The Schumpeterian Cost of Regulation on Entry and Innovation:

The Case of Bail Bonds

Erin L Scott National University of Singapore

elscott@nus.edu.sg

Anne Marie KnottWashington University in St. Louis

knott@wustl.edu

November 22, 2013

These results are preliminary. Please do not cite or quote without permission

_____________________Special thanks to members of the Professional Bail Agents of the United States for sharing their time and industry insights. This paper benefits from comments from seminar participants at Washington University in St. Louis, CCC at Massachusetts Institute of Technology, the Entrepreneurship/Innovation Seminar at UCLA, the Oliver E. Williamson Seminar on Institutional Analysis at UC Berkeley, the TIES seminar at Massachusetts Institute of Technology, the Atlanta Competitive Advantage Conference, AB Freeman School at Tulane, the ARISE workshop at the National University of Singapore and the strategy seminar at HEC Paris. This research was funded in part by the Ewing Marion Kauffman Foundation. All errors remain the authors.

The Schumpeterian Cost of Regulation on Entry and Innovation:

The Case of Bail Bonds

Entrepreneurship is valued in part because it stimulates Schumpeter’s gale of creative destruction. A popular view, reinforced by recent empirical work on regulatory reform, is that regulation inhibits entrepreneurship and innovation. However dismantling regulation is ill-advised, since generally regulation has primary goals other than to increase innovation. We study the impact of regulation on entry and innovation in such a setting, the bail bond industry. We build structural models of the entrepreneur’s entry decision and the incumbent’s continuation decision, then employ recovered parameter estimates in counterfactual policy environments that allow us to compare the impact of different regulatory regimes. Results from this exercise indicate that a combination of entry and operating regulations reduces entry yet increases innovation relative to a setting with no regulation. These benefits are above and beyond the intended benefits of the regulations. Our results suggest it is possible to design regulations without imposing the Schumpeterian cost on innovation, and provide preliminary indications of how to accomplish this. Moreover our results suggest that Schumpeterian cost itself may need to be redefined. Entry and innovation actually oppose one another in our setting. Not only does excess entry steal share, in so doing it also suppresses incentives to innovate.

2

1. Introduction

Entrepreneurship is valued in part because it stimulates Schumpeter’s gale of creative destruction. In

addition to bringing forth “the new commodity, the new technology, the new source of supply, the new

type of organization” (Schumpeter 1947: 84), entrepreneurial entry “strikes…existing firms…at their

foundations and their very lives” and thereby triggers two forms of incumbent response: an immediate

response of lowering prices, and often a lagged response of strategic investment to reduce cost and escape

competition (Aghion et al, 2001). It is these strategic investments by incumbent firms to escape

competition, in addition to the direct actions of entrepreneurs, wherein innovation occurs.

One factor shown to affect the level of entrepreneurship is the administrative cost of entry imposed

by governments. In a 2002 study, Djankov et al characterized the administrative entry costs to start a

business across 85 countries (including both direct payments and time spent complying with procedures).

The study revealed dramatic differences between entry costs (expressed as percentages of per capita

income) in the US (0.017) versus Europe (0.35 to 0.40) as well as other countries. In addition, the study

indicated that countries with lower entry costs had higher entry rates as well as higher productivity, as the

Schumpeterian gale anticipates. The Djankov entry cost data were subsequently employed in a number of

studies to further examine the basic findings. These studies revealed that entry costs had a significant

detrimental impact on entry rates (Fisman and Sarria-Alende 2004, Klapper Laevan and Rajan 2006,

Cicone and Papaioanno 2007, Dreher and Gassebner 2007) as well as total factor productivity (TFP)

(Nicolettti and Scarpetta 2003, Fisman and Sarria-Alende 2004, Loayza Oviedo Serven 2005, Barseghyan

2008).

Not surprisingly, these findings stimulated widespread policy reforms to simplify business startup:

193 reforms across 116 countries (World Bank 2008). The reforms in turn fueled a second wave of

empirical analyses employing new natural experiments with varied entry costs and administrative

procedures. The second wave of empirics largely corroborated the results from the initial wave. Studies

of SARE (Expedite Business Startup System-- a federal government program allowing startups to open

operations within one business day of application) in Mexico indicated that the startup rate (new entry as

a percentage of the number of incumbent firms) increased four to five percentage points after the

introduction of a more simplified, expedited new business registration system (Kaplan, Piedra, Siera 2007

and Bruhn 2008). In addition, the SARE studies found that the increased entry rate was associated with

increased job creation (Kaplan et al 2007) as well as decreased prices (Bruhn 2008). Similar results were

found for reforms in Brazil (Monteiro and Assuncao 2006), Russia (Yakovlev and Zhuravskaya 2007) and

India (Chari 2007). Moreover, these results were consistent with prior studies of US deregulation, such as

Olley and Pakes (1996) who found that telecommunications deregulation stimulated significant entry,

3

which in turn generated the downsizing and shutdown of unproductive plants resulting in substantial

increases in productivity growth. These studies have been used to reinforce the popular view that

regulation is counter to entry and innovation, expressed in political rhetoric, as well as in many policies

proposed in response to the economic downturn1.

Despite the apparent consensus regarding the relationship between entry costs and innovation, entry

regulations have been pervasive, as the first order concerns of regulations are rarely innovation.

Regulations are designed, among other reasons, to correct market failures, protect regional interests, and

contribute to the moral aspirations of a society (e.g reduce discrimination, enforce language laws). When

designing regulations for these first order concerns, policymakers are, and should be, concerned about

their impact on innovation. However, the literature offers little indication as to the impact of different

types of regulations on entry and innovation. This is because prior empirical work has either looked at

the impact of a specific regulation (Thomas 1990) or, more commonly, has employed measures of

regulatory extent – such as the strength of relevant political parties (Bertrand and Kramarz 2002), number

of entry procedures (Klapper Laeven Rajan 2006; Djankov 2002), length of the relevant administrative

code, or the size of the respective regulatory agency budget (Dudley and Warren 2010) – rather than the

nature of the specific regulations themselves. While these measures of regulatory extent were extremely

valuable in allowing researchers to compare the impact of administrative costs on entry and productivity,

a natural next step is examining the impact of a broader set of regulations.

Since industry regulation is rarely monolithic, there may be opportunity to choose from a menu of

regulations in an effort to preserve the first-order goals of regulation while minimizing the Schumpeterian

costs of regulation (reduced innovation). Indeed, most regulated settings already comprise multiple

regulations including entry restrictions and/or operating restrictions. Entry restrictions and operating

restrictions differ in their impact on entrant and incumbent behavior. Operating regulations typically

mandate specific firm behavior, such as minimum product quality. For example, in the child-care

industry regulations include a minimum staff-child ratio (Hotz and Xiao 2011). Accordingly, one concern

with operating regulations is they may limit firms’ ability to differentiate themselves. An additional

concern is that regulatory mandates may be driven by current processes and/or technologies and, as a

result, may limit firms’ ability to innovate or respond to future changes within the industry.

In contrast, entry regulations set no mandate for firm behavior. Rather they achieve their goals by

screening potential entrants based on human and/or physical capital requirements. For instance, state

banking commissions typically require potential entrants to have extensive industry experience as well as

invested capital in excess of a million dollars in order to enter. Entry regulations such as these may allow

for maximum flexibility in incumbent decision-making when compared to operating regulations. 1

4

However, by increasing the size and capital requirements for entrepreneurs there is concern these

regulations may indiscriminately screen out small, young firms (Klapper, Laeven, Rajan 2006).

Furthermore, by restricting the pool of potential entrants, such regulation may deter competition and

affect incumbent incentives to innovate (Posner 1975, Peltzman 1978, Stigler 1971).

Different regulatory frameworks may also provide perverse incentives to manipulate the regulations

to the benefit of regulatory agencies at the detriment of society. Indeed, the high economic costs to

regulation found in the previous literature may be due in part to regulatory capture, rather than a

consequence of entry regulations themselves. For example, the administrative costs from the Djankov

(2002) data tend to be highly correlated with the country corruption indices (Fisman and Sarria-Allende

2004). Regulatory “capture” occurs when regulation is used as an instrument for generating rents for

those controlling the regulation, be they incumbents (Stigler 1971) or politicians (Schliefer and Vishny

1993). Regulatory capture is possible because the diffuse incentives of outside individuals to fight

regulations are swamped by the financial incentives of the captured parties to pursue particular

regulations. This view is orthogonal to the traditional “Pigovian” view of regulation. Under the Pigovian

view, regulation is a policy tool to correct market failures arising from inherent characteristics of an

industry such as public goods, externalities, natural monopolies, or information asymmetries (Pigou,

1938). Regulations in these contexts attempt to correct the failures by internalizing externalities,

controlling monopolists, or providing the public with improved information (Dudley, 2005).

The empirical record provides considerable support for both regulatory views, with most regulations

likely being initiated for market correction purposes and ultimately yielding some regulatory capture.

When regulation is primarily of the capture form, it is not surprising that deregulation produces the entry

and productivity benefits seen in the reform studies. In these instances, deregulation is merely removing

another market imperfection. Therefore, while the consensus in the literature may be that entry regulation

substantially constrains productivity growth, this may not be borne out to the same extent in settings with

less regulatory capture. Moreover, understanding the relative performance of different regulations with

respect to their impact on entry and innovation may not only allow policymakers to make more informed

decisions regarding the Schumpeterian costs of their regulations, but may also allow for a more accurate

assessment of the degree of regulatory capture induced by different regulations.

