legal essay

Property Rights, Transaction Costs,

and the Limits of the Market.∗

Carmine Guerriero

University of Bologna

November 20, 2017

Abstract

Although the relevance of property rights and transaction costs for trade and inno- vation is well-known, our understanding of their origins and interplay is still limited. Within trade interactions, fully protecting the original owners’ property implies that some high-valuation potential buyers inefficiently refuse to buy it because of trans- action costs. When instead property rights are weak, low-valuation potential buyers inefficiently expropriate the original owners’ property. The trade-off between these two misallocations entails that property rights will be weaker the larger transaction costs are regardless of whether the latter are driven by frictions outside the control of traders or determined by the mix of the dispersion in their valuation and either the original owners’ market power or their privileged information. A similar conclusion holds true for an upstream firm’s property rights on an input necessary to a downstream firm to introduce a new technology and whose cost is random and ex ante non contractible. This time, transaction costs rise with the likelihood of a more productive technology. These implications survive if a group of traders/innovators has a larger political in- fluence on institutional design and if the disincentive to effort effect of weak property rights is considered. The model predictions are consistent with the negative effects of proxies for market frictions and failures on measures of the protection of personal, in- tellectual, and financial property that I document for a panel of 135 countries spanning the 2006-2015 period. This evidence suggests that the negative correlations between weak property rights and outcomes are partly spurious. Keywords : Property Rights; Transaction Costs; Market Frictions; Market Failures. JEL classification: D23; D40; D82; K11.

∗I would like to thank Andy Hanssen, Oliver Hart, Raffaella Paduano, Giorgio Zanarone and seminar participants at the university of Bologna for insightful comments. Address: Strada Maggiore 45, 40125 Bologna, Italy. E-mail: [email protected]

“Whenever transactions [. . .] are very expensive, [. . .] coercion is inherent [and] society will pick the

entitlement it deems favorable to the general welfare” [Calabresi and Melamed 1972, p. 1101].

“As soon as the land of any country has all become private property, the landlords, like all other men,

love to reap where they never sowed, and demand a rent even for its natural produce” [Smith 1976, p. 67].

1 Introduction

Albeit overwhelming evidence shows that strong property rights foster trade and innova-

tion, only recently economists have begun to study the determinants of this institution by

looking at the trade-off between the dispersed coercive power in a state of anarchy and the

predation by a central enforcement authority (Besley and Ghatak, 2010). Here, I qualify

these contributions by incorporating into the economics literature the key insight proposed by

a well-known legal scholarship (Calabresi and Melamed, 1972) and, notably, that incomplete

property rights can be efficient when transaction costs impede economic activities.

To characterize the trade-off between inefficient exclusion from trade/innovation and ex-

propriation guiding property rights selection, I study both the possibly consensual exchange

of economic value between its original owner and a potential buyer and a downstream firm’s

choice of whether to produce in-house through an old technology or to adopt a new one ne-

cessitating an upstream firm’s input. In the former case, I build on Guerriero (2016a), and I

study a society equally split into original owners, who have the same valuation for the only

good in the economy, and potential buyers, who are instead endowed with heterogeneous

valuations. The former are randomly matched to the latter by an intermediation technology

that allows each potential buyer to either obtain the good via consensual transfer by paying

the original owner’s valuation and bearing socially wasteful transaction costs or expropriate it

at no cost. I define property rights as the probability that an expropriated good is given back

to its original owner. When property is fully protected, some potential buyers with valuation

higher than that of the original owners are inefficiently excluded from trade due to transac-

tion costs. When instead the protection of property is weak, low-valuation potential buyers

inefficiently expropriate original owners. The trade-off between these two misallocations en-

tails that the protection of property, and thus the size of the market, will be more limited the

