incentive contracts and employee-initiated innovation ...first, employee-driven innovation...

0

Incentive Contracts and Employee-Initiated Innovation: Evidence from the Field

Wei Cai Harvard Business School

Susanna Gallani

Harvard Business School

Jee-Eun Shin University of Toronto

December 2, 2018

Preliminary and incomplete

Please do not circulate without permission.

Acknowledgments: We gratefully acknowledge the participants to the Harvard Business School Accounting and Management Brown Bag series, the participants to the Arizona State University seminar series, and the participants to the Harvard Business School Junior Faculty Research Workshop series for their helpful comments and suggestions. All errors remain our own.

1


Abstract Organizations often empower employees at all levels to propose innovation ideas that rely on their first-hand knowledge of their standard task (i.e. employee-initiated innovation). Many, however, struggle with motivating employees to develop innovative ideas that may benefit the firm, especially when the standard tasks for which employees are hired, measured and incentivized do not explicitly include innovation. Prior analytical research posits that low-powered incentives can motivate employees to pursue innovation opportunities by reducing the pressure to deliver on performance measures associated with their standard tasks included in the incentive contract. Using data from a Chinese manufacturing company where employment contracts for standard tasks exhibit significant variation in terms of composition of fixed and variable components of pay, we examine whether the structure of incentive contracts for the standard tasks influences employees’ propensity to engage in innovation activities. We find that employees under fixed-pay contracts are more likely to pursue innovation ideas that are valuable to the firm relative to employees under variable-pay contracts. Moreover, such efforts are concentrated on innovation ideas that are not specific to the standard task performed by the proposing employee, but are applicable to issues of greater breadth for the firm and/or with a long term view. We perform a battery of additional tests to rule out endogeneity concerns, to validate the robustness of our findings, and to validate the impact of contract structure on important organizational outcomes. Our results contribute to the literature on the effectiveness of using low-powered incentives to encourage unplanned employee-initiated innovation activities that are difficult to contract upon ex ante. Keywords: employee-initiated innovation, contract design, creativity, low-powered incentives JEL Codes: M41, M52, M54, M55 Data availability: The data used in this project is subject to a confidentiality agreement and cannot be shared without express consent of the organization’s legal representatives. Funding: This research did not receive any specific grants from funding agencies in the public, commercial, or not-for-profit sector.

2


1. Introduction

Many organizations encourage employees at all levels to propose initiatives for process

improvement, cost reductions, productivity enhancement, and improvements in the work

environment. These ideas, described in the literature as employee-initiated innovation activities

(Li, 2016), often rely on the first-hand knowledge of workers whose main job design does not

formally include responsibilities for innovation (i.e. developing new products or services,

generating new production or delivery methods for existing products or services), but are hired

and rewarded for delivering standard execution tasks. This study empirically examines whether

the design of incentive contracts for standard tasks influences employee-initiated innovation

activities.

Promoting employee-initiated innovation is critical to sustain competitive advantage in

many organizations. Some firms have formalized dedicated information and knowledge

management systems to collect and evaluate innovation initiatives proposed by employees.

Notable examples include Toyota’s iconic “Creative Ideas Suggestive System”, or Whirlpool’s

“idea labs”.1 Yet, the provision of incentives to encourage employee-initiated innovation activities

remains a significant challenge. First, employee-driven innovation activities arise unplanned such

that incentive contracts cannot be based on objectives that can be contracted upon ex-ante. Second,

employee-initiated innovations are often rewarded ex-post, contingent on the proposed innovation

being adopted and, in many cases, on their estimated incremental value. The uncertainty in the

1 Morgan, J. “The 5 Types of Innovation for the Future of Work. Pt 1: Employee Innovation” https://www.forbes.com/sites/jacobmorgan/2015/07/27/the-5-types-of-innovation-for-the-future-of-work-pt-1-employee-innovation/#3d8d489e7e20

3

outcomes associated with pursuing innovation opportunities may discourage employees from

expending effort in developing innovation ideas. Our study sheds light on how firms can influence

employees' innovation decisions through design choices of incentive contract for the standard

tasks.

Prior analytical work on innovation incentives examines the distribution of effort between

innovation activities and standard tasks as a multitasking problem, where employees allocate a

finite amount of effort between competing activities (i.e. standard task vs innovation) based on

their opportunity cost and the uncertainty in their measurement and reward (Holmstrom, 1989;

Holmstrom and Milgrom, 1991; Hellmann and Thiele, 2011). A fundamental insight from this

approach is that low-powered incentives for the standard task provide "breathing room" to pursue

potential innovation ideas by lowering the opportunity cost (potentially to the point of removing it

completely in presence of fixed pay) of the standard task. However, the assumption of finite

amount of effort that needs to be allocated between standard task and innovation activities as if in

a zero-sum game, is not necessarily realistic with respect to employee-initiated innovation. These

innovation ideas are generally less disruptive than the development of new products or new

production/delivery processes, and require less effort and time to formalize.2 In many cases,

workers plausibly come up or develop these ideas in their own time before submitting them for

approval or funding. In many cases, organizations reward successful employee-initiated

innovations with monetary rewards and other types of recognitions.

2 “An employee who just finished ringing up a customer at a retail store may have noticed a way to improve the employee experience, a call-center agent might have an idea to reduce call-time, a knowledge worker may have a suggestion for how to improve employee engagement, and a member from the janitorial staff might have an idea for how to reduce costs on wasted supplies”. Morgan, J. “The 5 Types of Innovation for the Future of Work. Pt 1: Employee Innovation” https://www.forbes.com/sites/jacobmorgan/2015/07/27/the-5-types-of-innovation-for-the-future-of-work-pt-1-employee-innovation/#3d8d489e7e20

4

Submitting employee-innovation ideas is also a way for workers to signal their type to

management. In organizations that foster a culture of empowerment, innovation, and

entrepreneurialism, innovative employees might increase their opportunities for future

compensation raises or promotions. At the same time, however, submitting innovation ideas

exposes workers to the risk of signaling low ability, to the extent that their ideas are of low quality

or fail to generate the expected benefits. A number of theories in strategy and management stipulate

that appropriate levels of "slack" and tolerance for failure can motivate employees to engage in

exploration of new opportunities as opposed to exploitation of existing assets (Manso, 2011,

Nohria and Gulati 1996, Damanpour 1991, Holmstrom 1989). While slack can be provided in

many ways, including target setting, budget allocations, or reduced time pressures on deliverables,

we posit that low powered incentives provide employees with lower uncertainty with respect to

the payoff associated with their standard task, and consequently weaken their fixation on the

standard task. As a result, we expect employees rewarded with low-powered incentives to engage

in innovation activities that benefit the organization to a greater extent than employees subject to

high-powered incentive contracts.

We test our prediction using field data from a Chinese manufacturing company where

employees are motivated to perform their standard operational tasks based on explicit incentive

contracts. The operations of the firm are very labor intensive, and require relatively low skill levels.

However, management encourages employees at all organizational levels to submit innovation

ideas (hereafter, employee-initiated innovation ideas) to improve firm productivity, quality,

working conditions, or reduce costs, with the potential to be rewarded upon positive review by

management (i.e. ex post settlement) in addition to their explicit incentive contracts. Innovation

ideas can vary widely in the scope of applicability and may extend beyond the employees’ standard

5

tasks. Compensation contracts for the standard task vary cross-sectionally and can be structured as

fixed pay only, variable pay only, or a combination of fixed and variable components. The variable

component is based on an output measure that summarizes the performance of the standard task to

which the employee is assigned (volume-based contract). While the assignment of a particular type

of contract structure to a worker depends on the worker’s main responsibilities (i.e. production

workers vs. management), institutional characteristics of this site lead to significant variation in

the composition of compensation contracts within equivalent organizational roles. We discuss

these institutional characteristics in section 4.

First, we examine whether the propensity to engage in innovation activities exhibits

significant differences depending on contract type. We predict and find that, compared to variable

pay, fixed-pay contracts are associated with significantly greater worker propensity to engage in

employee-initiated innovation activity. Employees rewarded with fixed pay are more likely to

expend effort to submit innovation ideas valued by management as being beneficial to the

organization. This result is robust to accounting for potential differences in the task nature, and

also to including control variables capturing employee characteristics that may affect an

individual’s propensity to expend effort on innovation-related activities.

Second, we examine the influence of compensation structure on the scope of applicability

of employee-initiated innovation ideas. Workers might focus on innovation that pertains narrowly

to their standard task in an attempt to improve output and/or reduce effort (examples include

changes in the order of operations within a manufacturing process, standardization of operating

tasks, solutions to reduce the incidence of rework or increased yield, etc.). Alternatively,

employees might develop ideas with a broader scope of applicability, such as improvements in the

general working conditions, indirect cost reductions, adoption of improved technology for

6

automation, or improvements in the long-term profitability and sustainability of the organization.

We posit that high-powered incentives might drive a narrow focus in the development of

innovation ideas, whereby employees might focus on ways to improve their individual

productivity, thus increasing their output-related payoff. Conversely, relaxing the fixation on the

standard task through the provision of low-powered incentive might induce a more collaborative

organizational culture, by which employees are more willing to share valuable information with

each other and generate innovation-related returns benefiting a larger number of constituents

within the organization. Our results indeed show that employees paid on a fixed schedule are more

willing to pursue innovation ideas with broader scope of applicability, whereas the occurrence of

narrowly defined task-specific innovation ideas does not exhibit any difference across different

incentive contract forms. Taken together, our results suggest that, compared to volume-based

incentive structures, low powered incentives are more conducive to employee-initiated innovation

activity and, in particular, to the generation of suggestions that impact a larger set of constituents,

even in conditions where the allocation of effort between the standard task and the innovation

activity is not subject to the zero-sum game constraints and trade-offs typical of a multitasking

environment adopted by prior literature.

We perform additional tests to explore the impact of contract structure on other important

organizational outcomes. Since innovation is not a primary responsibility for the workers we

examine in this study, we question whether the allocation of different contract structures might

influence dimensions of performance such as meeting productivity targets, production quality, or

employee turnover. We find no material differences in these organizational outcomes that we can

attribute to differences in the structure of incentive programs.

7

Next, we conduct a series of robustness tests to corroborate our main empirical findings.