To study the Schumpeterian cost of regulation we have assembled data on demand, firms, and

regulations across fifty-two markets and twenty state regulatory environments for the United States Bail

Bond Industry between 1990 and 2004. The United States bail bond industry has natural quasi-

experimental properties that make it ideal for studying the Schumpeterian cost of regulation. The industry

comprises roughly 14,000 owner-managed firms in over 3,000 discrete markets (county courthouses) that

5

differ in demand (demographics and crime rates), entry costs and restrictions (through entry regulations),

as well as profitability per defendant (through operating regulations). Furthermore, there are measures of

market productivity and innovation that are comparable across markets and with time. Finally, there is

substantial industry churn: new firms enter each year at a mean rate of 20% of the incumbent population

(versus 10% within the economy generally). Thus the setting presents an unusually sensitive test of

entrepreneurial response to differences/changes in market conditions (including regulations).

Our approach to examining the cost of regulation on entry and innovation utilizes structural models

of the entrepreneur’s entry decision and the incumbent’s continuation (not to exit) decision. These models

allow us to recover parameter cost estimates for specific regulations. We then employ these parameter

estimates in simulated firm decisions for unobserved regulatory regimes in a representative market over

several periods. The aggregation of entrant and incumbent decisions over these periods generates

evolutions in market structure and productivity for the representative market. These counterfactual policy

simulations allow us to better understand the extent to which differences in entry and innovation are

driven by differences in regulation.

Results from this exercise indicate that a combination of entry and operating regulations reduces the

equilibrium number of firms, yet increases innovation relative to a setting with no regulation. More

specifically, the analysis reveals that regulations affecting firm self-selection (such as entry or

continuation fees) increase innovation whereas operating regulations circumscribing firm behavior have

no impact on firm entry or innovation. The innovation benefits of these regulations are assumed above

and beyond their intended benefits in this setting (prohibiting transaction inequities and rights abuses).

These results suggest it is possible to design regulations without imposing Schumpeterian cost on

innovation. They also provide preliminary clues as to which regulations best accomplish this goal.

Finally, and perhaps most importantly our results suggest that Schumpeterian cost itself may need to be

redefined. Entry and innovation actually oppose one another in our setting. Not only does excess entry

steal share, in so doing it also suppresses incentives to innovate.

2. Setting: The US Bail Bond Industry

2.1 Introduction

Bail is a financial system for providing defendants freedom prior to trial while simultaneously

attempting to ensure their appearance at trial. The court holds the bail amount in exchange for pre-trial

release of the defendant. If the defendant appears at trial, the full bail amount is returned to the defendant

regardless of trial outcome. If the defendant fails to appear at trial however, the court retains the full bail

amount and issues an arrest warrant for the defendant. Surety bonds (the bail bond industry) are a market

mechanism for meeting bail in which a bail agent posts a surety bond with the court on behalf of a 6

defendant. The defendant pays the agent a non-recoverable fee (typically 10% of the bail) and also

provides collateral for the bond (from 25% to 200% of the bail). The agent is then liable for the bail

amount if the defendant fails to appear (FTA) at court, and so agents have considerable rights to monitor

and apprehend their defendants.

There is substantial debate about the relative merits of public enforcement versus private

enforcement (bail bonds) of pre-trial release. Despite concerns the industry is tainted by corrupt and/or

strong arm tactics, past economic studies of bail have indicated that private enforcement offers many

social advantages over public enforcement. In particular, competition among bail bond agents yields

bond fees that partially compensate for judicial discrimination in bail setting (Ayres and Waldfogel 1994).

In addition, private enforcement reduces FTA and fugitive rates relative to public enforcement. After

matching defendants on inherent flight risk, FTA rates are 23% lower and fugitive rates are 46% lower for

defendants on surety bonds (private enforcement) versus deposit bonds (public enforcement) (Helland &

Tabarrok 2004). These lower FTA and fugitive rates and their associated police and court costs translate

into estimated public savings of 63% per pre-trial release defendant for private enforcement relative to

public enforcement (Block and Twist, 1997). Accordingly, the bail bond industry offers benefit to its

customers (defendants) through release, and to communities through decreased FTA and fugitive rates.

2.2 Bail Contracts

Figure 1 outlines a typical bail transaction. Once a person is arrested, he or she must be arraigned

within thirty-six hours. The purpose of the arraignment is to file formal charges against the defendant and

to establish the terms of pre-trial release. In approximately 45% of cases defendants are released without

up-front cost (called non-financial release). Bail agents only become relevant in cases where the court

sets a bail, does not allow bail to be satisfied through an unsecured bond, and where the defendants

cannot raise bail on their own (roughly 21% of cases).

Defendants who wish to be released, but who cannot raise the full bail amount, select bail agents

either from a list of all agents maintained at the jail,from the yellow pages, or more recently, from online

advertisements. Historically yellow page ads havebeen the primary form of advertising and the largest

fixed expense of bail agencies.

From the perspective of the defendants, the service bail bond agents provide is undifferentiated.

Price competition in this industry occurs through the defendants (or persons on their behalf) calling

several agents to find the lowest commission fee and/or collateral requirements given the defendant’

history, charges, and bail amounts. Evidence of such price competition comes from Ayres and Waldfogel

(1994) who show competitive markets exhibit lower commission fees.

7

During these calls, the defendants (or persons on their behalf) file applications with prospective bail

agents. The application allows the agent to determine the risk-worthiness of a defendant in an

underwriting process that is similar to those for bank loans or insurance. In the case of bail, the

determination involves not only financial merit (income, assets), but also likelihood of flight and

likelihood of recapture in the case of flight. Thus the application includes substantial information about

defendants’ social networks. If the agent determines a defendant is a reasonable risk, then the agent and

defendant agree on the up-front fee (percentage of the bail) as well as collateral or co-signer requirements.

The fee is non-recoverable by the defendant, thus is irrelevant to the defendant’s ex post behavior.

Accordingly, the main financial instrument for ensuring a defendant’s appearance at trial is the collateral.

As collateral may be minimal given the financial limitations of most defendants, screening and

monitoring of clients are the primary mechanisms for ensuring appearance at trial.

Once defendants select a bail agent, they sign a contract outlining the agent’s rights in monitoring

them prior to trial and apprehending them in the event they “fail to appear” (FTA). The best outcome for

the agent is the defendant appearing at trial (roughly 75% of arrests). In these cases the agent keeps the

fee and releases the collateral to the defendant. If the defendant fails to appear, then the agent becomes

responsible to the court for the bail amount. States vary both in the recovery period (the length of time

agents have to produce the defendant to the court before they are liable for the bail) and in the amount of

bail that is forfeited. Often the forfeiture amount is tied to the length of time between the failure to appear

(FTA) and the date when a defendant is returned. In about 5% of arrests the defendant is not returned

within the recovery period, and is classified as a fugitive. This is the worst possible outcome for agents

because they must remit to the court the full amount of bail, and in addition have typically incurred

substantial apprehension costs. According to the American Surety Company, to be successful “a bail

agent must ensure at least 98.5% of their clients are present for all required court appearances.” Not

surprisingly, firm exit is common in the bail bond industry.

While screening and monitoring are done by the bail agent, apprehension is typically done by “bail

recovery agents” who are licensed separately from bail agents. As a prominent example, “Dog the

Bounty Hunter” is a bail recovery agent, rather than a bail agent and thus does not write bonds. States

vary in the extent to which bail agents can also be bail recovery agents, but even when co-licensing is

possible, bail agents often prefer to subcontract defendant recovery. States also vary in the restrictions

they impose on recovery.

2.3 Bail Bond Regulation

Regulation in the bail bond industry is performed at the state-level (typically by the state Department

8

of Insurance), to correct for informational asymmetries that exist between defendants and bail bond

agents. These bail bond regulations are designed to prevent consumer and human right abuses that might

arise due to 1) the informational asymmetry between bail agents and prospective clients, 2) the rare

frequency of the business transaction for most clients, and 3) the sensitive nature of the client situation.

Examples of potential consumer rights abuses include illegally acquiring clients (e.g. lying that a family

member sent the agent on the client’s behalf or paying government employees to promote their service),

distorting client costs (e.g. bribing judges to increase bail amounts to generate higher bail fees), and

exaggerating influence with courts (e.g. bail agent ability to alter release conditions and/or final court

adjudication to yield higher client payments). In addition to consumer rights abuses in this setting are

human rights abuses such as extortion, violence, kidnapping, and illegal searches of home or property to

lower monitoring costs and/or generate additional revenue and services. These rights abuses can be

directed toward either the client or individuals associated with the client (i.e. family members, friends,

and/or colleagues).2 In order to fully appreciate the regulatory objectives in the bail bond industry, one

should consider how these issues of uncertain quality and the government’s role therein contribute to a

particularly complex market environment.

Although the motivation for regulation may be similar across states, the specific regulations vary

across states and over time. This variation appears uncorrelated with demographics, population, or crime

levels. In most states, bail bond regulation comprises a combination of background restrictions and bail

skill requirements at the individual level as well as setup costs, required business practices, and

commission fee limits at the firm level. These restrictions affect firms’ entry and operating costs.

2.4 Entrepreneurship and Innovation in Bail Markets

Bail bond firms can be operated from a home office with less than $15,000 of invested capital (a

fidelity bond filed with the state, a yellow page ad, and collateral held with the underwriter) (Verocchi

2006). In this sense, the bail bond industry closely matches other professional service industries.