2

larger transaction costs are, regardless of whether the latter are driven by frictions outside

the control of traders—e.g., financial inefficiencies—or endogenously determined by the mix

of the dispersion in the traders’ valuations and either the original owners’ market power or

their privileged information.1 I reach similar conclusions when I turn to investment interac-

tions and I study a society equally split into randomly matched downstream and upstream

firms. The new technology requires the provision by the upstream firm of an input whose

random cost is realized after the downstream firm has borne an upfront expense. While both

costs and the firms’ payoffs are observable, unverifiable, and ex ante non contractible, only

the cost of the input is ex post contractible. Without loss of generality in particular, I main-

tain that after the preliminary phase the upstream firm has all the bargaining power and

so tries to hold-up his match by asking the high-cost realization. The downstream firm can

then either accept to pay the inflated cost, switch to the old technology, or turn to the legal

system. In the last case, the payoffs are determined by the upstream firms’ property rights,

which capture now the odds with which the downstream firm must accept the high cost.

This assumption squares with the idea that courts exploit unverifiable information to favor

the party they prefer regardless of the contract terms (Gennaioli, 2013). When the protec-

tion of property is strong, the risk of being held-up discourages the downstream firms from

innovating. When instead the upstream firms are only weakly protected, low-productivity

downstream firms inefficiently exploit the input. Balancing these two misallocations entails

that the protection of property rights (size of the market) will be weaker (larger) the greater

incomplete contracting costs are, i.e., the higher is the probability of low-cost realization.

Crucially, all the model implications survive when a group of traders/innovators has a

larger political influence on institutional design and when the disincentive to effort effect of

weak property rights is taken into account, and they apply to a wider range of cases. To

elaborate on the last point, the legal expropriation by a potential buyer (downstream firm)

is strategically similar to a condoned squatting of either a land or a building, the compulsory

1Transaction costs result then from “any impediments or costs of negotiating” [Calabresi and Melamed 1972, p. 1095]. A well-known law and economics literature considers this definition as inconsistent with the Coasean logic of organization (see however [Coase 1960, p. 2]), which would imply that transaction costs are instead those of “establishing and maintaining property rights” [Allen 1999, p. 898]. Nonetheless, ex-post misallocation and not ex-ante incentives to protect property and inquire about title is the key driver of institutional design if both prices and expropriation are endogenous (Dari-Mattiacci and Guerriero, 2015).

3

licensing granted to an intellectual property infringer, a majority stockholder’s legal capacity

to tunnel creditors’ and minority shareholders’ resources out of a firm, and the discretion of a

principal to lawfully segment a market (an agent’s power to successfully breach a contract).

To evaluate the central model predictions, I analyze a panel of 135 countries for which

the Executive Opinion Survey—EOS, hereafter—run by the World Economic Forum report

measures of the protection of personal, intellectual, and financial property and proxies for

transaction costs between 2006 and 2015 (see table 1). The EOS is the longest-running survey

of the opinions of business leaders on a broad range of topics for which statistics are unreli-

able, outdated, or nonexistent (WEF, 2015).2 Starting with the metrics of property rights, I

consider three indicators ranging between one and seven and gaging respectively the protec-

tion of generic property including financial assets—i.e., Property-Rights, the defense of in-

tellectual property rights including anti-counterfeiting measures—i.e., Intellectual-Property,

and the safeguard of the interests of minority shareholders, i.e., Shareholders-Protection (see

table 2 for the definition and sources of these and the remaining variables used in the present

study). Being these rights typically assigned to the downstream firms providing the main

productive resources [Burk and McDonnel 2007, p. 594], the three aforementioned indexes

constitute an inverse (direct) metrics of the property rights of the upstream (downstream)

firms. The maps in figure 1 visualize the strong correlations among these three measures—

nowhere lower than 0.81—and their sizable variation across countries. To draw these maps, I

averaged each variable over time, and then I divided the range of each average into four equal

intervals. In the empirical test instead, I use the continuously measured yearly variables.

Furthermore, I document in the Internet appendix that the evidence remains essentially the

same when I consider objective regulation-based measures of property rights and in partic-

ular the length of adverse possession of personal property and the investor protection index

developed by the Doing Business project.3 Turning to the proxies for transaction costs,

I consider measures of the severity of both market frictions—i.e., excessive regulation and

financial inefficiencies—and failures, i.e., lack of the competitiveness of corporate activity,

2The 2015 edition gathered more than 14,000 responses in 144 countries (WEF, 2015). I substitute missing observations with the closest data points. This choice is immaterial to the gist of the analysis.