First, we provide evidence to reduce the concern that the assignment of contract types may be

endogenously determined. A contract type determinant model estimates that managerial roles are

more likely to be rewarded with fixed pay contracts, and that fixed pay contracts are more likely

to be assigned to workers hired after the company underwent a merger which resulted in significant

changes in the top management team (our sample period starts immediately after the merger). To

disentangle the effect of the new management from the effect of the contract type, we re-estimate

our main test including an interaction between type of contract and the time of hire, which

continues to support our main finding that low-powered incentives are associated with greater

likelihood of employee-initiated innovation activity. Second, even though we include department-

level fixed effects in all our main analyses, we confine our sample only to departments that exhibit

variation in their hiring practices. Due to the nature of the performed function, some departments

are less likely to exhibit variation in the form of employee incentive contracts – that is, every

employee in the department exhibits the same type of contract. Our results are robust to the sample

restriction. Third, because the contract type is time-invariant in our sample, we estimate cross-

sectional regressions at the employee-level (whereas our main tests involve panel data to capture

variability in the company’s operations over time as well as seasonality), and confirm that

employees hired on a fixed contract are more likely to engage in innovation-related activities at

any point in time within our sample time period. Finally, we employ an instrumental variable

approach using the month of hire as instrument predicting the type of contract assigned to

employees, and continue to find results consistent with our main tests.

Our study is one of the first to show how standard incentive contracts interact with

employee unplanned innovation activities using field data. A large literature focusing on the

8

optimal use of incentives and rewards for innovation activities assumes the availability of a

performance measure that can proxy for innovation activities (Holthausen et al., 1995, Lerner and

Wulf, 2007, Kachelmeier et al., 2008). However, innovation measures that are contractible ex-ante

are only rarely available in practice, especially with respect to employee-initiated innovation as

defined in our study. Additionally, while prior literature examines employees for whom creativity

and innovation falls within their main responsibilities (e.g. R&D or advertising staff), we focus on

unplanned innovation activities performed by employees whose main responsibility is to perform

standard execution tasks, for whom innovation activity cannot be contracted upon ex ante. Our

findings empirically corroborate the theoretical prediction put forth by multitasking theory that

low-powered incentives provide employees with "breathing room" to "jump out of their box"

(Hellmann and Thiele, 2011), even in absence of the stringent zero-sum game assumed by

analytical research with respect to effort allocation between activities measured and rewarded with

different degrees of uncertainty.

This study also contributes to the debate on the appropriateness of complete versus

incomplete incentive contracts. On the one hand, stemming from the work of Holmstrom (1979),

extant literature on the design of optimal incentive contracts has adopted the informativeness

principle, by which incentive contracts should include all variables that provide information about

the agent's actions. On the other hand, a different stream of literature has explored the benefits

associated with "incomplete contracts" (Grossman and Hart 1986; Hart 1995; Hart and Moore

1990, 2008), which take into consideration elements of performance that are not foreseeable or

contractible ex-ante as part of the firm's standard operational processes. This study adds to the

literature by comparing incentive contracts structured as either fixed or variable pay based on an

assigned standard task, and examining the effect on unplanned employee-initiated innovation

9

activities that are not outlined in the agent’s standard incentive contract. Thereby, we shed new

light on how contract structure can mitigate contracting difficulties of motivating unplanned

employee-initiated innovation activities that are difficult to contract upon ex ante.

The remainder of the paper is organized as follows. In Section 2, we provide an overview

of the extant literature that examines the provision of incentives for innovation. In Section 3, we

describe relevant theoretical predictions and develop our hypotheses. Next, we describe our

research setting and explain the suitability of our research site to address our research question

empirically (Section 4). We describe our research design in Section 5. Section 6 summarizes the

results of our statistical tests and our main inferences. We conduct additional tests and validate the

robustness of our results in Section 7, and conclude with Section 8.

2. Prior Literature on Incentive Systems for Innovation

A growing body of literature in accounting and economics has explored mechanisms to

foster creativity and innovation in organizational settings. Successful models of organizational

innovation exhibit a culture of appreciation for innovation activities manifested through

recognition and rewards, and of support for exploration and risk-taking through availability of

time, resources, autonomy, and tolerance for failure (Amabile, 1988; Damanpour, 1991). Research

studies have addressed various determinants of innovative activities in organizations, including

monetary and non-monetary incentive mechanisms.

Appropriately structured performance-based financial incentives can be effective in

motivating innovation and can lead to superior results compared to settings where financial

incentives are absent (Ederer and Manso, 2013). Several studies recommend the inclusion of long-

term oriented provisions and protections from early failure and external pressure in the design of

incentive systems aiming at stimulating innovation activities. Analytical work by Manso (2011),

10

supported by controlled experiments by Ederer and Manso (2013) and Ederer (2013), as well as

empirical studies (Baranchuk et al, 2014; Holthausen et al, 1995; Lerner and Wulf, 2007) support

the use of stock options with long vesting periods, profit sharing, and golden parachutes as

effective incentive mechanisms for innovation activities.

While considerations that organizational culture starts at the top (Amabile, 1988) motivate

studies exploring the effectiveness of executive compensation incentive structures on innovation

activity at the organization level, the role of middle managers and front-line employees is critical

for a successful culture of organizational innovation (Amabile, 1988, Baumann and Stieglitz et al.

2014; Holthausen et al., 1995). Many organizations encourage rank-and-file employees to generate

innovative ideas that are informed by their first-hand knowledge, skills, and experience. In many

cases, these employees are not “full-time innovators” (Li, 2016), in that their primary

responsibilities relate to execution tasks that require little, if any, creativity. We refer to this type

of innovation activity as “employee-driven innovation”, where the term “driven” highlights the

feature of innovative ideas that are generated and selected by employees without formal

assignments by their superiors (Li, 2016). The use of stock options or protection devices such as

golden parachutes are rarely observed in compensation contracts for rank-and-file employees.

For many employees, pursuing innovation opportunities means going above and beyond

their job descriptions (Birkinshaw and Duke, 2013, Hellmann and Thiele, 2011), and effective

goal-directed incentives need to be in place to ensure that innovation activity happens and that it

is performed to develop ideas that are beneficial to the organization (Grabner, 2014, Baumann and

Stieglitz et al., 2014). Incentives for innovation can be provided through either explicit incentives

associated with the innovation task itself, or by lowering the marginal opportunity cost of the

11

innovation task by reducing or removing incentives on alternative activities that compete for the

effort of the worker (Holmstrom and Milgrom, 1991).

Organizations tend to include creativity-related incentives in their compensation contracts

relatively infrequently (Ittner et al. 1997). Plausible reasons include the possibility that these

rewards might be ineffective to stimulate innovation output, or that measuring creativity and

innovation performance might be difficult (Kachelmeier et al. 2008). Incentivizing and rewarding

employee-driven innovation performance presents several challenges (Holmstrom, 1989). First,

employee-driven innovation ideas arise unplanned, thus impeding the definition of performance

metrics that can be contracted upon ex-ante and leaving ex-post settlement as the next best solution

to reward innovation performance. Second, innovation ideas are difficult to compare with one

another, thereby creating difficulties in the evaluation of effort associated with each initiative, and

in the matching between innovation ideas and rewards. Even when explicit incentives for

innovation are present, organizations lament lower than expected participation in innovation

activities, which they attribute to employees experiencing short-term pressure to deliver upon their

main job responsibilities (Li, 2016).

Prior literature has identified organizational slack and reduced time pressure as powerful

mechanisms to ease short-term performance pressure and provide degrees of freedom that may

allow employees to act upon opportunities for exploration (Nohria et al., 1996, Birkinshaw and

Duke, 2013, Damanpour, 1991, Li, 2016). Organizational slack can be provided through setting

targets and budgeting that are easier to achieve, by providing flexibility in the time requirements

set for workers’ deliverables. To the extent that the structure of employee compensation and the

relation between standard task performance and compensation payoffs contributes to determining

12

the pressure to deliver in the standard task, reducing the power of incentives with respect to the

standard task is another way to provide employees with organizational slack.

3. Theoretical Background and Hypotheses Development

In multitasking settings, the cost of the incentive associated with a particular task depends

on the mix of tasks that the individual employee is assigned, and the characteristics of their

performance measures. In general, employees make allocation choices between activities that are

measured with different degrees of sensitivity and precision (Banker and Datar, 1989) and

rewarded with different degrees of predictability (e.g. subjective measurement, relative

performance evaluation, etc.). Multitasking theory predicts that employees will shift effort toward

activities that are easy to measure and are more clearly associated with greater monetary payoffs

(Holmstrom and Milgrom, 1991). Analytical studies and controlled lab experiments concur that,

compared to canonical pay-for-performance agreements, low-powered incentives for standard

tasks are better suited to incentivize employee-driven innovation in multi-tasking environments,

(Baumann and Stieglitz, 2014, Manso, 2011, Hellmann and Thiele, 2014; Ederer, 2013, Ederer

and Manso, 2013). Multitasking theory, however, assumes that employees are endowed with a

finite amount of effort to allocate across a finite number of competing tasks, and that rewards for

each task can be determined based on signals informative of the employee effort - albeit sometimes

through ex-post bargaining (Hellmann and Thiele, 2011).

Relaxing the requirement of a finite amount of effort to allocate between competing tasks

– as it is often the case with respect to employee-initiated innovation – converts the incentive

structure into an additive model, whereby an employee engaging in innovation activities that are

considered valuable by the organization will likely face little (if at all) trade-offs in the allocation

of effort between competing tasks, and add the reward for the innovation activity to the

13

compensation for the standard task. We explore whether and how, in absence of a strict zero-sum

game allocation requirement, different levels of incentive intensity associated with the standard

task lead to different workers’ propensity to engage in innovation activities. We consider opposite

extremes in the continuum between fixed pay and variable pay.

On the one hand, employees rewarded with fixed (variable) pay experience less (more)

pressure to deliver on their standard task and might be more (less) inclined to expend effort on

innovation, contingent that the expected reward for the innovation idea is greater than the cost of

effort required to develop it. On the other hand, however, fixed (variable) employees are less

(more) likely to benefit directly from improved productivity, due to the weak (strong) link between

volumes produced and incentive payoffs. To disentangle these competing predictions, we test the

following hypothesis, which we express in null form:

H1(null): Low-powered incentives for standard tasks are associated with similar likelihood of pursuit of employee-driven innovation ideas, compared to high-powered incentives. Employee-initiated innovations vary in their scope of applicability. Employees may

leverage on their first-hand knowledge to develop improvement initiatives narrowly defined

around their standard task (e.g. improved throughput, reduced production downtime, improved

production flow, etc.). Alternatively, employees might propose ideas that offer broader

applications and benefit a larger set of constituents (e.g. cost reduction initiatives, improvement of

the general work environment, promotion of collaboration between departments, etc.).

Management is generally interested in both kinds of ideas to the extent that the proposed

innovation converts directly or indirectly into increased efficiency, productivity, or profitability.

Employees, however, might benefit differently from the ideas they propose. Narrowly defined

innovations might have a more direct effect on individual productivity (e.g. changes in the

14

production flow might eliminate a bottleneck in the production process, thereby increasing

employee output per unit of time), whereas broadly defined innovations might generate returns for

the proponents that are more difficult to identify and might materialize sometime in the future (e.g.

cost reductions might allow the company to convert corresponding savings in greater job security

or future pay raises for all employees). Our rich research setting allows us to observe the incidence

of innovation ideas of different nature and scope of applicability. Therefore, in exploratory fashion,

we analyze the relation between the structure of the compensation contract associated with the

standard task and the type of innovation activity in which workers might be more likely to engage.