Entrepreneurs entering a given market are typically drawn from the local population of bail agents. In

general, these agents typically have low human capital. For example, not uncommon licensing

requirements are a GED (General Equivalency Degree—a certification for people not completing high

school), no felony convictions, and no insurance violations.

The most significant variable cost for bail bond firms is the premium for bond underwriting

(typically 2-5% of the bail amount, and therefore approximately 50% of revenue). This cost is largely

outside their control until they reach a scale where pooled risk allows them to negotiate lower premiums

or to self-insure. The primary costs that are under firms’ control are (1) ex ante screening—the process 2

9

for gathering defendant data and the decision rule for determining which defendants to bond and (2) ex

post monitoring—various procedures to ensure defendants appear at trial. These costs (reflected in

differences in firm FTA rate) are the primary sources of firm heterogeneity.

Innovation in this context resembles the process innovation in Duranton and Puga (2001) who model

the role of diversified cities in incubating innovation. In their model, nascent entrepreneurs may know what

product/service they want to provide, but may not know all the details of how to produce it. They try to find

their ideal production process by sequentially mimicking processes already in use locally. Accordingly it is in

their best interest to locate in diversified cities during this search process. Later when they have discovered

their ideal process (depending on relocation costs) they may move to specialized cities to avoid congestion.

Interviews with bail bond entrepreneurs indicate they learn basic principles of the industry from

mandatory coursework, but that much of their ultimate “production function” is discovered through trial-and-

error experimentation. For example, one founder followed the standard practice of purchasing a yellow page

ad when he entered the industry, but discovered it to have a low return on investment. The half-page ad cost

$2000 per month, but didn’t generate much traffic because the ordering of ads is based on size as well as

seniority. Thus his ad was behind all ¾ and full-page ads as well as all the half-page ads of older firms.

Further the clients he obtained through these ads were of very low quality, because by the time they reached

his ad they had already been rejected by several firms.

In an effort to find a more effective way to obtain clients, he enrolled in a community college course,

“how to use the internet to drive business”. After completing the course he paid $2000 to create a website and

another $200 per month for search engine optimization. His phone began ringing off the hook. Furthermore,

since this was the late 1990s, these internet clients tended to be affluent, which meant they were reliable and

had adequate collateral to ensure appearance at trial.

Thus while there is no R&D in bail bonds, there is investment in general purpose technologies in

conjunction with organizational innovation to exploit these technologies. Some of the relevant technologies in

this setting include websites as described above (to drive traffic), as well as later technologies such as criminal

and credit databases (for screening), automated phone call systems (for monitoring) and GPS tracking

technology (for monitoring)). This characterization of innovation is consistent with the definition of

innovation in Mansfield (1966) as the utilization of an invention, with Rosenberg’s (1982) definition of

process innovation as the introduction of elements into a firm’s production or service operation, as well as

Guadalupe, Kuzmina and Thomas (2012) operationalization of process innovation as the combination of

new technology and new methods of organizing production.

3. Data

Data for our study come from three sources. Demand (arrest) and demographic data were collected

10

from the State Court Processing Statistics (SCPS) through the National Pretrial Reporting Project. The

SCPS includes arrest histories and outcomes on all felony defendants from the month of May across forty

of the most populous counties in the United States for each even year between 1990 and 2004. The data

set includes information regarding their demographics, prior arrest history and adjudication, current arrest

charges, current release method, and subsequent court appearances for the current charges. Most

significantly for our purposes, for most (98.5%) of those individuals released on bail it includes

information on their bail amount, their payment type (cash, property, or surety bond), whether they failed

to appear for court (FTA), and if they were classified as a fugitive at the one year mark. The complete

sample of municipalities and annualized arrest counts appear in Table 1 and Figure 2a.

Firm counts for each county and year were compiled from the National Establishment Time-Series

(NETS) Database, a longitudinal collection of Dun & Bradstreet records developed by Walls and

Associates (see Figure 2b for a histogram of firm counts across sample counties). These data include firm

age and firm-employment information. As a check on the quality of the NETS data in our setting, we also

obtained yellow page ads from the National Archives for all municipalities in all years. Comparison of

the firm entries and exits from the Yellow Page ads to the NETS data indicated near perfect alignment.

Graphical summaries of distributions for market outcomes (entry and exit as well as release rate,

annualized FTA rate and fugitive rates) are presented in Figures 2c-2g, respectively.

The twenty state-level bail bond regulatory environments for the counties in the sample were

recreated for the years from 1990-2004 by reviewing each state’s archival statutes and administrative

code. To collect these data we began by reviewing current regulations in each state. We then found the

corresponding sections in the current state statutes and administrative code. We tracked those sections of

statutes and codes backward through the sample period to characterize all the bail bond regulations in

place in each state-year, careful to check for statutes that may no longer be in effect. For each state, all of

the bail bond regulations were classified based on their respective regulation type, e.g, pre-requisite

coursework is a regulation “type” which can take a number of forms (required hours and/or course

content).

This process resulted in a set of over forty regulation “types” related to bail agents and firms. From

those forty types we restricted attention to a subset of six with sufficient variance across states. The

regulations included requirements for potential entrants to have industry experience, pass an exam and

pass a background check, post a cash bond with the court (often $25,000),maintain contact with their

clients, keep daily records, maintain a public place of business, and complete continuing education

courses. Using these specific regulatory types allows us to overcome some of the challenges in much of

the regulation literature, which use proxies or measures of regulatory extent rather than specific

11

regulations.

The entire data set thus comprises data on demand, firm counts and regulations for the forty most

populous counties in the United States between 1990 and 2004. Note, however, as a constraint of the

SCPS data, the forty counties change somewhat with time. As a result, the data set is not a complete

panel and is treated as a cross-sectional set. Multiple observations in the same county are currently

treated independently. Additional control data relating to population, personal income, and interest rates

were collected from the US Bureau of Economic Analysis for each region-year. House price indices were

collected from the US Federal Housing Finance Agency (the Case-Shiller House Price Index was not

used, as it did not contain all relevant counties). For a complete listing of the counties, the number of

observations per county (listed by relevant municipality), average firm counts, and average number of

arrests see Table 1. Summary statistics are provided in Table 2.

4. Empirical Approach

Our empirical approach to examining the Schumpeterian cost of regulation employs structural

models of the entrepreneur’s entry decision and the incumbent’s continuation decision to identify the

impact on firm entry and innovation3. This is performed as two-step estimation. First, we use incumbent

firm characteristics and their continuation decisions to recover firm-specific FTA rates, fixed costs,

variable costs, and operating regulatory costs. Second, the incumbent estimates are incorporated in the

entrant estimation along with the number of potential entrants in the market to recover entrant FTA rates

and entry costs (including entry regulations). The full set of incumbent and entrant parameter estimates

are then employed in simulated firm decisions for unobserved regulatory regimes. The aggregation of

entrant and incumbent decisions over several periods in a representative market generates evolutions in

market structure and innovation (measured as improvement in release rate and FTA rate) for that market.

These counterfactual policy simulations allow us to better understand the extent to which differences in

entry and innovation rates are driven by differences in regulation. They also facilitate tradeoff analyses

between regulations, tradeoff analyses between outcomes (entry versus innovation) without the inherent

biases of reduced form models with respect to selection effects.

4.1 Incumbents continuation (or exit) decision

In each period, incumbents assess whether to continue operations or to exit the industry. Because

over 96% of firms are single location operations, the firm is deciding whether to stay to in operation, not

where to stay in operation. Furthermore, because bail industry characteristics and regulations prevent

short-term losses (e.g. using future profits to fund current productivity enhancements), firms in this 3

12

industry operate under a strict profit constraint. Accordingly as soon as a firm’s operations result in

losses, we expect the firm to exit the market. As such, the incumbents’ decision to continue in operation

compares the firm’s profit for the year to its continuation costs to remain in the market the following year:

π imt ≥ ContCostsmt (6)

where π imt is the profit for firm, i, in market, m, at time, t. Since the only regulation affecting continuation

costs is the continuing education requirement, we model ontinuation costs as a linear function of that

requirement in each market. Note there is no constant term in the continuation costs; regions without a

continuing education requirement have continuation costs of zero.

ContCostsmt=γContEduc ContEd ucmt (7)

where ContEducmt takes a value of one if at time t market m is in a state with a continuing education

requirement.

The incumbent, i, has the following estimated profit equation4:

π imt=Feemt ∙ q imt¿−FC mt−ScreenVCmt ∙ qimt

¿−MonitorVCmt ∙( qimt¿)2−q imt

¿FTA imt Recoverymt

(8)

which reflects firms’ revenues less their fixed and variable costs and expected costs from client FTAs,

recovery activities, and fugitives. The commission , Feemt, is the percentage of the bail amount the firm

receives from the client either as mandated by the state or determined competitively. Quantity, q imt¿, is

the amount in hundred thousand dollars the firm bails out (this is expressed in dollars rather than number

of contracts to account for risk variance across contracts). Fixed cost, FCmt , is modeled as a constant

plus regulations requiring a place of business scaled by the House Price Index.

FCmt=γ FCConstant +γ MaintainPublicBus PublicBusmt HPImt (9)

The first variable cost, ScreenVCmt, is the variable cost per contract. It is modeled as a function of

mean personal income in the market to reflect the opportunity costs and be constant in the number of

4

13

clients.

ScreenVCmt=γ VCCost Scalar PersonalIncomemt (10)

where PersonalIncomemt is the average personal income in market m in year t. Total screen cost is the

variable cost times the amount bailed out. While bail firms screen more defendants than they take on as

clients, we assume the acceptance rate is reasonably stable and thus bail amount is a useful proxy for the

amount and complexity of screening.

