolves the MH problem. Our insights complement

theirs, as they are relevant in a static setting. Fu, Lu, and Lu (2012) showthat g rants mayfacilitate

greater competition in a contest between researchers with asymmetric capital endowments. We

abstract from competition to focus on the role of information.

Many studies have explored trade-offs between high- and low-powered incentives. Due to

the “effort-substitution problem,” H¨

olmstrom and Milgrom (1991) show the potential for “fixed-

wage” contracts.4This fixed wageis independent of any signal received by the principal; the grant

in our model depends on investment, but is independent of performance. Baker (1992) shows that

performance-pay may be muted if performance is weakly correlated with verifiable measures.

Low-powered incentives may also arise as a means of risk-sharing (see, e.g., Prendergast, 1999);

2Weuse the ter ms “observable,”“contractible,” and “verifiable” interchangeably.

3Studies close to this analysis include Lewis and Sappington (2000a), Lewis and Sappington (2000b), and Ollier

and Thomas (2013). There are notable exceptions, which will be discussed.

4The effort-substitution problem arises in multitasking models when efforts are substitutes. As a result, a stronger

incentive on one task reduces effort on the other task.

C

The RAND Corporation 2020.