Employees rewarded with variable pay contracts are primarily incentivized based on the

productivity of their assigned standard task which may limit their attention to innovation ideas that

are highly task-specific. On the other hand, employees compensated with fixed contracts may

allow for a broader focus also encouraging the submission of non-task-specific innovation ideas.

Accordingly, the design of the incentive contract based on the standard task may significantly

interact with the incentives to pursue different types of innovation opportunities.

4. Research Setting

4.1. Company Description

Our field data is obtained from a Chinese manufacturing firm that produces packaging

materials and supplies. The firm maintains a stable client base such that its revenue stream is

largely predictable. However, production orders exhibit seasonal fluctuations – the firm’s busiest

months of operations are in the summer and fall driven by the production orders of their two major

clients, whereas production slows down in the winter months. Due to the small margins typical of

this industry, firm profits largely depend on its ability to reduce production inefficiencies (such as

15

quality defects, machine downtime due to technical issues, and rework) and to improve cost

efficiency.

The production process is organized into 11 phases that each constitute a department.

Examples of such departments include the box-gluing department, the laminating department, the

printing department, the storage and transportation department, etc. Therefore, employees within

each department are assigned a primary task (i.e. “standard task”) that is crucial in maintaining the

overall flow of the production process. The tasks assigned to each individual department differ in

their nature, but are fairly comparable in terms of task complexity, and can be measured using

readily available performance measures based on the volume of units completed.

4.2. Incentive Contracts of Rank-and-file Employees

Firm productivity is dependent on the rank-and-file employees who are directly involved

in the firm’s production processes. They are rewarded based on explicit incentive contracts where

total incentive compensation is determined by combinations of fixed and variable components.

These contracts can assume one of three different forms: a contract that is (1) only based on a fixed

component (Fixed), (2) only based on a variable component (Variable), or (3) a combination of a

fixed and a variable component (Mixed). The variable component is determined based on the output

measure that summarizes the productivity of units completed with respect to each department.

The type of contract is determined at the time of the employee hire, and which type of

contract is being offered to the employee depends on institutional characteristics as well as the

type of role defining the main responsibility of the employee (for example, employees hired as

managers are more likely to be offered a fixed contract, whereas a combination of contract

structures is observed among front-line workers). Institutional factors influencing the choice of

contract assigned to a new hire include the time of the year in which a particular employee is added

16

to the roster, as well as management preferences for fixed contracts. During the busiest months of

the year, the company is more likely to hire front-line workers. Because of the high volumes of

production in those months, variable contracts are attractive to prospective employees. During idle

production times, the management is more likely to hire back-office staff or maintenance workers,

who are typically rewarded with fixed contracts due to the weak link between their job

responsibilities and production volumes. Finally, during off-peak production months, management

might offer fixed or variable contracts in its hiring practices, which encompass all roles in the

organization. Management, however, exhibits preferences for fixed contracts, due to the greater

predictability of the related financial commitment, as well as the intent to provide greater income

stability to its workers.6 We note that while the time of the year might influence the type of contract

offered to the employee, it does not impact the likelihood of retention of the new hire. 7

Additionally, absolute compensation amounts do not significantly differ across different contract

types for similar roles.8 Finally, in our sample period, we do not observe a change in the employee

contract type from their initial hire.

4.3. Promotion of Employee-initiated Innovation Ideas

Due to the small margins and the labor-intensive nature of its main operations, the firm

empowers its front-line employees to contribute to improving profitability through the proposal of

ideas that might improve efficiency, productivity, and profitability. Accordingly, in addition to the

6 A recent merger event at the beginning of 2014 resulted in significant changes in the composition of top management. While employees hired prior to the merger event were retained and their contracts were not modified, the current management is more likely to offer fixed contracts to employees hired for organizational roles previously rewarded with variable pay. This variation in contracting preferences allows us to examine the influence of contract structure on the propensity to engage in innovation behavior across employees with similar main task responsibilities. 7 In particular, we do not observe the hiring of temporary workers in our sample period. 8 We were able to obtain compensation amount data for each employee at one single point in time during our sample period. T-tests comparing the compensation amounts between the three different contract types do not reveal any differences that are statistically significant.

17

explicit incentive contract related to employees’ standard task, the firm rewards the submission of

feasible and advantageous employee-initiated innovation ideas.

Employees at all levels of the organization can submit innovation ideas, which vary widely

in their scope of applicability – ideas that directly relate to the employees’ standard task, but also

ideas that can benefit processes in other departments or the overall organization are encouraged

for submission. However, not all innovation idea submissions are rewarded. Each idea submission

is evaluated by management, and employees receive a monetary award only if management

assesses the idea to be valuable to the firm. The amount of the award is not pre-determined, but

decided ex-post on a case-by-case basis. Additionally, there is no objective evaluation system for

the submitted innovation ideas. Instead, management subjectively assesses how the submitted idea

can potentially enhance overall firm performance. Approved ideas result in an additional award

opportunity ranging from 1%-3% of the proponent’s monthly pay.

We provide examples of innovation ideas submitted by employees in the Appendix. The

decision to submit innovation ideas depends on the employee’s willingness to share their first-

hand knowledge with other stakeholders of the organization. Even though submissions of any ideas

are associated with potential additional award opportunities, employees may exhibit significant

variation for the type of innovation ideas they propose across the different contract types.

To evaluate the submitted innovation ideas, management maintains a framework by which

they distinguish between different types of innovation ideas. The innovation categories and the

corresponding descriptions used to evaluate the submitted ideas are provided in the Appendix, and

can be largely categorized into two broad categories – task-specific innovation ideas (narrow

scope) and non-task-specific innovation ideas (broad scope). Task-specific innovations include

ideas that improve efficiency (e.g. speed, throughput, etc.), quality of the process (e.g. incidence

18

of rework, defects, etc.), or the standardization and streamlining of the production process (e.g. 5S

initiatives). In contrast, non-task-specific innovations include suggestions that will benefit the

organization via improvements in activities other than the employees’ standard tasks. Examples

include initiatives that promote collaboration across teams or departments, ideas that increase

automation, reduce costs, or improve the long-term sustainability, morale or culture of the

organization.

5. Research Design

5.1. Sample and Data

Our sample includes monthly employee-level data from March 2014 to December 2016.

There are total of 513 unique employees, and the sample comprises of a total of 6,016 employee-

month observations. For each month in the sample period, we collect information on the number,

type, and quality of innovation idea submissions by all employees. In addition, we obtain data on

the design of the incentive contract for each employee, and additional background and

demographic characteristics. In line with local business practices, the company operates its

production lines ten or eleven months each year, with January and February corresponding to idle

time. A detailed description of the variables of interest for our analyses are provided below.

5.2. Variables of interest

5.2.1. Dependent Variables: Innovation

We are interested in studying the drivers of the employee’s propensity to propose an

innovation idea. We proxy for it using an indicator variable Submissioni,t which assumes value one

if employee i submits an innovation idea in month t, and zero otherwise. Next, we analyze

innovation activity based on the classification used by the company, which maps into our two

broader categories based on the breadth of the scope of applicability (degree of task-specificity) of

19

the innovation idea (see Appendix). We construct indicator variables representing idea

submissions for each innovation type, assuming, respectively, value one if employee i submitted

an idea of the particular type in month t. Broad scope categories include ideas with long term

benefits (Sub_lti,t), ideas benefiting a group or a team (Sub_groupi,t), ideas benefiting a different

department (Sub_diffdepi,t), ideas to improve the technology, automation, and computerized

systems of the firm (Sub_techi,t), and ideas aiming to reduce overhead costs (Sub_costi,t). Narrow

scope categories include ideas to improve standardization and streamlining of operating tasks

(Sub_5Si,t), ideas improving the quality of the production process and lowering defects and rework

(Sub_qualityi,t), and ideas improving efficiency and throughput (Sub_efficiencyi,t).

Table 1, Panel A provides descriptive statistics on the innovation-related variables.

Innovation submissions occurred only in about 6% of our employee-months. Moreover, we

observe that the submitted ideas are predominately considered to be viable by management – about

95% of the submitted ideas have subsequently been rewarded with a bonus. We proxy for the

quality of the submitted innovation idea using the indicator variable Approvedi,t, which assumes

value one if the innovation idea by employee i in month t is considered to be viable and therefore

is rewarded with a bonus, and zero if not. Whereas Submission captures an input measure of

innovation, Approved captures an output measure. The policy to award monetary rewards only to

the proponents of valuable innovation ideas provides a signal to employees that submitted

innovation ideas will be evaluated, and only those deemed valuable will be rewarded. In absence

of such a policy, employees might be tempted to maximize the quantity of innovation submissions

at the expense of their potential quality in the hope that some of those ideas might be approved

and rewarded. This possibility would be particularly salient for employees under fixed contracts

as they do not enjoy any upside potential associated with their assigned standard task (i.e. more

20

effort does not translate in greater payoffs). We interpret the high approval percentage together

with the low incidence of innovation submissions as a signal of employees being selective with

respect to their innovation ideas. While submitting innovation proposals might signal an

entrepreneurial type of employee, submitting bad ideas might also factor negatively into

management’s future decisions with respect to job assignment, promotions, or pay increases.

Additionally, as shown in Table 1, Panel B, only about 15% of the employees engage in innovation

activities, suggesting a high concentration of innovation activity within a limited sample of

employees. Learning about the relation between contract structure and individual propensity to

innovate might shed light on ways to increase the percentage of “innovators” in organizations.

Looking at the different innovation types, we note a higher frequency of submission of

innovation ideas aiming at reducing costs, improving efficiency, and promoting long term

organizational outcomes. The ratio between submission and approval of the innovation idea is

largely consistent across innovation categories.

5.2.2. Independent Variables: Contract Type

In our field setting, we observe three different types of employee incentive contracts

pertaining to the employee’s assigned standard tasks. These contracts differ depending on the

extent by which compensation is based on fixed versus variable components. We define the

different types of contracts as Fixedi, Variablei, and Mixedi, respectively. The type of contract for

each employee is determined at the time of hire, and does not change over the course of our sample

period. In other words, in our setting, the type of contract constitutes an employee-level fixed

effect.12 Employees under a Fixed contract receive the fixed pay amount at the end of each month

12 A concern might arise with respect to endogenous selections of contract type. That is, one could expect that contract structure might reflect, among other things, the propensity of the employee to generate innovative ideas. We determine that this is not the case by estimating a determinant model for the type of contract, and by implementing a 2SLS

21

regardless of their performance on their standard task. Whereas employees under a Fixed contract

are not subject to downside risk with respect to their monthly compensation, they do not enjoy any

upside potential for greater performance on the standard task. Employees under a Variable contract

do not have a guaranteed pay amount, and their end-of-month pay is entirely dependent on their

performance on their standard task. Employees under a Variable contract enjoy unbounded

incentive compensation, but are exposed to greater downside risk as their compensation has no

floor. Employees under a Mixed contract are guaranteed a fixed amount at the end of each month,

and can earn additional compensation based on the performance on their standard task. As shown

in Table 1, Panel C, about 58 percent (13 percent) [30 percent] of all employees are hired with a

Fixed (Variable) [Mixed] contract.