3Adverse possession is a form of property rights acquisition such that a possessor becomes the legal owner of a good without the original owner’s consent, but by virtue of a sufficiently long, open, continuous, and notorious possession. The relative proxy however has no time variation (Dari-Mattiacci and Guerriero, 2015).

4

lemons-type distortions, and incomplete contracting costs due to asset specificities.

OLS estimates suggest that the protection of personal, intellectual, and financial property

is the weakest where market frictions and failures are the largest and asset specificities are

the most limited. Albeit consistent with the main model predictions, these results may be

capturing reverse causality, may be driven by the confounding effect of omitted variables, or

may be attenuated by the error in the measurement of transaction costs.

Accordingly, I pursue three strategies to determine if the correlations I uncover are,

in fact, causal. First, I control not only for country and year fixed effects, but also for

the development level, inclusiveness of political institutions, level of nonproduced output,

external and internal conflicts, and human capital. Considering these observables together

leaves the results almost intact. Second, I use insights from Altonji, Elder, and Taber (2005)

to calculate how much greater the influence of unobservable factors, relative to observables,

would need to be to explain away the negative links between property rights and transaction

costs. I find that the influence of unobservables would have to be on average sixty times

greater than that of all observables considered together, which seems unlikely. Finally,

I devise a 2SLS approach based on the positive dependence of market failures (incomplete

contracting costs) on the dispersion in the traders’ valuations (likelihood of a more productive

technology) foreseen by the model. To elaborate, I document strong negative (positive) first-

stages between the measures of exogenous transaction costs, the extent of market power, and

lemons-type distortions (asset specificities) and proxies for both the availability of the latest

technologies and the quality of math and science education. These results are consistent with

recent firm-level evidence suggesting that the distance to the technology frontier—and thus

the difference in the desirability of old and new products—is smaller in countries endowed

with a human capital more apt to absorb ideas and knowledge and in which therefore the

likelihood of a more productive technology is larger (Añón Higón et al., 2017). Conditional

on all observables, the 2SLS estimates are very similar to the OLS ones, and the validity

of the exclusion restriction is vindicated by two extra results. First, the Kleibergen-Paap

test always rejects under-identification at the 1 percent or less. Second, the overidentifying

restrictions cannot be rejected at a level nowhere lower than 5 percent, and the excluded

instruments have no direct role in the semi-reduced form regressions. All in all, this evidence

5

makes it difficult to envision that OLS correlations are spuriously driven by reverse causation,

a mechanism different from the one I model, or measurement error. Accordingly, I take them

as consistent with, if not proving, causality running from transaction costs to property rights.

The present study is strictly related to several strands of literature. First, a legacy

of recent contributions shows that weak property rights can be optimal in an endowment

economy (Jordan, 2006; Bar-Gill and Persico, 2016; Segal and Whinston, 2016; Arruñada,

Zanarone, and Garoupa, 2017). Not only do I extend this result to production economies

by clarifying how weak property rights can curb market frictions and partially solve market

failures,4 but I also highlight the relationships between the latter and either the dispersion

in the traders’ valuations or the likelihood of a more productive technology. Accordingly, my

analysis is also related to a body of research—still in its infancy—on endogenous transaction

costs (Dari-Mattiacci, 2012; Barry, Hatfield, and Kominers, 2014).5 Finally, Acemoglu and

Johnson (2005) focus on the relative importance of protecting property rights and enforcing

contracts, whereas I examine the determinants of the trade-off between these two institutional

strategies created by the possibility of transferring value without consent. Differently from

this contribution moreover, I also emphasize that weak property rights are society’s response

to the existence of sizable transaction costs and thus their negative correlations with economic

outcomes are partly spurious (see for a similar argument Aghion et al., [2010]).

The paper proceeds as follows. In section 2, I discuss the basic relationships between

property rights and transaction costs to motivate my more general model, which I illustrate

in section 3. Next, I assess whether these correlations are indeed causal in section 4. Finally,

I conclude in section 5, and I gather the proofs, tables, and figures in the appendix.