The second variable cost, Monitor VCmt, is modeled as a constant plus a linear function of

regulations affecting variable costs (such as requiring contact with the defendant and maintaining daily

records). This captures the costs associated with monitoring clients pending the trial date, and is assumed

to increase (at a decreasing rate) with the amount of bail (e.g. needing more support staff or more

automation to stay in touch with all clients).

Monitor VCmt=γ VC+γ VCContact Contactmt+γ VC DailyRecords DailyRecordsmt (11)

where γVCcaptures the average monitoring costs across regions, Contactmt takes a value of one if at time t

market m is in a state with regulations requiring bail bond firms to maintain contact with their clients

between release and the trial date, and DailyRecordsmt takes a value of one if at time t market m is in a

state with regulations requiring bail firms to maintain daily records.

Firm heterogeneity is captured by the firm-specific FTA rate. Improvements to the firm’s FTA rate

can be the result of trial and error experimentation, adoption of new technologies and/or scale economies.

Accordingly, for incumbents, the expected firm-specific FTA rate is a function of the market level FTA

rate, firm age, and firm size. Age is included to capture the effects of trial and error experimentation and

is also used to identify a firm’s entry cohort. The firm’s entry cohort captures conditions at the time of

firm birth affecting its later productivity (Stinchcombe 1965, Nelson 1991, Henderson and Mitchell 1997,

Hannan & Freeman 1984). Finally, firm size is included in the firm-specific FTA rate to account for scale

economies as well as opportunities to better pool FTA risk.

FTAimt=f (FTAmt , FirmAgeimt ,Cohort imt , FirmSizeimt)+εimt (12)

where FTAmt is the market failure to appear rate, Ageimtis the age of firm i, measured in years since

14

founding. ¿¿ imt ¿ is firm size, measured as number of employees. Note that firm size is considered

exogenous in the current model. The firm-specific shock, ε imt, is drawn from a normal distribution

N (0 ,σ I2). This firm-specific shock is how a firm’s realized firm-specific FTA rate differs from the

expected firm-specific FTA rate given the firm’s characteristics (size, age). This shock is firm-year

specific. Thus it does not capture time-invariant unobserved firm characteristics. Rather, it reflects a

sudden, distinct increase in FTA rate from a particularly bad (or good) draw of clients.

While the FTA rate is firm specific, the expected costs associated with clients who fail to appear,

Recoverymt , is market specific. This is because recovery efforts are most often outsourced to bail

enforcement agents, or bounty hunters.

Recoverymt=(1−FUGmt ) ∙ReturnPercentagemt+FUGmt (13)

The expression (1−FUGmt ) captures the percentage of FTA clients who will be recovered within one

year, ReturnPercentagemt is the expected percentage of the full bail amount the firm will remit to the

court for those clients. In contrast, FUGmt is the percentage of clients who remain fugitives at year end

and for whom the firm remits the full bail amount to the court. The total cost of clients who fail to appear

is thus this recovery rate times the firm FTA rate and the amount bailed out. This is the primary cost for

firms in this industry and the primary source of heterogeneity among established firms.

The incumbent firm’s objective function is thus to choose the amount of bail contracts, qE ,mt¿, to

maximize profits given local regulatory costs, recovery costs, and firm-level expected FTA rate,:

q imt¿=

Feemt−ScreenVCmt−FTA imt Recoverymt

2 ∙MonitorVCmt (14)

4.2 Potential entrant’s entry decision

In each market year, we assume the set of potential entrants equals the number of non-owner, bail

bond employees currently in the market (Figure 3 shows the number of potential entrants relative to the

number of entrants). In addition to matching the stylized fact that most entrepreneurs found firms in their

current geographic region and industry (Stuart and Sorenson 2003), this assumption reflects experience

regulations requiring founders to work for existing firms for a given number of years prior to starting their

own firm. Thus potential entrants are deciding whether and when to enter their local market, rather than

selecting which market to enter.

15

We further assume that within a market, all entrants have the same expectation of future profits and

therefore the same probability of entry. This is done to accommodate the large number of potential

entrants and because we lack individual-level data. As such, in contrast to many structural models of firm

entry or labor entry there are no “high” and “low” entrant types in the model. All potential entrants

within a market are considered homogenous until entry.

Entrant firms, E, enter market, m, at time, t, if their expected profits for the year meet or exceed the

entry costs. As with the incumbent firm decision, there is a strict profit constraint. Thus potential

entrants only consider the next year’s profits when deciding whether to enter.

E πE ,mt ≥ EntryCostsmt (15)

where E πE ,mt is the expected profits for identical entrants, E.

EntryCostsmtare modeled as a linear function of the initial customer acquisition costs and the market

specific regulations affecting up-front costs (requiring an industry exam, posting a cash bond with the

court, having industry experience and passing a background check)

EntryCostsmt=γ AC ( Nmt

Demandmt)+γ Exam Exammt+γ CashBond CashBondmt Interestmt+γ Exper Expermt+γ Background Backgroundmt

(16)

Initial customer acquisition costs are captured by Nmt , the number of incumbents in the market (scaled by

average firm size), divided by Demandmt , the market demand for bail. Acquisition costs account for the

higher costs later entrants have to pay to acquire customers as market demand is met. Regulations

requiring an exam are captured by an indicator variable, Exammt . Regulations requiring the posting of a

cash bond (often around $20,000) are captured by the variable CashBondmt , scaled by the interest rate at

the time. Apprenticeship or industry experience requirements are captured by the dummy variable,

Expermt. Finally, Backgroundmt is a dummy for regulations requiring either a background check or a

requirement that agents have no felony or misdemeanor convictions. The estimated entry costs should

reflect both the direct financial cost of the regulations and the screening / selection effect of potential

entrants due to the regulations.

All potential entrants have the following expected profit:

16

E πE ,mt=Feemt ∙ qE ,mt¿−FC mt−ScreenVCmt ∙ qE, mt

¿−MonitorVCmt ∙ (qE ,mt¿ )2−qE ,mt

¿ E FTA E , mt Recovery mt

(17)

which matches the incumbent profit equation. For potential entrants, the expected firm-level FTA rate is

still firm specific, however, all potential entrants within market m at time t share the same expected firm-

level FTA rate. The firm-level FTA rate is a function of the market level FTA rate, FTAmt , and a firm-

specific shock, ε E , mt, which is drawn from a normal distribution N (0 ,σ E2) . The distribution of the

entrant error term captures the variance in entrants’ beliefs regarding their expected firm-level FTA. This

dispersion in beliefs may be due to differential access to market-level data (demand, FTA and fugitives

rates) or to differences in performance expectations. Only the distribution of potential entrants behavior

is recovered, not specific potential entrants’ decisions.

E FTAE ,mt=βFTAE FTAmt+ε E ,mt (18)

As in the incumbent decision, potential entrants select the amount of bail contracts, qE ,mt¿, to

maximize their expected profits, given local regulatory costs, local recovery costs, and their firm-level

expected FTA rate:

qE ,mt¿=

Feemt−ScreenVCmt−E FTAE ,mt Recoverymt

2 ∙ MonitorVCmt (19)

4.3 Model constraints

We impose two major constraints on the incumbent model. First, we try to force the total estimated

bail bonds in each market-year to match the actual amount in the data for that market-year by minimizing

the bail bond quantity moment condition.

Q̂mt=qE , mt¿ ( Emt )+∑

iq imt

¿ ( Iimt ) (20)

Quantity Moment=(Q̂mt−Qmt)2 (21)

where Emt and I imt capture the behavior of firms that have just entered, those continuing in the operations, 17

and those exiting. Since the churn rate is quite high in this industry it is important to include the

productivity of failed firms in the moment condition.

Second, we try to force the total estimated FTA rate in each market-year to match the FTA in the data

for that market-year by minimizing the FTA rate moment condition.

F̂TAmt=FTA E, mt qE,mt

¿ ( Emt )+∑i

FTA imt q imt¿ ( I imt)

qE ,mt¿ ( Emt )+∑

iqimt

¿ ( I imt ) (22)

FTA Moment=( F̂TAmt−FTA mt)2 (23)

Additionally, constraints are imposed such that total fixed cost, total screening cost and total

monitoring cost are positive. Note that individual regulations are allowed to be positive or negative.

FCmt ≥ 0 (24)

ScreenVCmt ≥ 0 (25)

MonitorVCmt ≥0 (26)

No constraints were placed on the continuation costs (in the incumbent decision model) or on the

entrant costs (in the entrant decision model).

4.2.4 Estimation Technique and Constraints

The model as described above was estimated in two stages. First, we estimate the incumbent

decision by maximizing the likelihood of seeing incumbents continue or exit in a market-year given the

market-year bail quantity and FTA moment conditions and cost constraints. This allows us to recover the

fixed, variable, and continuation cost estimates, as well as the firm-specific FTA rate,. Note that

incumbent estimation includes entrants that have just entered and are deciding to continue after their first

year in operation.

Second, we estimate the entrant decision by maximizing the likelihood of seeing the number of

entrants in a market-year given the number of potential entrants and the recovered parameters from

incumbent estimation. This second estimation allows us to recover the entry costs (including cost of entry

regulations) and the distribution of entrant beliefs regarding their expected firm-specific FTA.

In both estimations, all FTA rates were constrained to fall between zero and one. We accomplish this

18

by transforming the incumbent rate, FTA imt using a scaled cumulative normal distribution function.