5.2.3. Control Variables: Employee Characteristics

We control for employee individual characteristics that may be associated with the

employee’s propensity to engage in innovation-related activities (Table 1, panel C). We include

DormEmpi, an indicator variable assuming value one if the employee lives in company-provided

housing (dormitory), and zero otherwise. Workers living in the company dormitory are generally

single, and enjoy a significantly shorter commuting time compared to their colleagues living at

home. Additionally, sharing common areas, such as cafeterias, exercise facilities, or leisure spaces

might increase the opportunity to exchange ideas and develop innovations collectively. Moreover,

reduced commitments to family obligations and commuting time might provide these workers with

more time to engage in innovation-related activities outside their assigned standard tasks. About 7

percent of all employees in our sample live in dormitory facilities.

estimation of our main model using instrumental variables. Details of the analysis and estimation results are reported in Section 7.1.

22

Next, we control for gender, using the indicator variable Femalei, which assumes value one

if the employee is female, and zero otherwise. About 38 percent of all employees in our sample

are female. Gender is likely associated with personality traits such as creativity, extroversion,

confidence, selflessness, etc., which might impact the employee’s propensity to propose

innovation ideas.

Further, we control for the age of the employee (variable Agei) measured in number of

years.13 Agei may correlate with an employee’s experience level and knowledge base which may,

in turn, impact the ability to identify opportunities and generate innovation ideas. The average

employee in our sample is about 33 years old. We also control for Tenure, which measures the

length of the contractual relation between the organization and the employee in years. On the one

hand, employees that have been with the company for a longer time might have accumulated

greater institutional and technical knowledge which they can leverage to develop valuable

innovation proposals. On the other hand, relatively new hires might be in touch with more recent

technological developments, organizational solutions that they might have seen in other firms, or

simply hold an unbiased view of the needs and processes of the operations, which might lead them

to propose fresh innovation ideas. The average tenure in our sample is 1.8 years, and spans between

a minimum value of 1 year to a maximum of 17.

We also control for the rank of the employee within the company. Mgmti is an indicator

variable assuming value one if the employee performs managerial functions, and zero otherwise.

Similar to age, a higher rank within the company may be associated with better ability and/or

experience, which may impact innovation-related activities. About 8 percent of all employees in

our sample perform a management function.

13 Age is measured at the beginning of our sample period and maintained constant over the months included in our sample.

23

Finally, we control for the time of hire of each employee included in our sample. As

previously noted, the company underwent a merger event at the beginning of 2014, which led to a

significant change in the composition of top management. While interviews with the current

management team indicated no explicit relation between the contract offered to new hires and their

propensity to innovate, it is possible that changes in the selection criteria might confound our

results. While the encouragement of employee-initiated innovation has always been central to the

firm, and the incentive scheme to reward innovation idea submissions existed even prior to the

merger, we conservatively control for the time of hire of employee i using an indicator variable

(JoinAfterMergeri) assuming value one if the employee was hired after the merger event and zero

otherwise.

--- Insert Table 1 here ---

6. Statistical Tests and Empirical Results

6.1. Incentive Contracts and Innovation

Table 2 summarizes the correlations between the variables previously described. Variable

contracts and mixed contracts appear to be negatively correlated with innovation submissions,

while fixed contracts are positively correlated with ideas submissions. Female employees and

younger ones are less likely to submit ideas, while employees with greater tenure and performing

management roles are more likely to engage in innovation activities. Employees that joined the

company after the beginning of 2014 tend to innovate less than those that were already in the ranks

at the time of the merger.


24

Table 3 provides the estimation results of our tests for our main hypothesis. H1, expressed in null

form, predicts that employees rewarded with incentive contracts of different volume-intensity are

similarly likely to propose innovative ideas. We estimate the following model:

!"#$%&&%'(),+ = - + /01%234) + /56%234) + /78'%(9:;3<63<=3<) +

/?@'<$A$B) + /C13$DE3) + /F9=3) + /G6=$;) + /HI3("<3) + J (1)

We estimate model (1) using logistic regressions.14 The dropped (base) case with respect

to contract types is Variablei. Because we estimate our model using a panel data sample, all

estimations include month fixed effects and cluster standard errors at the department level. Our

results reject the null hypothesis expressed with H1, as we detail hereafter.

The coefficients reported in the two columns result, respectively, from estimations without

department fixed effects (column (1)), or with department fixed effects (column (2)). The results

are consistent with the prediction prior analytical predictions that employees rewarded with fixed

pay are more likely to pursue innovation ideas. Month fixed effects allow us to control for

seasonality of the company’s business operations, and department fixed effects allow us to control

for any department-level characteristics such as task nature, such that we exploit contract type

variation within each department to isolate the effect of contract type on the employee’s propensity

to pursue innovation ideas. After including month and department fixed effects, the significantly

positive coefficient on Fixed (b1= 0.903, p<0.05) suggests that employees under a fixed contract

are 71.1 percent more likely to submit innovation ideas compared to employees rewarded with

14 We focus our analyses on estimations using panel data. While the contract type is defined at the employee level and it does not change over time (i.e. time-invariant employee characteristic), the characteristics of the operations in our field setting (seasonality, idle months, high productivity months) are likely to influence the likelihood of innovation activity, which is our main response variable of interest. Nonetheless, in section 7.4. we clarify that estimations at the employee cross-sectional levels yield equivalent results.

25

Variable contracts. Mixed contracts are also significantly more likely (69.9 percent) to submit

innovative ideas compared to Variable pay employees (b2= 0.842, p<0.10).15

In all estimations of model (1) reported in Table 3, the coefficients corresponding to our

control variables are largely intuitive and consistent across specifications. For example, we

observe that employees who reside in company dormitory facilities, or who perform a management

function are associated with a higher propensity to submit (valuable) innovation ideas. Age,

gender, and tenure, however, are not significantly associated with the likelihood to submit ideas.


6.2. Innovation Scope of Applicability

We explore the relation between the type of contract related to standard tasks and the

likelihood of observing innovation ideas with varying degrees of scope of application. We estimate

model (1) by specifying different dependent variables to account for the categories of innovation

utilized by management to classify proposed ideas. Estimation results reported in Table 4 show

that, relative to employees paid with Variable contracts, employees paid on a fixed schedule are

more willing to pursue innovation ideas that extend beyond their standard task and benefit a

broader set of constituencies within the organization.

In fact, when we estimate model (1) using Sub_Xi,t (where X is the category of innovation),

the coefficient associated with Fixedi is significantly positive with respect to all broad scope

innovation ideas (i.e. Sub_lti,t, Sub_groupi,t, Sub_difdepi,t, Sub_costi,t, Sub_Techi,t).16 Relative to

Variable contracts, Mixed contracts are also significantly associated with a higher of submission

15 The coefficients are log of the odd ratio. To interpret the coefficients, we transform the log of the odds back to a probability: p = exp(0.903)/(1+exp(0.903)) = .711; p = exp(0.842)/(1+exp(0.842)) = .699 16 We are not able to estimate our models when the dependent variable is Sub_morale. The reason is that there is no variation in the type of contract associated with proponents of these types of innovation ideas. More specifically, upon investigating the data, we observe that Fixed perfectly predicts Sub_morale. In other words, in our setting, innovation idea submissions related to improving morale are proposed uniquely by employees under fixed contracts.

26

of broadly defined ideas benefiting the firm in the long term (Sub_lti,t) and ideas benefiting a team

or a group (Sub_groupi,t). Taken together, these results suggest that fixed contracts can motivate

employees to engage in more pro-organizational behavior, instead of instilling a narrow focus on

the performance measure that relates specifically to the standard task included in their explicit

incentive contract. Additionally, the relation between compensation and task-specific innovation

productivity does not seem to be influenced by the structure of the incentive contract.

These results have significant implications for incentive contract design choices at firms

where innovation critically hinges on hands-on knowledge of front-line employees who are

directly engaged with the core operations. In broader sense, our findings suggest that contract

design choices can mitigate incentive problems arising from limited availability of performance

measures, as it is the case with respect to unplanned innovation activities. In particular, the income

smoothing effect of fixed incentive contracts can reduce employee uncertainty. Thereby, we show

that fixed contracts can provide a means to empower employees with “organizational slack” that

can motivate desired behaviors beyond the performance measures outlined in the contract.


7. Supplemental Tests

7.1. Endogeneity in Contract Assignment

A potential concern that may limit the interpretation of our empirical results is that the

assignment of different incentive contract types may be informed by the employee’s propensity to

produce innovative ideas. For example, if management assigns fixed contracts to employees that

are likely to be more engaged in innovation-related activities, it would be difficult to disentangle

whether the documented effects are due to the contract type or management’s screening of different

employee types. Immediately prior to our sample period, the company underwent a merger, which

27

resulted in a significant change in the composition of the top management team. Interviews with

management revealed a preference for offering fixed contracts where possible as part of the

company’s initiative to improve employee well-being, and provide them with a more stable income

stream. Nearly all newly hired employees were assigned fixed or mixed contracts.17 However, the

contracts for employees that had joined the company before the merger remained the same,

regardless of their individual contractual form.18 Approximately 72 percent of the employees in

our sample were hired after the merger as evidenced by the descriptive statistic of JoinAfterMergeri

(see Table 1, Panel C).

To rule out the possibility that contract assignment could be endogenous to particular

employee characteristics correlated with innovation propensity, we conducted additional analyses.

First, we confirmed with management during interviews that contract assignment decisions are not

deliberately based on employees’ potential for innovation. Specifically, management emphasized

that the location of our study is a manufacturing site employing workers with relatively low levels

of education, and their main responsibilities involve tasks that are fairly standard and non-

innovation related. Moreover, they indicated that the variation observed in the structure of contract

types depends on the timing of the hire during the year cycle and on the characteristics of the jobs,

as we explain later in this section. Second, we estimated a determinants model to identify and

control for potential drivers of contract assignment that might be also correlated with innovation

activities (section 7.1.1). Third, we re-estimate our main analysis using an instrumental variable

approach (section 7.1.2)

7.1.1 Contract Type Determinant Model

17 A small number of employees were still hired based on variable contracts. However, such hires are confined to departments where the industry norm is that workers are paid based on piece rates. 18 The merger event was a friendly merger, and there were no drastic changes in the company’s firm operations. The only effective change resulting from the merger was a change in ownership.