FTA imt=δ Φ (1−FTAI imt )+εimt (27)

where δ is a scaling factor on the cumulative normal distribution (CDF) function allowing the firm-

specific FTA rate to vary across a specific range of values. This scaling factor was estimated outside the

model. The interior component of the CDF is inverted to match the natural shape of the relationship

between firm-specific FTA rate and age. The inverted rate, FTAI imt is a function of FTAmt, FirmAgeimt,

FirmSizeimt, and the firm’s cohort, where cohorts are defined as the decade of entry. The mean and

variance of the cumulative normal distribution function was estimated within the model to allow for

maximum flexibility of the firm-specific FTA range form. The functional form with the inclusion of

FTAConstant and the division of all parameters by βFTAi is simply to allow for this flexibility.

FTAI imt=FTAmt+α Age1

βFTAiFirmAgeimt ( FirmAge imt ≤ 10 )+

α Age 2

βFTAiFirmAge imt (10<FirmAgeimt ≤ 20 )+

α Age3

βFTAi21 ( FirmAgeimt>20 )+

βempl

β FTAiFirmSizeimt+

FTAConstantβFTAi

(28 )

The standard deviation on the incumbent FTA error term, ε imt , was constrained such that the error

shocks would keep the incumbent FTA as a rate with values between 0 and 1.

0 ≤ σ I ≤ 0.25 (29)

Similarly, to maintain the entrant FTA rate as a rate with values between 0 and 1, the entrant expected

rate, E FTAE ,mt, as defined in equation 12 is constrained such that

0≤ βFTAE ≤ 3 (30)

and the standard deviation on the entrant FTA error term was constrained such that

0≤ σ E ≤0.25 (31)

19

5. Parameter Estimates

5.1 Parameters Recovered from the Incumbent Continuation Decision

Fixed and Variable Cost Parameters

Estimates for the recovered parameters are presented in Table 3a; estimated cost estimates are

presented in Table 3b. Looking first at the common profit equation, both constant fixed costs and those

associated with the place of business regulation were found to be negligible. This is not surprising given

(1) this is not a fixed cost intensive business (as previously discussed, the primary fixed-cost expenditure

for a firm is a yellow-page advertisement) and (2) a place of business may scale from shared office space

to a full store-front depending on the firm’s quantity bailed out. As such, the cost associated with place of

business may be captured in the variable cost components.

Variable costs comprise both the constant (per bail amount) screening variable cost and the increasing

(per bail amount) monitoring variable cost. Screening cost was found to be just a few dollars per hundred

thousand dollars bailed out. This is roughly in line with credit checks. Monitoring cost is found to be

roughly 0.4% of the bail amount ($403 per hundred thousand dollars of bail). The regulations for

maintaining contact and daily records were found to have minimal impact. This could be due to the fact

that the cost of these practices is negligible, that firms even in unregulated regions also follow them, or

that these regulations are difficult to monitor and enforce by the regulatory body.

Incumbent Firm FTA Rate Parameters

The estimates indicate the main cost for bail bond firms is clients’ failure to appear (FTA). This is

consistent with what we learned in interviews. Figure 4a shows the estimated firm FTA rate as a function

of firm age implied by the age coefficients in Table 3a (the spread is due to variance in firm size). Not

surprisingly, we see rapid improvement in FTA rate among younger firms and less dramatic rates of

improvement among the middle cohort. What is surprising, however, are the distinct shifts of FTA rates

among the different cohorts. One would expect older firms to have the lowest firm FTA rates if trial and

error learning was driving the improvement. However the figure seems to indicate the firm’s birth cohort

plays a substantial role. This could be driven by the time-variant conditions at firm entry, e.g, a high level

of competition at entry such that only highly productive firms survive, or the availability of new

technologies that are more likely to be adopted by later cohorts. Surprisingly, firm size was found to be

insignificant. This is counter to our expectation that larger firms might enjoy scale economies, or might

be able to pool FTA risk over a larger client base.

We combine the effects of all these costs in Figure 4b, which shows the estimated range of profits for

20

firms in our sample. These profit estimates match with expectations of firm profitability in the industry

based on interviews and data provided in the NETS database.

Continuation Cost Parameters

The cost associated with continuing education requirements is estimated to be $562. This was on the

high end of fees we found online for bail bond courses. So while it may reflect actual out of pocket costs,

it may also reflect firms’ making a more conscious decision to remain in the industry. When required to

exert effort to stay in business, lower performing firms may decide to exit. In the absence of this cost,

low performing firms (particularly those that are part-time) linger in the market until they are forced out

by a high FTA rate shock.

5.2 Parameters Recovered from the Potential Entrants Entry Decision

Potential Entrants Expected Firm FTA Rate Parameters

The mean expected FTA rate of potential entrants was estimated to be 0.3654 of their market rate.

Thus potential entrants expect an FTA rate of 3%, compared to a market average rate of 12.5%. This

suggests either significant overconfidence or complete naivete among potential entrants. Such optimism

by potential entrants is aggravated by performance uncertainty (Wu and Knott 2006), which is quite high

in this context (the estimated FTA error shock variance was 0.244). This explains a significant portion of

the variation in the entry behavior. This high level of variation in expected firm-level FTA could be due

to the general variation in potential entrants’ own capability, beliefs about that ability (overconfidence)

and knowledge of the industry and market conditions.

The high level of uncertainty is certainly consistent with the fact that the actual market FTA rates are

not readily known. Courts do not closely monitor and/or aggregate the FTA rate of their defendant

population, nor are they likely to share such information when it is available. Recall, even the SCPS data

set used in this study is (1) available for only defendants in the month of May (or less), (2) is collected

only in even years, and (3) and is made available to the public only after a considerable lag. Although

incumbent firms do closely monitor and aggregate their own FTA rate, they do not publicize their

‘failures’ and closely guard their firm-level FTA rate.

Overall, this overconfidence by potential entrants regarding their performance relative to the market

is not surprising, although the magnitude may be a bit extreme. Despite the magnitude, this does not

indicate potential entrants are behaving irrationally when entering the market. In particular, the high level

of uncertainty may contribute to market optimism that contributes to sustained entry.

21

Entry Cost Parameters

Entry costs have two components: costs of acquiring customers and costs of entry regulations. The

parameter relating to initial customer acquisition costs, AC indicates that as the number of firms (size

weighted) increases relative to the level of demand, it is more expensive to enter the market. The mean

cost of entry in the data from the initial customer acquisition piece is $195,000. This seems exceptionally

high for an industry with expected profits of less than fifty thousand dollars. However, this accounts for

both the actual costs and the expected losses from clients’ failures to appear (which becomes more likely

as firms relax bonding qualifications in an effort to obtain clients in a crowded field of firms). If the costs

reflected only the actual entry costs, all potential entrants would enter once the market appeared

profitable, which is not reflected in the data.

The entry costs associated with entry regulations net to $158,698. Again, this accounts for the actual

cost of the regulations and how the entry regulations alter potential entrant behavior. For instance,

passing an industry exam is unlikely to cost an owner $28,265; nor is a background check likely to cost

$73,765. However both these regulations alter the opportunity costs of potential entrants and, as such,

increase the profitability threshold required for entry. The estimated cost for the cash bond requirement

($56,758)is closely tied to the direct cost of posting the general bond with the court . The regulation

requiring owners to have prior industry experience does not seem to affect the implicit cost of entry. This

is consistent with our assumption that potential entrants are drawn principally from existing bail agents.

Overall, the parameter estimates show that entry regulation need not be financially binding to

potential entrepreneurs to increase entry costs; both financial constraints, such as requiring the posting of

an initial cash bond, and time and capability (opportunity cost) constraints, such as requiring the passing

of an industry exam had similar entry costs.

6. Counterfactual Policy Experiment

In order to investigate the impact of regulation on firm entry and innovation we conducted

counterfactual policy experiments using four different bail bond regulatory regimes: no regulation, entry

regulation only, operating regulation only, and both entry and operating regulation. All the regulatory

environments begin with the same initial condition. This initial condition sets the number and distribution

of incumbent firms (firm-age and firm-size), the number of potential entrants, and market conditions

(arrests, FTA and fugitive rates). In this policy experiment, the initial condition was based on the year

1990 data for Santa Cruz, California. The simulation recalibrates from the initial condition to the

estimated regulatory environments. As a result, the implications from the counterfactual experiments are

22

driven not from the changes from the initial condition, but, rather, from the differences between the

regulatory environments.

6.1 Counterfactual Technique

The counterfactual policy experiment utilizes the same decision logic as described in the entrant and

incumbent models outlined previously. Potential entrants enter if their expected profits exceed entry

costs, and incumbent firms continue in operation if their current profits exceed continuation costs.

In each of the regulatory environments, two phases are simulated for each year. First, the number of

entrants is determined from the number of potential entrants and the probability of entry from the ‘entrant

decision’. To accommodate the fact that entrants vary in their initial size, we draw size (number of

employees) from the distribution of actual entrant size in the data. Entering firms are carried through to

the next period.

In the next phase, both entrants and incumbents determine their optimal quantity, q imt¿, based on their

firm-specific FTA rate (comprising both their permanent FTA rate plus their current period FTA shock).

Each firm’s profitability is then calculated. Firms observe this profit, then decide whether to continue in

the market or to exit. Firms who exit in a given year, firms are assumed to have bailed out half their q imt¿.

Continuing firms bail out their full q imt¿. The quantity of firms and the quantity-weighted FTA values of

all firms are summed to obtain the simulated market productivity measures.

Continuing firms remain as incumbents for the next period and all exiting firms are dropped from the

counterfactual. The total number of potential entrants is then calculated for the subsequent period given

the total number of employees in the market. The entrant and incumbent phase of the counterfactual is

repeated for the next even year (as market data is driven by even years). Seven iterations are conducted

for each policy environment. Rather than comparing each policy environment to the actual data, the

counterfactual policy environments should be compared to each other to account for establishment of a

new equilibrium as we move from the actual environment to the counterfactual environment and to

account for any unique idiosyncrasies of the model.

As there is no firm size decision in the above model, firms retain their size (number of employees)

throughout the experiment (either their firm size in the initial condition or their firm size upon their entry

in the simulation).

6.2 Counterfactual Results

Regulatory impact on entry and exit

23

The first set of counterfactual policy experiments looks at the effect of different regulatory

environments (no regulation, entry regulation only, operating regulation only, and both entry and

operating regulation) on entry and exit. The entry regulations in the counterfactual experiments are the

same as those in the structural model (exam, background check, experience, and cash bond), as are the

operating regulations (place of business), variable costs (maintain contact, daily records), and

continuation costs (continuing education).

Figure 6 indicates that the levels of entry and exit are lower with entry regulations as compared to

both unregulated environments or environments with only operating regulation. This low level of churn

is driven by higher entry costs associated with entry regulations. As there is little difference in expected

costs while in operation (regulations affecting fixed and variable costs were found to be nominal),

unregulated environments and those with operating regulations only show similar initial entry rates.

With regard to exit, environments with operating regulations have higher firm exit. This is due to

continuation costs driving less profitable firms to exit the market. Continuation regulations two effects: a

direct effect of increasing the likelihood of exit, and a dynamic effect on entry. As they decrease the

number of incumbent firms, they lower the customer acquisition cost (more customers relative to the

number of firms). These lower acquisition costs stimulate entry. Thus we see higher entry and exit

(churn) under operating regulations, than we do under no regulation despite similar profit expectations.

Thus environments with both entry and operating regulations have less churn and lower firm counts

than the other environments. Few potential entrants decide to enter due to the higher entry costs and, of

those firms that enter, few stay in the market due higher continuation cost. The higher entry costs

associated with entry regulation seem to offset the lower customer acquisition costs from operating

regulations (continuation costs). Thus entry and operating regulations have complementary effects in

reducing the churn rate and the equilibrium number of firms.

Regulatory impact on innovation

The second set of counterfactual policy experiments looks at the effect of different regulatory

environments on innovation, where innovation is captured by improvements in the surety release/FTA rate

frontier. The exact trade-off between improvements in release rates and market-level FTA rates is not a

topic of this paper. Instead, we view improvements to market-level FTA rates without significant

decrease in the release amount (and vice versa) as innovation.

Looking first at the release rate, we see that quantity bailed out increases over time in environments

with operating regulation only, with entry regulations only, as well as with no regulation (Figure 5). Only

in the environment with both operating and entry regulations is the release rate (quantity bailed out)

24

relatively flat. Notably, in no regulatory environment does the release rate decrease over time. This

suggests there is no regulatory cost with respect to release rate.

Results for the market-level FTA rate exhibit a similar pattern. The FTA rate is both consistently

lower and decreases most in a regime with both operating and entry regulations. The lower market-level

FTA rate is driven largely by a low industry churn rate (Figure 6). New firms enter with high FTA rates

and ultimately are forced out of the market, but their temporary presence increases the market-level FTA

rate. Market level FTA rates can be improved either by reducing the churn rate (higher entry costs) or by

decreasing the time to exit of poor performing firms (higher continuation costs). When both entry costs

and operating costs are in place, the costly churn is reduced and the market-level FTA rate drops as

incumbent firms gain experience and scale and only more productive entrants stay in the market.

Further exploration

We examine this explanation further by conducting a second set of counterfactual experiments that

look at specific components of operating regulations separately (Figures 7 and 8): those affecting the

incumbent profit function through fixed and variable costs (i.e. place of business statute, maintain contact,

and daily records) and those affecting continuation costs (i.e. continuing education courses). This set of

counterfactual policy experiments indicate that environments with operating regulations affecting the

incumbent profit function were comparable to environments with no regulations. Thus the primary effect

of operation regulations on market productivity comes from the continuation costs. Requiring incumbent

firms to re-invest at yearly intervals drives low performing incumbent firms from the market, thereby

reducing FTA rates. Thus, the primary effect of entry and continuing regulations is to reduce churn of

low quality firms, thereby increasing the proportion of high quality firms with low FTA rates.

To better understand the extent to which firm selection is driving the counterfactual results, the final

set of counterfactual experiments considers the same regulatory environments and outcome measures, but

in an environment without firm-level innovation (Figures 9 and 10). Notably, the separation of regulatory

environments still occurs, and entry and operating regulations still having the greatest productivity

improvement (lowest market FTA rate). However, there is almost no improvement in market productivity

over time. Accordingly the innovation we see in Figure 7b seems to reflect the fact that reducing churn

allows firms to capture the returns from investments to reduce FTA rates.

Taken together these counterfactual policy experiments demonstrate that the combination of entry and

operating regulations operate by limiting the number of low quality firms (both by reducing their entry

and be expediting their exit). This yields two benefits: the immediate benefit of higher productivity

(lower FTA rates) from the higher percentage of quality firms, and the longer-term benefit of higher

25

innovation because the quality firms have sufficient scale to capture returns to their investments.

This result that markets with higher quality firms also have higher innovation is consistent with the

complementarity between firms’ initial productivity and the extent of innovation (Guadalupe, et al 2012).

While the selection in the case of Guadalupe et al comes from foreign partners, whereas our setting

features self-selection, the basic assumptions of the model hold as do the conclusions. Firms with higher

initial productivity innovate more and that innovation increases in expected market size. In their case the

larger market size comes from access to the foreign parent’s market. In our case the larger market size

comes from the fact that entry regulations leave fewer firms to split the market. This innovation increases

both firm sales and labor productivity.

While we don’t observe innovation directly as they do, we do observe (as do they) that productivity

increases with initial productivity as well as with lower levels of entry (available market share).

Moreover, the innovation they observe resembles the type innovation we discovered in conversations with

firms. In both cases the innovation that high-productivity firms engage in is process innovation—

moreover it is most likely to be a combination of new technology and new methods of organizing

production (new machines was negative in isolation in Guadalupe et al).

7. Discussion

A popular view of regulation is that it inhibits entry and innovation. This view is supported by recent

empirical work indicating that countries with lower administrative entry costs had higher entry rates as

well as higher productivity. However, wholesale reduction in regulation to achieve these benefits is ill-

advised. Regulations are designed primarily for first-order concerns such as correcting market failures,

protecting regional interests, and contributing to the moral aspirations of a society. The literature offers

little guidance regarding the impact of regulations with these first order concerns on entry and innovation.

Our goal was to extend prior work both by examining regulation’s adverse impact on entry and

innovation, and by looking at specific regulations rather than regulatory extent. The intent was to assess

the feasibility of designing regulation that preserves first order goals while minimizing adverse impact.

To do so, we examined the Schumpeterian cost of regulation in the United States Bail Bond industry.

The industry had natural experiment-like properties: roughly 14,000 owner-managed firms in 3,000

discrete markets that differ in demand and regulations, with outcome measures that differ from measures

driving firm entry. We employed structural models of the entrepreneur’s entry decision and the

incumbent’s continuation decision given local demand and regulations in this setting. This allowed us to

estimate the costs of specific regulations and to recover other parameter estimates associated with the

shared entrant and incumbent profit equation. We then employed these estimates in counterfactual policy

26

experiments. These analyses yielded a number of findings.

The primary finding is that a combination of entry and operating (continuation cost) regulations

reduces the amount of market churn (firm entry and exit), yet increases innovation relative to a setting

with no regulation. The beneficial effects of the entry cost/continuation cost regulatory bundle stem from

fewer low quality firms. This has an immediate benefit of lower average FTA rates in a market, but more

importantly leads to greater innovation. This occurs because high quality firms are left with sufficient

scale to capture returns to investment in FTA-reducing technology. While we do not model these

investments explicitly, they show up in differential learning rates across cohorts. Notably, these benefits

to innovation are above and beyond the intended benefits of regulation in this setting (i.e. prohibiting

transaction inequities and rights abuses). These results suggest it is possible to design regulations without

Schumpeterian cost on innovation.

Perhaps our most important result however is that “Schumpeterian cost” may need to be redefined.

Entry and innovation actually oppose one another in our setting. Not only does excess entry steal share

(Mankiew and Whinston 1986), in so doing it also destroys incentives to innovate because the expected

market share for the focal firm shrinks. Conversations with bail bond firms indicated that when entry

restrictions were relaxed in one market, it attracted more and lower quality firms. Rather than stimulating

innovation, this led to fee competition as well as illicit efforts to attract clients (such as advertising a 2%

fee up-front, but later requiring clients to pay an undisclosed 8% fee to recover their collateral). Such

illicit behavior under intense competition has been observed in other settings (Bennet and Pierce). Our

result that innovation decreases in the level of entry runs counter to the prior studies showing that

reducing administrative entry cost increases both entry and productivity. It is however consistent with the

results in Guadalupe et al discussed previously. We believe our results differ from the Djankov studies in

that the regulations in his and subsequent studies were subject to considerable regulatory capture

(administrative entry costs were correlated with countries’ corruption indices). Accordingly their removal

corrected a market failure. In contrast the primary intent of bail regulations is to reverse consumer abuses

and human rights abuses, thus those regulations appear less prone to regulatory capture.