28

In section 6.1 we discussed the results of our statistical estimation of model (1), which

described the relation between contract and individual characteristics and the likelihood of

submitting innovation ideas in a given month. Our results, reported in Table 3, indicated

JoinAfterMerger, Mgmt, and DormEmp as significant predictors of employee-initiated innovation

activity. To explore the possibility that these characteristics are also driving the assignment of

contracts of different structure, we estimate the following determinant model:

K'(;<DL;IMB3) = - + /08'%(9:;3<63<=3<) + /5@'<$A$B) + /713$DE3) +

/?9=3) + /C6=$;) + /FI3("<3) + J (2)

where the dependent variable ContractTypei is a categorical variable that assumes value one if the

contract is a Variable contract, two if the contract is a Mixed contract, and three if the contract is

a Fixed contract. All other variables are defined as previously described.

We estimate the determinant model (2) using ordered logit regressions. We include

department fixed effects and we cluster the standard errors at the department level. Table 5 reports

the estimation results. Consistent with the policy described by the new management of the site, the

coefficient on JoinAfterMergeri is significantly positive. Moreover, the significantly positive

coefficient on Mgmti indicates that employees with manager functions are more likely to be

awarded fixed pay contracts. This is not surprising, given that Variable contracts generally assume

the availability of accurate performance measures that can account for the agent’s output.19 Front-

line employees at this company are assigned standard execution tasks for which output is readily

measurable. In contrast, management performance is difficult to assess based on volume-based

performance measures, such that the incentive contracts of employees with manager functions is

19 Variable contracts (or the variable portion of mixed contracts) in our settings exhibit a structure similar to a piece-rate contract. Employees are rewarded based on a fixed amount per unit of output.

29

more likely to include a fixed component. None of the other characteristics associated with

individual employees appear to be significant determinants of the contract type.


Given the significance of the variable JoinAfterMerger in the determination of the contract

type, and that a large portion of our sample was hired under the new management, we perform an

additional analysis to explore the incremental effect of contract type on the propensity to submit

innovation ideas among employees that were hired after the merger. We estimate the following

model:

!"#$%&&%'(),+ = - + /01%234) ∗ 8'%(9:;3<63<=3<) + /56%234) ∗ 8'%(9:;3<63<=3<) +

/71%234) + /?6%234) + /C8'%(9:;3<63<=3<) + /F@'<$A$B) + /G13$DE3) + /H9=3) +

/O6=$;) + /0PI3("<3) + J (3)

Estimation result are reported in Table 6 and indicate that even under the new hiring policy

enacted by management after the merger, contract type is a significant driver of submission of

employee-initiated innovations, whereby employees that joined after the merger and are rewarded

with fixed pay contracts or mixed contracts with respect to their standard tasks are more likely to

submit innovation ideas than employees rewarded with variable contracts.


7.1.2 Instrumental Variables

As described before, the operations of the company are subject to significant seasonality

and, due to industry and local norms, stop their manufacturing functions in certain months of the

year (idle months). Based on our interviews with firm representatives, management resorts to

offering Variable contracts to attract workers during the busiest months of the annual production

30

cycle, whereas, during months when the site only runs supporting functions, while main production

functions are shut down, managers are more likely to offer Fixed contracts to prospective hires.

We exploit the variation in hiring policy within the year to implement an instrumental

variable (IV) approach, by which we estimate innovation behavior using a two-stage least square

(2SLS) framework. To qualify as proper instruments, variables need to satisfy a validity

requirement by being correlated with the endogenous regressors - the contract type – and an

exclusion restriction requirement, by being uncorrelated with the error terms of innovation

behavior regressions. We construct three instrumental variables and show that they satisfy the

validity requirement and the exclusion restriction.

First, we adopt as an instrument the hiring period during the annual production cycle. That

is – we construct an ordinal variable (JoinPeriod) that assumes value -1 if the month in which

employee i was hired falls within the busy months of the year, value +1 if the month of hire falls

within the idle months of the annual cycle, and 0 if the employee is hired during months of normal

production. Table 7, Panel A, reports our estimation results.

In the first stage (column 1), we estimate contract determinant model (model (2)) including

JoinPeriod as a predictor, and we show that JoinPeriod satisfies the validity requirement, as the

associated coefficient is positive and significant (b = 0.460, p<0.01), confirming that employees

that join the firm during idle production times are more likely to be offered a Fixed contract, and

that employees joining during busy months are more likely to be offered a Variable contract (recall

that the dependent variable ContractType assumes value 1 if the contract is Variable, 2 if the

contract is Mixed, and 3 if the contract is Fixed). The second stage (columns 2 and 3), reports the

estimation of model (1), which predicts the likelihood of innovation ideas submission. In this

estimation, the value of the predictors related to the contract type is substituted by the fitted value

31

from the first stage regression. We continue to find evidence that employees rewarded with Fixed

or Mixed contracts are more likely to submit innovation ideas compared to Variable contract

employees. However, our results are not as strong in presence of department fixed effects, likely

due to the fact that different department are more likely to hire in different periods of the annual

cycle. For our instrument to be a valid one, we need to show that, in our context, the timing of

joining the firm is not correlated with the employee’s innovation behavior. In columns 4 and 5 we

provide evidence of a satisfactory exclusion restriction, by showing that JoinPeriod is not

correlated with the error term of the estimation of model (1).

To further explore these relations we repeat the 2SLS estimation using different

specifications of our instrumental variable. Specifically, we construct an indicator variable

(JoinBusy) assuming value 1 if the employee is hired during busy months and value 0 otherwise;

we also construct a similar indicator variable (JoinIdle) assuming value 1 if the employee is hired

during idle months and value 0 otherwise. Table 7, Panel B and Panel C report the estimation

results related to the use of these two instruments. We find equivalent results to those reported in

Panel A. However, the advantage of building these two additional instruments while having only

one endogenous regressor (i.e. contract type) is that it allows us to conduct an overidentification

test of whether the instruments satisfy the exclusion restriction. The Hansen-Sargan J statistic for

the over-identification test has a p-value of 0.8089, which means that we are unable to reject the

null hypothesis that both instruments are exogenous. Collectively, our evidence suggests that the

chosen instruments are not correlated with the error term of the main regressions, thereby further

satisfying the exclusion restriction.


7.2. Contract Structure and Organizational Outcomes

32

Employee-initiated innovations emerge from workers whose primary responsibility is the

execution of standard tasks for which they were selected and hired. While, as we showed, fixed

compensation contracts rewarding standard tasks are more conducive to innovation activities and,

in particular, those with broader scope of applicability, it is important to explore the relation

between the structure of the compensation contract and some key organizational outcomes. We

analyze the association between contract type and the propensity to meeting operational targets,

the incidence of production quality issues, and employee turnover. We estimate the following

model:

Q";L'$3),+ = - + /01%234) + /56%234) + /78'%(9:;3<63<=3<) +

/?@'<$A$B) + /C13$DE3) + /F9=3) + /G6=$;) + /HI3("<3) + J (4)

where the dependent variable (Outcome) is substituted by proxies measuring productivity, quality,

and turnover.

A possible negative externality associated with fixed contracts is that employees might

extract rent by operating at a reduced level of effort, as their standard payoff is independent from

their volume of production. In our setting, each month, management flags individual employees

as having met or not having met their assigned targets.20 We follow this classification and create

an indicator variable Meti,t which assumes value one if the employee i has met or exceeded its

budgeted output in month t, and zero otherwise. Our goal is to examine whether employees,

especially those under a Fixed contract, are more prone to undermining their performance on their

standard task.

20 In our setting, targets are agreed upon at the beginning of the year and allocated to each month of activity. Given the high predictability of the revenue stream and the order volume for the firm, renegotiations of monthly targets are extremely rare.

33

Employee rent extraction might also occur in the form of lower quality of work. Product

quality has great influence on firm growth, profitability, and sustainability. Bad quality has

typically very negative and lasting effects on customer satisfaction and retention. In our setting

management monitors employee contribution to production quality by tracing quality defects or

complaints to the employees that participated to the production process in which the quality issue

was generated. Employees are flagged in the company’s control system every time a complaint is

filed. We construct an indicator variable (BadQualityi,t), which assumes value 1 if employee i is

flagged for bad quality in month t and zero otherwise.

Finally, we relate contract structure to turnover dynamics. Employee retention is a priority

for most organizations, and the costs associated with high turnover are typically significant.

Variable contracts provide workers with the least income predictability and with the highest

pressure to deliver on the standard task. Fixed contracts, on the other hand, provide relatively

greater job security and financial stability. We construct an indicator variable (Departurei,t)

assuming value 1 if employee i leaves the firm in month t, and zero if the employee continues their

employment in that month.

Table 5 reports the logit estimations of model (4) respectively for each dependent variable

corresponding to the organizational outcomes of interest. The results show no significant

differences across contract types with respect to each organizational outcome, with the exception

of a positive relation between mixed contracts and higher incidences of quality issues. We

conclude that offering fixed contracts might be superior to other contract structures, in that fixed

pay is associated with greater innovation and no particular downsides, at least with respect to the

limited list of organizational outcomes considered in this study.


34

7.3. Subsample Test: Departments with Contract Type Variation

In order to sharpen our empirical tests that estimate the effect of contract types on the

propensity to pursue innovation ideas, we conduct robustness test that confine our analyses only

to departments that exhibit contract type variation.24 For some departments, the industry practice

is such that incentive contracts follow one particular form. We exclude such departments, and

estimate our model (1) using a sample restricted to only on the departments where management

discretion to award different types of contracts is not limited by industry norms. The results of our

estimations are tabulated in Table 9, and are qualitatively similar to our main results in Table 3

and offer further support to reject the null with respect to our main hypothesis H1.


7.4. Cross-sectional Test

As noted previously, in our setting, contract type constitutes an employee-fixed

characteristic as employees are assigned incentive contracts at the time of hire, and the contract is

not subject to changes over the course of their employment with the firm (or, at least, we do not

observe any changes in contract structure at the individual employee level during our sample

period). This feature prevents us from estimating a difference-in-differences model specification

where we can estimate more precisely a causal effect between contract type and innovation

activities based on employees that experienced a change in contract type. Instead, our results are

rather based on associations stemming from the analysis of cross-sectional variation in contract

structure – i.e. whether employees with fixed contracts are associated with a higher number of

(high-quality) innovation idea submissions.

24 Note that our main analyses already include department fixed effects, and thus, already account for differences at the department level such as task nature that may affect our results.

35

Our main analyses reported in Table 3 are based on a panel sample. In other words, we

retain the monthly observations for each employee, and thus, the level of analysis is at the

employee-month level. In order to match the level of analysis of the independent variable of

interest (i.e. contract type) with the dependent variable, we conduct a cross-sectional analysis at

the employee-level (pooled sample). To do so, we define new employee-level dependent variables

for innovation-related activities. Precisely, we define SubmissionEi as an indicator that assumes

value one if the employee has ever submitted an innovation idea in our sample period, and zero

otherwise. Similarly, ApprovedEi is defined as an indicator variable that assumes value one if the

employee ever received a bonus on a submitted innovation idea in our sample time period, and

zero otherwise. The results of the estimation of model (1) with these specifications are tabulated

in Table 10, and are qualitatively similar to our main results.


8. Conclusion

In this study, we empirically examine the relationship between the design of incentive contracts

and employee-initiated innovation activities. Prior analytical work proposes that low-powered

incentives can induce employees to expend effort pursuing innovation-related activities, and

especially activities that benefit a broader set of constituents within the organization rather than

the individual employee. A fundamental premise in these models is that employees face an effort

allocation decision involving a standard task, for which performance measures are readily

available and contractible ex-ante, and innovation-related activities that arise unplanned, are

uncertain in their value and thus require an ex-post settlement process in order to be rewarded. To

examine the question empirically, we use field data from a company that (1) encourages the

submission of employee-initiated innovation ideas and records these ideas and their outcomes in

36

their information system, and (2) exhibits contract structure variation within similar organizational

roles. The characteristics of our field setting allow us to obtain a systematic measure that can proxy

for employee-initiated innovation that may not be available otherwise and to attribute our results

to specific contract types, ruling out endogenous factors that may be correlated with the firm’s

decision to assign a particular type of incentive contract.

We find empirical evidence that is consistent with theoretical predictions. Employees

rewarded with fixed contracts for standard tasks exhibit a higher propensity to submit innovation

ideas, compared to employees earning variable pay. More importantly, in our setting, such

innovation activity is not simply a signaling game by which the employee wants to signal their

high-innovation type. In fact, innovation submissions are low in volume and the submitted idea

are predominantly considered to be valuable to the firm, as indicated by the fact that they result in

actual reward payouts in the ex-post settlement process. Furthermore, we find evidence supporting

the prediction that low-powered incentives are associated with greater propensity to engage in

ideas where the returns to the innovation are beneficial to broad subset of workers and in line with

long-term strategic goals.

Like any archival field study, this research is subject to several limitations. First, our study

is subject to external validity concerns due to the fact that we include information pertaining to

only one organization, and our results might therefore be influenced by idiosyncratic

characteristics of the field setting. Second, we are restricted to the contract types that the company

is using. The contracts used at our research site allow us to compare low-powered versus high-

powered incentive contracts. However, it may be that alternative contract types or performance

measures may be more suitable in motivating employee-initiated innovation activities. Despite

these limitations, our study sheds new light on how contract structure can mitigate contracting

37

difficulties in motivating unplanned employee-initiated innovation activities for which

performance measures are hardly available. Future research may examine some additional

unintended consequences associated with fixed contracts (e.g. effect on collaboration and joint

innovation projects), explore conditions under which fixed or variable contracts may be more

desirable, or examine the influence of different structures of ex-post bargaining (i.e. fixed bonuses

versus benefit sharing arrangements) for successful innovation activities.

38

Appendix: Types of Employee-Initiated Innovations

Category Type Description Examples Variable

Non Task-Specific Innovation Ideas (Broad Scope)

Long-term Ideas that enhance the long-term success of the company

"At this stage, our company does not have a complete proofing management standard. As a result, illegal operations often occur. We shall draft a formal proofing management standard that workers should follow."

Sub_lt

Group Ideas that promote collaboration

"Due to the building setup, the offset printing plant is now separated by the detention area of the outgoing products, resulting in poor communication and inconvenience. I hope that the outer wall can be removed so that the collaboration among the workers in the offset printing plant can be largely improved."

Sub_group

Different Department Ideas that benefit other departments

One employee from storage department suggests that "defective products in stock cannot be sold and may be used to print internal documents and labels."

Sub_diffdep

Cost Ideas that decrease overhead expenses

"There are two machines that are damaged for different reasons. We can assemble the good parts of one machine to the other. As a result, we only need to buy one new machine rather than two machines."

Sub_cost

Technology

Ideas that enhance to company’s computerized processes and automation

"The booster pump of the company's fire protection system is pressurized every 3-5 minutes due to the sensitivity of the pressure switch and the leakage of the pipeline, resulting in the pump being often damaged and the water pressure being insufficient. I suggest to add a timing device to the pump control circuit, which not only provides a higher water pressure in the pipeline, but also increases the pressurization interval to around 20 minutes."

Sub_tech

Morale Ideas that improve team/group morale

"We can celebrate office birthdays on a monthly basis. This is a way to gain employees’ sense of belongings and increase employee satisfaction." Sub_morale

Task-Specific Innovation Ideas (Narrow Scope)

5S Ideas that enhance the standardization process of the standard task

"I suggest to draw a paper diagram depicting the model, configuration and operation of the laminating machine." Sub_5s

Quality Ideas that decrease the number of bad-quality (standard task) outputs

"There is no waste disposal area between the two templates in the middle of die cutting area, which increases the probability of defective projects. I suggest to add a 3mm waste disposal area in the middle of die cutting area, so that workers can verify each product during the process."

Sub_quality

Efficiency Ideas that enhance the speed of executing the standard task

"“400 per roll” of material is currently used, resulting in too frequent machine shutdowns as materials need to be replaced. This results in wasting a lot of printing time. I suggest to order the “800 per roll” material instead."

Sub_efficiency

39

References

Amabile, T. M. 1988. A model of creativity and innovation in organizations. Research in Organizational Behavior 10:123-167.

Banker, R. D., and S. Datar. 1989. Sensitivity, precision, and linear aggregation of signals for

performance evaluation. Journal of Accounting Research 27 (1):21-39. Baranchuk, N., R. Kieschnick, and R. Moussawi. 2014. Motivating innovation in newly public

firms. Journal of Financial Economics 111 (3):578-588. Baumann, O., and N. Stieglitz. 2014. Rewarding value-creating ideas in organizations: The power

of low-powered incentives. Strategic Management Journal 35 (3):358-375. Birkinshaw, J., and L. Duke. 2013. Employee-led innovation. Business Strategy Review 24:46-51. Damanpour, F. 1991. Organizational innovation: A meta-analysis of effects of determinants and

moderators. The Academy of Management Journal 34 (3):555-590. Ederer, F. 2013. Incentives for parallel innovation. In SSRN: https://ssrn.com/abstract=2309664

or http://dx.doi.org/10.2139/ssrn.2309664. Ederer, F., and G. Manso. 2013. Is Pay for Performance Detrimental to Innovation? Management

Science 59 (7):1496-1513. Grabner, I. 2014. Incentive system design in creativity-dependent firms. The Accounting Review

89 (5):1729-1750. Grossman, S. J., and O. D. Hart. 1986. The costs and benefits of ownership: a theory of vertical

and lateral integration. Journal of Political Economy 94 (4):691-719. Hart, O. D. 1995. Firms, contracts, and financial structure. New York: Oxford University Press,

USA. Hart, O. D., and J. Moore. 1990. Property rights and the nature of the firm. Journal of Political

Economy 98 (6):1119-1158. Hart, O. D., and J. Moore. 2008. Contracts as reference points. Quarterly Journal of Economics

123 (1):1-48. Hellmann, T., and V. Thiele. 2011. Incentives and innovation: A multitasking approach. American

Economic Journal: Microeconomics 3 (1):78-128. Holmstrom, B. 1979. Moral Hazard and Observability. The Bell Journal of Economics 10 (1):74-

91.

40

Holmstrom, B. 1989. Agency costs and innovation. Journal of Economic Behavior and Organization 12:305-327.

Holmstrom, B., and P. Milgrom. 1991. Multitask principal-agent analyses: incentive contracts,

asset ownership, and job design. Journal of Law Economics & Organization 7:24-52. Holthausen, R. W., D. F. Larcker, and R. G. Sloan. 1995. Business unit innovation and the structure

of executive compensation. Journal of Accounting and Economics 19:279-313. Ittner, C. D., D. F. Larcker, and M. V. Rajan. 1997. The choice of performance measures in annual

bonus contracts. The Accounting Review 72 (2):231-255. Kachelmeier, S. J., B. E. Reichert, and M. G. Williamson. 2008. Measuring and motivating

quantity, creativity, or both. Journal of Accounting Research 46 (2):341-373. Lerner, J., and J. Wulf. 2007. Innovation and incentives: Evidence from corporate R&D. Review

of Economics and Statistics 89 (4):634-644. Li, X. 2016. Management control and employee-driven innovation. Doctoral Dissertation,

Accounting and Management Unit, Harvard University, Harvard Business School. Manso, G. 2011. Motivating innovation. The Journal of Finance 66 (5):1823-1860. Nohria, N., and R. Gulati. 1996. Is slack good or bad for innovation? The Academy of Management

Journal 39 (5):1245-1264.

41

Table 1: Descriptive Statistics

Panel A: Panel Data N mean p50 sd min p25 p75 max

Innovation-related Variables Submission 6016 0.060 0.000 0.238 0.000 0.000 0.000 1.000 Approved 6016 0.057 0.000 0.233 0.000 0.000 0.000 1.000 Sub_lt 6016 0.014 0.000 0.119 0.000 0.000 0.000 1.000 Sub_group 6016 0.009 0.000 0.094 0.000 0.000 0.000 1.000 Sub_diffdep 6016 0.003 0.000 0.055 0.000 0.000 0.000 1.000 Sub_cost 6016 0.046 0.000 0.210 0.000 0.000 0.000 1.000 Sub_tech 6016 0.005 0.000 0.068 0.000 0.000 0.000 1.000 Sub_5s 6016 0.007 0.000 0.083 0.000 0.000 0.000 1.000 Sub_quality 6016 0.007 0.000 0.085 0.000 0.000 0.000 1.000 Sub_efficiency 6016 0.017 0.000 0.129 0.000 0.000 0.000 1.000 Sub_morale 6016 0.002 0.000 0.041 0.000 0.000 0.000 1.000

Panel B: Cross-Sectional Data

N mean p50 sd min p25 p75 max Innovation-related Variables Submission 513 0.154 0.000 0.361 0.000 0.000 0.000 1.000 Approved 513 0.152 0.000 0.359 0.000 0.000 0.000 1.000 Sub_lt 513 0.014 0.000 0.116 0.000 0.000 0.000 1.000 Sub_group 513 0.031 0.000 0.174 0.000 0.000 0.000 1.000 Sub_diffdep 513 0.023 0.000 0.151 0.000 0.000 0.000 1.000 Sub_cost 513 0.109 0.000 0.312 0.000 0.000 0.000 1.000 Sub_tech 513 0.031 0.000 0.174 0.000 0.000 0.000 1.000 Sub_5s 513 0.047 0.000 0.211 0.000 0.000 0.000 1.000 Sub_quality 513 0.049 0.000 0.216 0.000 0.000 0.000 1.000 Sub_efficiency 513 0.078 0.000 0.268 0.000 0.000 0.000 1.000 Sub_morale 513 0.012 0.000 0.108 0.000 0.000 0.000 1.000