Finally, as a byproduct of our entry model, we discovered a substantial disconnect between potential

entrant’s expected FTA rates and the market-level FTA rate. On average, potential entrants expected their

rate would be two-thirds lower than the market rate. This suggests significant entrepreneurial

overconfidence. Such overconfidence has been shown to increase in the level of uncertainty regarding

own performance (Wu and Knott 2006), which was considerable in this setting. The result is higher entry

rates than would otherwise seem rational (Camerer and Lovallo 1999, Wu and Knott 2006). Since excess

27

entry leads to higher FTA rates and lower innovation, an alternative solution to regulatory change is to

reduce uncertainty (and thereby overconfidence and excess entry) by publishing FTA rate distributions for

each market.

A number of caveats are in order. First, the unique market conditions in the bail bond industry may

limit the applicability of our results to other settings. The industry has low sunk and fixed costs, low

industry-specific human capital investments, and considerable part-time firm activity. Despite these

limitations, the bail bond industry was selected precisely because it offered econometric features not

found in many industries (i.e. diverse regulatory environments, distinct markets, exogenous demand, and

measurable outcomes that were distinct from outcomes driving entry decisions). These features allowed

us to investigate many types of regulations and their impact. So while there are limitations to this setting,

the study does provide considerable insight into the impact of individual regulations on entry and

innovation.

In addition to caveats regarding the setting are ones regarding the empirical approach. First, the

structural model presents a static and myopic decision framework. Potential entrants decide to enter

based on next year’s profitability and incumbents exit once they are no longer profitable. This matches the

logic in the bail bond industry (revealed in interviews) where there are restrictions preventing loans to

firms. However, while it is the best match for this industry, a static decision framework does not capture

the long-term strategies seen in many other entrepreneurial settings. Second, while the model allows for

considerable firm heterogeneity, it does not allow for sustained competitive advantage: two firms of the

same size and the same age differ only in their time variant shock. Third, given the number of firms in

the industry, there is no strategic interaction. A firm’s quantity decision has no impact on the decisions of

other firms.

Despite these caveats, the analysis highlights the need to carefully consider the impact of specific

regulations (rather than measures of regulatory extent) on entrepreneurship and innovation. This paper is

an initial attempt at doing so. Moreover our results suggest it is possible to design regulations without

imposing the Schumpeterian cost on innovation. They also provide preliminary results as to which

regulations best accomplish this goal.

28

ReferencesAghion, Philippe et al. “Competition and Innovation: An Inverted-U Relationship.” Quarterly Journal of Economics 120.2 (2005): 701-728.

Ayres, Ian, and Joel Waldfogel. “A Market Test for Race Discrimination in Bail Setting.” Stanford Law Review 46.5 (1994): 987-1047.  

Barseghyan, Levon. "Entry costs and cross-country differences in productivity and output." Journal of Economic Growth 13.2 (2008): 145-167.

Bertrand, Marianne, and Francis Kramarz. “Does Entry Regulation Hinder Job Creation? Evidence from the French Retail Industry*.” Quarterly Journal of Economics 117.4 (2002): 1369-1413.

Block, Michael K. and Steven J. Twist, "Evidence of a Failed System: A Study of the Performance of Pretrial Release Agencies in California," American Legislative Exchange Council, Washington, D.C., April 1995.

Branstetter, Lee G., Francisco Lima, Lowell J. Taylor, and Ana Venâncio. “Do Entry Regulations Deter Entrepreneurship and Job Creation? Evidence from Recent Reforms in Portugal.” NBER Working Paper No. 16473. October 2010.

Bruhn, Miriam. 2008. “License to Sell: Business Start-up Reform in Mexico.” Policy Research working paper 4538, World Bank, Washington DC.

Camerer, Colin, and Dan Lovallo. "Overconfidence and excess entry: An experimental approach." The American Economic Review 89.1 (1999): 306-318.

Chari, A. 2007. “License Reform in India: Theory and Evidence.” Department of Economics, Yale University, CT.

Ciccone, Antonio, and Elias Papaioannou. "Red tape and delayed entry." Journal of the European Economic Association 5.2‐3 (2007): 444-458.

Djankov, Simeon. 2002. “The Regulation of Entry.” Quarterly Journal of Economics, 117(1): 1-37

Dranove, David, and Ginger Zhe Jin. Quality disclosure and certification: Theory and practice. No. w15644. National Bureau of Economic Research, 2010.

Dreher, Axel, and Martin Gassebner. 2007. “Greasing the Wheels of Entrepreneurship? Impact of Regulations and Corruption on Firm Entry.” KOF Working Paper No. 166.

Dudley 2005

Dudley, Susan and Melinda Warren. “A Decade of Growth in the Regulators’ Budget: An Analysis of the U.S. Budget for Fiscal Years 2010 and 2011.” 2011 Annual Report.

Fisman, Raymond, and Virginia Sarria-Allende. 2004. “Regulation of Entry and the Distortion of Industrial Organization.” NBERWorking Paper 10929. Cambridge, MA.

Fombrun, Charles, and Mark Shanley. "What's in a name? Reputation building and corporate strategy." 29

Academy of management Journal (1990): 233-258.

Hannan, Michael, and John Freeman

Helland, Eric, and Alexander Tabarrok. “The Fugitive: Evidence on Public versus Private Law Enforcement from Bail Jumping.” Journal of Law and Economics 47.1 (2004): 93-122.

Henderson and Mitchell 1997

Hotz, V. Joseph & Mo Xiao. "The Impact of Regulations on the Supply and Quality of Care in Child Care Markets," American Economic Review, American Economic Association, vol. 101(5), pages 1775-1805, August 2011.

Kaplan, David, Eduardo Piedra, and Enrique Seira. 2007. “Entry Regulation and Business Start-ups: Evidence from Mexico.” Policy Research Working Paper 4322. World Bank, Washington DC.

Klapper, Leora, Luc Laeven, and Raghuram Rajan. “Entry regulation as a barrier to entrepreneurship.” Journal of Financial Economics 82.3 (2006): 591-629.

Kleiner, Morris M. and Kudrle, Robert T., Does Regulation Affect Economic Outcomes? The Case of Dentistry. Journal of Law and Economics, October 2000

Loayza, Norman V., Ana Mar Oviedo, and Luis Serven. 2005. "The Impact of Regulation on Growth and Informality: Cross-Country Evidence," Policy Research Working Paper Series 3623, The World Bank.

Mankiew and Whinston

McChesney, Fred S. "Rent Extraction and Rent Creation in the Economic Theory of Regulation," Journal of Legal Studies 16(1987): 101-118.

Monteiro, Joana, and Juliano Assuncao. 2006. “Outgrowing the Shadows: Estimating the Impact of Bureaucratic Simplification and Tax Cuts on Informality and Investment.” Pontifica Universidade Catolica, Department of Economics, Rio de Janeiro, Brazil.

Nicoletti, Giuseppe, and Stefano Scarpetta. "Regulation, productivity and growth: OECD evidence." Economic policy 18.36 (2003): 9-72.

Nelson 1991

Nelson, R. and S. Winter 1982 An Evolutionary Theory of Economic Change, The Belknap Press

Olley, G. Steven, and Ariel Pakes. "The Dynamics of Productivity in the Telecommunications Equipment Industry, Econometrica 64." n1297 (1996): 1263.

Peltzman, Sam. “Toward a More General Theory of Regulation.” Journal of Law and Economics 19.2 (1976): 211-240.

Pigou, Arthur. 1938. The Economics of Welfare, London: Macmillan.

Posner 197530

Schumpeter, Joseph A. Caplitalism, Socialism, and Democracy. 2d ed. New York: Harper & Bros., 1947.

SCOCAL, McDonough v. Goodcell, 13 Cal.2d 741

Shleifer, Andrei, and Robert W. Vishny. "Corruption." The Quarterly Journal of Economics 108.3 (1993): 599-617.

Stigler, George J. "The theory of economic regulation." The Bell journal of economics and management science (1971): 3-21.

Stinchcombe, Arthur L. 1965. “Social Structure and Organizations.” Pp. 142-193 in Handbook of Organizations, edited by James G. March. New York: RandMcNally.

Stuart, Toby E., and Olav Sorenson. "Liquidity events and the geographic distribution of entrepreneurial activity." Administrative Science Quarterly (2003): 175-201.

Thomas, Lacy Glenn. "Regulation and firm size: FDA impacts on innovation." The RAND Journal of Economics (1990): 497-517.

Verrochi, Richard, 2006. How To Start A Bail Bond Business And Become A Bail Bondsman.

Walls, Donald W., National Establishment Time-Series Database.

Wiggins, Steven N. “Product Quality Regulation and New Drug Introductions: Some New Evidence from the 1970s.” The Review of Economics and Statistics 63.4 (1981): 615-619.

Wolak, Frank A. “Regulatory Barriers to Lowering the Carbon Content of Energy Services.” http://www.kauffman.org/state-and-regulatory-hurdles-can-slow-clean-energy-innovation.aspx

World Bank. 2008. Doing Business 2009. Washington DC.

Wu, Brian, and Anne Marie Knott. "Entrepreneurial risk and market entry." Management Science 52.9 (2006): 1315-1330.