42

Panel C: Employee-Level Data Contract-related Variables Variable 513 0.127 0.000 0.333 0.000 0.000 0.000 1.000 Mixed 513 0.296 0.000 0.457 0.000 0.000 1.000 1.000 Fixed 513 0.577 1.000 0.495 0.000 0.000 1.000 1.000 Employee Characteristics JoinAfterMerger 513 0.719 1.000 0.450 0.000 0.000 1.000 1.000 DormEmp 513 0.068 0.000 0.252 0.000 0.000 0.000 1.000 Female 513 0.382 0.000 0.486 0.000 0.000 1.000 1.000 Age 510 33.081 31.354 10.620 16.000 24.375 41.059 66.720 Mgmt 513 0.076 0.000 0.265 0.000 0.000 0.000 1.000 Tenure 513 1.816 1.000 1.319 1.000 1.000 2.091 17.000

Notes: Table 1 reports the summary statistics for all variables used in the empirical tests. All variables are defined in Section 5.2. Panel A reports the descriptive statistics corresponding to our complete panel data sample. In Panel A, Innovation-related variables are defined as indicator variables assuming value 1 if the employee has submitted at least one innovation idea of the indicated kinds during the month zero otherwise, and the indicator variable Approved assumes value 1 if any idea submitted by the employee has been approved during the month and zero otherwise. Panel B reports the descriptive statistics relative to our data collapsed to the cross-sectional employee-level. In Panel B, Innovation-related variables are defined as indicator variables assuming value 1 if the employee has submitted at least one innovation idea of the indicated kinds during our sample period and zero otherwise, and the indicator variable Approved assumes value 1 if any idea submitted by the employee has been approved during our sample period and zero otherwise. Panel C reports the descriptive statistics related to employee characteristics, including their contract type and demographic information.

43

Table 2: Correlations 1 2 3 4 5 6 7 8 9 1. Submission 1.0000

2. Sub_lt 0.4780*** 1.0000

3. Sub _group 0.3756*** 0.4609*** 1.0000

4. Sub _diffdep 0.2162*** 0.1718*** 0.2207*** 1.0000

5. Sub _cost 0.8703*** 0.5427*** 0.4064*** 0.2050*** 1.0000

6. Sub _tech 0.2699*** -0.0083 0.0453*** -0.0037 0.1591*** 1.0000

7. Sub _5s 0.3309*** -0.0102 0.0555*** -0.0046 0.0669*** 0.1996*** 1.0000

8. Sub _quality 0.3387*** -0.0104 0.0952*** 0.0667*** 0.1944*** 0.1948*** 0.1100*** 1.0000

9. Sub _efficiency 0.5183*** 0.0057 0.0558*** 0.0400*** 0.4180*** 0.1802*** 0.0972*** 0.3363*** 1.0000 10. Sub _morale 0.1610*** -0.0049 0.0826*** -0.0022 0.1656*** 0.0572*** -0.0034 -0.0035 0.0579*** 11. Variable -0.1068*** -0.0696*** -0.0547*** -0.0244* -0.1011*** -0.0167 -0.0251* -0.0132 -0.0337*** 12. Mix -0.0449*** 0.0482*** 0.0080 -0.0129 -0.0404*** -0.0349*** -0.0083 -0.0149 -0.0543*** 13. fixed 0.1292*** 0.0212 0.0409*** 0.0317** 0.1205*** 0.0429*** 0.0285** 0.0236* 0.0734*** 14. JoinAfterMerger -0.0951*** -0.0746*** -0.0540*** -0.0148 -0.0834*** -0.0348*** -0.0467*** -0.0336*** -0.0484*** 15. DormEmp 0.0007 0.0721*** 0.0182 0.0020 0.0115 -0.0020 -0.0135 0.0068 -0.0166 16. Female -0.0244* 0.0433*** 0.0267** 0.0050 0.0120 -0.0458*** -0.0562*** -0.0223* -0.0336*** 17. Age -0.0474*** -0.0246* -0.0248* -0.0348*** -0.0562*** 0.0235* 0.0637*** -0.0202 -0.0550*** 18. Mgmt 0.2293*** 0.1255*** 0.0848*** 0.0577*** 0.1897*** 0.0507*** 0.0941*** 0.0450*** 0.1233*** 19. Tenure 0.0617*** 0.0772*** 0.0328** 0.0071 0.0689*** 0.0033 0.0150 0.0035 0.0223*

10 11 12 13 14 15 16 17 18 10. Sub _morale 1.0000

11. Variable -0.0234* 1.0000

12. Mix -0.0208 -0.2933*** 1.0000

13. fixed 0.0373*** -0.6289*** -0.5588*** 1.0000

14. JoinAfterMerger -0.0301** -0.4520*** 0.1929*** 0.2352*** 1.0000

15. DormEmp -0.0174 0.4595*** 0.0608*** -0.4481*** -0.3817*** 1.0000

16. Female -0.0250* 0.2091*** -0.0849*** -0.1123*** -0.0565*** -0.0957*** 1.0000

17. Age 0.0038 0.3157*** -0.0132 -0.2632*** -0.2289*** 0.1409*** 0.1606*** 1.0000

18. Mgmt -0.0044 -0.1914*** -0.1633*** 0.2988*** -0.1850*** -0.1243*** -0.1839*** 0.0287** 1.0000 19. Tenure 0.0071 0.3751*** -0.1638*** -0.1921*** -0.7691*** 0.3827*** 0.0079 0.2816*** 0.1971***

44

Table 3: Contract Type and Innovation Activities

(1) (2) Submission Submission

Fixed 2.585*** 0.903* (3.97) (1.95) Mix 1.538 0.842* (1.48) (1.74) JoinAfterMerger -0.879* -0.546 (-1.78) (-1.17) DormEmp 1.596** 1.120** (1.98) (2.51) Female 0.356 -0.195 (0.78) (-0.73) Age -0.020 -0.011 (-1.12) (-0.54) Mgmt 1.487*** 0.899** (3.36) (2.15) Tenure -0.016 -0.120 (-0.27) (-1.26) Intercept -5.302*** -5.588*** (-6.57) (-8.11) N 5833 5833 pseudo R-sq 0.180 0.321 Month FE Yes Yes Department FE No Yes

Notes: Table 3 reports the coefficients estimated for model (1) using logit regression. Variable is the base (dropped) case. Estimations in columns (2) include department fixed effects. All estimations include month fixed effects and are performed clustering standard errors at the department level. Two-tailed statistical significance is indicated as follows: * = p<0.10; ** = p<0.05; *** = p<0.01.

45

Table 4: Contract Type and Innovation Types Broad Scope Innovations Narrow Scope Innovations

(1) (2) (3) (4) (5) (6) (7) (8) Sub_lt Sub_group Sub_diffdep Sub_cost Sub_tech Sub_5s Sub_quality Sub_efficiency Fixed 10.926*** 13.739*** 12.366*** 1.575* 1.647** 0.827 0.612 -0.466

(5.33) (8.53) (10.66) (1.70) (2.46) (1.09) (0.71) (-1.04) Mixed 13.622*** 15.373*** 0.098 1.213 0.000 1.415*** 0.270 -0.682

(4.49) (12.43) (0.09) (1.62) (.) (4.15) (0.48) (-1.15) JoinAfterMerger 2.833* -1.359* 1.091* -0.271 -1.015 -1.585 -1.360* -0.761**

(1.90) (-1.86) (1.69) (-0.48) (-1.08) (-1.04) (-1.65) (-2.55) DormEmp 0.000 -1.934 14.918*** 1.911** 0.767 -0.542 0.637 0.022

(.) (-1.23) (11.05) (2.29) (0.69) (-1.47) (1.06) (0.06) Female -0.124 0.555 -0.745 0.054 -1.237 -2.359** -0.155 -0.118

(-0.07) (0.67) (-0.66) (0.17) (-1.25) (-1.97) (-0.26) (-0.46) Age -0.161** -0.031 -0.091* -0.013 0.052** 0.076** -0.003 -0.040

(-2.03) (-1.29) (-1.70) (-0.62) (2.12) (2.12) (-0.08) (-1.61) Mgmt 0.207 -0.227 1.615 0.579 0.411 1.804** 0.463 1.219**

(0.10) (-0.63) (1.09) (1.54) (0.47) (2.57) (0.53) (2.48) Tenure 0.494* -0.216 0.086 -0.058 -0.105 -0.224 -0.229 -0.118 (1.87) (-0.98) (1.03) (-0.73) (-0.22) (-0.45) (-1.28) (-1.24) Intercept -18.675*** -20.449*** -14.615*** -9.064*** -5.537*** -7.320*** -4.873*** -3.765*** (-8.44) (-11.08) (-6.68) (-11.86) (-7.03) (-9.24) (-3.43) (-4.24) N 1029 1887 546 5656 1185 2849 3146 4571 pseudo R-sq 0.765 0.448 0.129 0.410 0.131 0.294 0.129 0.276 Month FE Yes Yes Yes Yes Yes Yes Yes Yes Department FE Yes Yes Yes Yes Yes Yes Yes Yes

Notes: Table 4 reports the coefficients estimated for model (1) using the propensity to produce innovation ideas (Panel A) for individual categories of innovation. All estimations include month and department fixed effects, and are performed clustering standard errors at the department level. Two-tailed statistical significance is indicated as follows: * = p<0.10; ** = p<0.05; *** = p<0.01.

46

Table 5: Determinants of Contract Type

ContractType JoinAfterMerger 2.028***

(2.68) DormEmp -1.987

(-1.50) Female 0.247

(0.64) Age -0.018

(-0.97) Mgmt 3.290*** (2.97) Tenure -0.016 (-0.12) N 509 pseudo R-sq 0.397 Department FE Yes

Notes: Table 5 reports the coefficients of model (2) estimated using ordered logit regression. The dependent variable ContractType is a categorical variable assuming value 1 if the contract is Variable, 2 if the contract is Mixed, and 3 if the contract is Fixed. The estimation includes department fixed effects and is performed using standard errors clustered at the department level. Two-tailed statistical significance is indicated as follows: * = p<0.10; ** = p<0.05; *** = p<0.01.

47

Table 6: Effects of Being Hired by New Management

(1) (2) Submission Submission Fixed * JoinAfterMerger 12.510*** 12.863***

(13.96) (17.33) Mix * JoinAfterMerger 9.879*** 9.530***

(7.65) (8.64) Fixed 2.117*** 0.124

(3.62) (0.30) Mixed 1.844* 1.490***

(1.80) (3.90) JoinAfterMerger -13.244*** -13.253***

(-14.88) (-14.72) DormEmp 1.095 0.452

(1.51) (1.27) Female 0.305 -0.327

(0.67) (-1.14) Age -0.023 -0.016

(-1.19) (-0.75) Mgmt 1.510*** 0.989**

(3.43) (2.28) Tenure -0.031 -0.167

(-0.55) (-1.30) Intercept -4.711*** -4.692*** (-5.77) (-6.29) N 5833 5833 pseudo R-sq 0.188 0.332 Month FE Yes Yes Department FE No Yes

Notes: Table 6 reports the coefficients estimated for model (3) using logit regression. The coefficients associated with the interaction terms Fixed*JoinAfterMerger and Mix*JoinAfterMerger represent the incremental effect of contract assignment on innovarion propensity controlling for potential systematic changes in the hiring practices by the new management of the company. Estimations in column (2) include department fixed effects. All estimations include month fixed effects and are performed clustering standard errors at the department level. Two-tailed statistical significance is indicated as follows: * = p<0.10; ** = p<0.05; *** = p<0.01.

48

Table 7: Instrumental Variable Tests

Panel A: Instrument JoinPeriod

First Stage Second Stage Exclusion Restriction (1) (2) (3) (4) (5)

DV ContractType Submission Submission Submission Submission Fixed

4.101*** 0.682 2.561*** 0.913* (4.29) (0.51) (4.06) (1.95)

Mixed

2.700*** 0.542 1.533 0.877* (4.10) (0.70) (1.46) (1.75)

JoinAfterMerger 1.963**

-1.053** -0.593 (2.55)

(-1.97) (-1.41)

DormEmp -1.813 2.109*** 0.611 1.788* 1.179** (-1.41) (4.86) (1.28) (1.95) (2.35)

Female 0.301 0.558*** -0.072 0.390 -0.176 (0.67) (4.04) (-0.45) (0.83) (-0.67)

Age -0.017 0.013 -0.002 -0.017 -0.010 (-0.91) (1.52) (-0.37) (-0.90) (-0.53)

Mgmt 3.335*** 0.038 0.460** 1.519*** 0.907** (2.74) (0.14) (2.34) (3.55) (2.22)

Tenure 0.005 0.244*** 0.032 -0.019 -0.116 (0.04) (4.41) (0.35) (-0.32) (-1.26)

JoinPeriod 0.460***

0.429 0.108 (3.30)

(1.30) (0.41)

Intercept

-6.480*** -3.303*** -5.260*** -5.594*** (-5.61) (-2.95) (-6.10) (-8.09)

N 509 5866 5833 5833 5833 pseudo R-sq 0.400

0.188 0.321

Month FE

Yes Yes Yes Yes Department FE Yes No Yes No Yes

49

Panel B: Instrument JoinBusy


DV ContractType Submission Submission ContractType Submission Fixed

3.535*** 0.838 2.599*** 0.918**

(2.78) (0.62) (3.97) (1.98) Mixed

2.313*** 0.628 1.552 0.871*

(2.65) (0.81) (1.46) (1.74) JoinAfterMerger 1.951**

-1.107** -0.628

(2.49)

(-2.04) (-1.47) DormEmp -1.889 1.856*** 0.663 1.752** 1.180**

(-1.44) (3.28) (1.39) (1.98) (2.46) Female 0.283 0.490*** -0.056 0.377 -0.174

(0.66) (2.81) (-0.35) (0.81) (-0.66) Age -0.017 0.008 -0.002 -0.019 -0.010

(-0.90) (0.75) (-0.28) (-1.01) (-0.53) Mgmt 3.316*** 0.189 0.440** 1.489*** 0.899**

(2.81) (0.55) (2.23) (3.47) (2.15) Tenure 0.010 0.217*** 0.043 0.001 -0.109

(0.08) (3.03) (0.46) (0.01) (-1.15) JoinBusy -0.434***

-0.517 -0.187

(-3.28)

(-1.11) (-0.51) Intercept

-5.807*** -3.428*** -5.141*** -5.543***

(-3.82) (-3.05) (-6.96) (-7.81) N 509 5866 5833 5833 5833 pseudo R-sq 0.400

0.185 0.321

Month FE


50

Panel C: Instrument JoinIdle


DV ContractType Submission Submission ContractType Submission Fixed

4.944*** 0.216 2.523*** 0.903*

(4.71) (0.08) (4.06) (1.94) Mixed

3.286*** 0.276 1.507 0.849*

(4.54) (0.18) (1.46) (1.79) JoinAfterMerger 2.055***

-0.859 -0.546

(2.81)

(-1.62) (-1.17) DormEmp -1.856 2.467*** 0.453 1.716* 1.130**

(-1.45) (5.23) (0.50) (1.91) (2.30) Female 0.277 0.661*** -0.119 0.389 -0.192

(0.66) (4.40) (-0.42) (0.83) (-0.72) Age -0.019 0.020** -0.004 -0.017 -0.010

(-1.02) (2.13) (-0.39) (-0.90) (-0.55) Mgmt 3.316*** -0.198 0.523 1.544*** 0.902**

(2.86) (-0.68) (1.40) (3.48) (2.24) Tenure -0.031 0.298*** 0.001 -0.045 -0.121

(-0.20) (4.73) (0.01) (-0.92) (-1.30) JoinIdle 0.856**

0.824 0.044

(2.12)

(1.39) (0.08) Intercept

-7.482*** -2.927 -5.481*** -5.601***

(-5.91) (-1.34) (-5.45) (-8.01) N 509 5866 5833 5833 5833 pseudo R-sq 0.401

0.188 0.321

Month FE


Notes: Table 70 reports the coefficients of the 2SLS estimation of model (1). In Panel A we report estimations where we adopt as instrument JoinPeriod, an ordinal variable assuming value -1 if the period in which employee i is hired is a busy month, value +1 if the period is idle, and 0 if it is a period of regular production. In Panel B we report estimations where we adopt as instrument JoinBusy, an indicator variable assuming value 1 if the period in which employee i is hired is a busy month, and 0 otherwise. In Panel C we report estimations where we adopt as instrument JoinIdle, an indicator variable assuming value 1 if the period in which employee i is hired is a month with idle production, and 0 otherwise. The Hansen-Sargan J statistic for the over-identification test considering JoinBusy and JoinIdle as instruments for the estimation of our relation has a p-value of 0.8089, which means that we are unable to reject the null hypothesis that both instruments are exogenous.

51

Table 8: Effect on Organizational Outcomes

(1) (2) (3) Met BadQuality Departure Fixed 0.045 0.554 0.771

(0.19) (1.42) (1.64) Mixed -0.295 0.619** 0.091

(-1.59) (2.09) (0.16) JoinAfterMerger 0.173 -0.007 -2.035

(0.46) (-0.02) (-0.98) DormEmp 0.354 0.206 -0.418

(0.88) (0.98) (-0.66) Female 0.181* -0.576 0.287

(1.76) (-1.12) (0.97) Age -0.006 0.008 -0.031*

(-0.43) (0.90) (-1.71) Mgmt 0.282 -0.061 -0.751**

(1.40) (-0.10) (-2.37) Tenure -0.165 0.057 -1.186 (-1.04) (0.39) (-1.13) Intercept -4.002*** -3.116*** 5.377* (-3.58) (-5.08) (1.68) N 5799 5672 509 pseudo R-sq 0.072 0.139 0.172 Department FE Yes Yes Yes

Notes: Table 8 reports the coefficients of model (4) estimated using ordered logit regression and adopting different dependent variables representing important organizational outcomes. Respectively, column (1) reports the estimates of model (2) using the dependent variable Met, an indicator variable assuming value 1 if employee i met or exceeded her assigned target in month t and zero if the employee i missed the target; Column (2) reports the estimates of model (4) using the dependent variable BadQuality, an indicator variable assuming value 1 if the activity for which employee i is responsible was associated with a quality complaint in month t and zero otherwise. Column (3) reports the estimates of model (4) using the dependent variable Departure, an indicator variable assuming value 1 if employee i left the company in month t and zero if employee i was still on staff in month t. All estimations include department fixed effects and is performed using standard errors clustered at the department level. Two-tailed statistical significance is indicated as follows: * = p<0.10; ** = p<0.05; *** = p<0.01.

52

Table 9: Robustness Test: Restricting to Departments with Contract Variation

(1) (2) Submission Submission Fixed 1.888*** 0.871*

(3.68) (1.76) Mixed 0.384 0.546

(0.89) (1.44) JoinAfterMerger -1.548*** -1.376***

(-6.18) (-4.73) DormEmp 0.557 0.798**

(0.97) (2.32) Female -0.134 -0.217

(-0.22) (-0.55) Age -0.017 -0.014

(-0.50) (-0.53) Mgmt 1.477*** 1.272**

(2.60) (2.07) Tenure -0.082 -0.251 (-1.01) (-1.22) Intercept -4.428*** -4.780*** (-7.08) (-6.83) N 4237 4237 pseudo R-sq 0.207 0.278 Month FE Yes Yes Department FE No Yes

Notes: Table 9 reports the coefficients estimated for model (1) using logit regression restricted to a sample corresponding only to departments that exhibited variation in the contract structure offered to their employees. Estimations in column (2) include department fixed effects. All estimations include month fixed effects and are performed clustering standard errors at the department level. Two-tailed statistical significance is indicated as follows: * = p<0.10; ** = p<0.05; *** = p<0.01.

53

Table 10: Cross-sectional Test

(1) (2)

SubmissionE SubmissionE Fixed 1.941*** 1.317**

(4.04) (2.45) Mixed 1.001* 0.955

(1.74) (1.48) JoinAfterMerger -1.253** -0.979

(-2.06) (-1.40) DormEmp 1.229 1.490

(1.64) (1.49) Female -0.434 -0.387

(-1.00) (-1.08) Age -0.006 -0.002

(-0.32) (-0.11) Mgmt 1.609*** 1.560**

(3.08) (2.35) Tenure 0.128 0.078 (0.37) (0.17) Intercept -2.660* -3.375** (-1.92) (-2.38) N 509 506 pseudo R-sq 0.206 0.267 Department FE No Yes

Notes: Table 10 reports the coefficients estimated for model (1) using logit regression on the cross-section of employees in our sample. The unit of analysis in this estimation is the employee (as opposed to the employee-month observation that was adopted as the unit of analysis in Table 3). The model is estimated adopting the indicator variable SubmissionE, assuming value 1 if employee i has ever proposed an innovation idea and zero otherwise, as the dependent variable. Estimations in column (2) include department fixed effects. All estimations are performed clustering standard errors at the department level. Two-tailed statistical significance is indicated as follows: * = p<0.10; ** = p<0.05; *** = p<0.01.

incentive contracts and employee-initiated innovation ...first, employee-driven innovation...

Documents