United States Department of Justice. Office of Justice Programs. Bureau of Justice Statistics. State Court Processing Statistics, 1990-2006: Felony Defendants in Large Urban Counties. ICPSR02038-v4. Ann Arbor, MI: Inter-university Consortium for Political and Social Research. 2010-09-03. doi:10.3886/ICPSR02038

Yakovlev, Evgeny, and Ekaterina Zhuravskaya. 2007. “Deregulation of Business.” New Economic School, Moscow.

31

Table 1. Number of Observations per County and Average Firm Counts and Arrests

Firm Count Arrest Count Freq.Muni Mean Std. Dev. Mean Std. Dev.

Atlanta 37.67 12.97 12184.00 3476.40 6Austin 21.67 6.03 5280.00 253.99 3

Baltimore 37.67 6.81 4024.00 395.51 3Baltimore City 59.33 14.01 12336.00 3197.59 3Birmingham 28.33 7.31 6068.00 1099.76 6

Bronx 7.00 0.00 12168.00 0.00 1Brooklyn 7.00 0.00 7224.00 0.00 1Buffalo 3.00 0.00 3367.20 998.66 5

Cincinnati 4.20 2.68 5702.40 562.75 5Clearwater 22.40 9.07 8097.60 1924.17 5Columbus 10.00 1.73 4944.00 937.23 3

Dallas 44.57 6.21 11396.57 2522.00 7Detroit 3.57 1.51 10416.00 5804.91 7El Paso 19.67 1.15 3796.00 103.46 3Fairfax 6.40 2.61 4106.40 619.86 5

Fort Worth 71.20 13.77 9134.40 1090.56 5Ft. Lauderdale 53.00 18.92 9504.00 2023.46 8

Honolulu 4.67 0.58 1984.00 508.74 3Houston 100.86 20.09 22038.86 4342.65 7

Indianapolis 24.20 3.70 10190.40 1575.46 5Jacksonville 11.00 0.00 11712.00 0.00 1Kansas City 26.33 4.16 3996.00 727.46 3Los Angeles 223.86 67.57 62016.00 11427.06 7

Martinez 20.50 2.12 3414.00 466.69 2Memphis 29.43 6.50 8557.71 1140.01 7

Miami 109.13 44.56 26130.00 4164.99 8New Haven 13.00 0.00 5712.00 0.00 1

Newark 13.50 5.21 13100.00 3877.66 6Norristown 2.75 3.20 2307.00 399.97 4

Oakland 31.33 3.98 7764.00 935.78 6Orange 66.33 22.51 11936.00 2885.66 6Orlando 38.00 11.53 10936.00 2333.35 3

Philadelphia 1.75 2.22 16800.00 952.77 4Phoenix 36.13 5.72 18435.00 4944.99 8Pittsburg 5.00 1.87 3052.80 413.40 5

Redwood City 14.00 1.73 2104.00 206.92 3Riverside 54.67 15.63 13168.00 2809.78 3Rochester 2.50 0.71 2670.00 161.22 2

Sacramento 35.60 12.93 10166.40 485.25 5Salt Lake City 17.40 11.55 4581.60 1432.74 5San Bernardino 38.25 12.43 9183.00 4131.82 8

San Diego 71.60 35.39 14678.40 2892.25 5San Francisco 12.50 3.51 7962.00 974.96 4

Santa Cruz 32.88 8.64 9729.00 1312.90 832

Seattle 10.80 2.17 6355.20 1123.19 5St. Louis 5.40 1.14 3796.80 687.55 5Suffolk 1.00 0.00 2880.00 0.00 1Tampa 41.20 5.07 8323.20 1711.54 5Tucson 5.67 1.21 5310.00 1031.77 6Ventura 17.00 1.00 2304.00 369.48 3

Washington DC 10.00 4.53 2918.40 773.39 5West Palm Beach 31.50 13.48 7032.00 1165.59 4

Total 37.23 45.06 10787.60 10879.90 239

33

Table 2. Summary Statistics

Variable Obs Mean Std. Dev. Min MaxRegional Controls

Population 239 4021904 3763282 680942 18700000Personal Income 239 $29,106 $7,814 $16,529 $54,755

Arrest Count 239 10788 10880 1440 77808Regional Firm Data

Entrants 239 4.674 6.061 0.000 43.000Exits 239 2.305 3.198 0.000 22.000

Incumbents 239 32.556 39.967 0.000 272.000Potential Entrants 212 147.910 168.961 0.000 847.000Total Employment 239 145.820 167.251 0.000 913.000

Mean Firm Age 236 9.674 4.602 0.600 38.000Mean Firm Employment 236 5.784 11.025 1.500 104.250

Regional Outcomes (Per Year)Surety Bond (# ) 239 2477.322 2518.452 24.000 12672.000Surety Bond ($) 239 $29,300,000 $45,200,000 $21,000 $389,000,000

FTA Surety Bond (#) 239 416.485 402.196 0.000 2256.000FTA Surety Bond ($) 239 $4,742,429 $9,558,431 $0 $109,000,000

FTA All (#%) 239 0.125 0.068 0.022 0.423FTA Surety (#%) 239 0.183 0.109 0.000 0.625FTA Surety ($%) 239 0.156 0.112 0.000 0.732

Fug Surety (#) 239 83.498 93.431 0.000 576.000Fug Surety ($) 239 $1,121,692 $2,465,207 $0 $24,200,000Fug All (#%) 239 0.258 0.128 0.038 0.750

Fug Surety (#%) 222 0.225 0.197 0.000 1.000Fug Surety ($%) 221 0.248 0.250 0.000 1.000

Regional Regulations - Model SortReg: Retain Contact 239 0.059 0.235 0.000 1.000

Reg: Place of Business 239 0.155 0.362 0.000 1.000Reg: Daily Records 239 0.251 0.435 0.000 1.000

Reg: Continuing Educ 239 0.234 0.424 0.000 1.000Reg: Background Check 239 0.155 0.362 0.000 1.000

Reg: Experience Req 239 0.192 0.395 0.000 1.000Reg: Post Bond 239 0.046 0.210 0.000 1.000

Reg: Exam 239 0.201 0.401 0.000 1.000

34

Table 3a. Parameter Estimates

Incumbent Decision

Incumbent FTA

α Age1 0.0196**

α Age2 0.0115**

α Age3 0.0090**

βempl 0.0017

FTAConstant 0.0205*

βFTAI 0.3164*

σ I 0.0338*

δ 0.1900

Constant Screening Variable Costs

γVC Cost Scalar 6.1159e-05*

Increasing Monitoring Variable Costs

γVC 403.3338*

γVC Contact -13.0307

γVC DailyRecords 25.3351

Fixed Costs

γ FCConstant 6.6854**

γ MaintainPublicBus 0.0733**

Continuation Costs

γContEduc 562.0726**

Entrant Decision

Entrant FTA

βFTAE 0.3654**

σ E 0.2439**

Entry Costs

γ AC 7.5705e+06**

γ Exam 2.8265e+04**

γCashBond 5.6758e+04**

γ Exper -0.5322

γBackground 7.3675e+04*

35

Table 3b. Estimated Costs of Parameter Estimates

Incumbent Decision

Constant Screening Variable Costs

ScreeningVC <$5

Increasing Monitoring Variable Costs

Monitoring VC 0.4%*q2

MaintainContact Regulation 0%

Daily Records Regulation 0%

Fixed Costs

¿Cost Constant $0

Maintain Public Business Regulation $0

Continuation Costs

Continuing Education Regulation $562

Entrant Decision

Entry Costs

Customer Acquisition Costs $195,000 (mean value)

Exam Regulation $28,265

InitialCash Bond Regulation $56,758 (mean value)

Experience Regulation $0

Backgroud Reg u lation $73,675

36

Figure 1. Overview of Bail Bond Process

37

Figure 2. Summaries of data distribution across municipalities

Figure 2a. Arrest Distribution

Figure 2b. Firm Count Distribution

38

Figure 2c. Entry distribution

Figure 2d. Exit distribution

39

Figure 2e. Market outcome distribution: Release rate

40

Figure 2f. Market outcome distribution: FTA rate

41

Figure 2g. Market outcome distribution: Fugitive Rate Conditional on FTA

42

Figure 3. Number of Entrants compared to pool of potential entrants

43

Figure 4. Estimated Results Regarding Firm FTA rate and Profitability

Figure 4a. Estimated Firm Level FTA Rate

Figure 4b. Estimated Firm Profitability (and added firm profitability shock)

44

Figure 5. Entry/Exit Results from Counterfactual Policy Experiments

45

Figure 5a. Entry Count Figure 5b. Exit Count

Figure 5c. Firm Count

Figure 6. Performance Outcomes from Counterfactual Policy Experiments

Figure 6a. Market Quantity Bailed ($100k)

Figure 6b. Market FTA Rate

46

Figure 7. Performance Outcomes from Counterfactual Policy Experiments – Operating Regulations Split

Figure 7a. Market Quantity Bailed ($100k)

Figure 7b. Market FTA Rate

47

Figure 8. Entry/Exit Results from Counterfactual Policy Experiments – Operating Regulations Split

Figure 9. Performance Outcomes from Counterfactual Policy Experiments – No Firm Learning48

Figure 8a. Entry Count Figure 8b. Exit Count

Figure 8c. Firm Count

Figure 9a. Market Quantity Bailed ($100k)

Figure 9b. Market FTA Rate

Figure 10. Entry/Exit Results from Counterfactual Policy Experiments – No Firm Learning

49

50

Figure 10a. Entry Count Figure 10b. Exit Count

Figure 10c. Firm Count