2 Property Rights and Transaction Costs

While all legal systems punish theft and embezzlement and provide remedies for the

dispossessed owners, the protection of property against private expropriation is almost never

complete as clearly displayed in the maps in figure 1. When transaction costs are sizable

4Albeit based on the same basic model, Guerriero (2016a) focuses instead on the relationship between property rights and the heterogeneity in the potential buyers’ valuations and does not consider either the issue of technology adoption or the related problem of identifying the drivers of incomplete contracting costs.

5While the former emphasizes the feedback of legal rules on transaction costs, the latter draw a link between transaction costs and how egalitarian is the distribution of the collective gains from Coasean bargaining.

6

indeed, a private party is often allowed to take someone else’s property with or without

paying compensation (Bouckaert and De Geest, 1995). To cast a first glance at whether this

pattern accurately describes also the EOS data, I consider one to seven indexes capturing

the severity of financial inefficiencies and the relevance of the market failures analyzed in the

model, i.e., market power, lemons-type distortions, and incomplete contracting costs.

Starting with the first one, I consider an index measuring the financial sector difficulties

in providing products and services to businesses, i.e., Unavailability-Financing. A value of

one suggests that it provides a wide variety, whereas a value of seven implies that it does not

provide them at all. For what concerns market power, I employ an index capturing the lack

of competitiveness of corporate activity, i.e., Market-Dominance. A value of one suggests

that corporate activity is spread among many firms, whereas a value of seven implies that

it is dominated by few business groups. Turning to lemons-type distortions, I focus on an

index falling with the extent of information used by buyers to make purchasing decisions, i.e.,

Asymmetric-Information. A value of one suggests that purchases are based on a sophisticated

analysis of attributes, whereas a value of seven implies that they are based solely on the lowest

price. Finally, I capture asset specificities with a measure of the competitive advantage of

the country’s companies in international markets, i.e., Asset-Specificities. A value of one

suggests that it is represented by low cost labor or natural resources, whereas a value of

seven implies that it is constituted by unique products and processes.

Conditional on country and year fixed effects, the partial correlations between the mea-

sures of property rights and Unavailability-Financing, Market-Dominance, and Asymmetric-

Information (Asset-Specificities ) in figures 2 to 4 (figure 5) reveal that there are strong nega-

tive (positive) links between the protection of the original owners’ (downstream firms’) prop-

erty and market frictions and failures (asset specificities), which are based on a sizable within

variation and not driven by a handful of abnormal observations.6 Columns (2), (3), (5), and

(6) of table 3 report estimates of these correlations from regressions obtained from the sub-

samples for which I observe all the additional controls illustrated in section 4.1.1 and with

standard errors allowing for clustering by country. In particular, a one-standard-deviation

rise in Unavailability-Financing, Market-Dominance, or Asymmetric-Information (Asset-

6My results are similar when I exclude the outliers identified through the Cook’s distance (Cook 1977).

7

Specificities )—roughly 1 (0.9)—is associated with about a 0.3(0.4)-standard-deviation—i.e.,

around 1—fall (rise) in the strength of property rights. These coefficients are all significant

at 5 percent or better and are coherent with the evidence coming from two alternative one

to seven indexes of market frictions and failures (see columns (1) and (4) of table 3). While

the former gages the burden of administrative requirements on firms—i.e., Over-Regulation,

the latter measures the impact of non-tariff barriers on the ability of imported goods to com-

pete with domestic ones, i.e., Trade-Barriers. More generally, I document in the Internet

appendix that the negative (positive) links between the original owners’ (downstream firms’)

property rights and market frictions and failures (asset specificities) survive when the latter

are proxied by either the necessity of bribery for the daily activity of firms, their difficulty to

obtain a bank loan, the hurdle they face when raising money by issuing shares on the stock

market, or the lack of competitiveness of local markets (extent of production sophistication).

Next, I present a model of the design of property rights on exchangeable value and

innovation activities rationalizing several legal issues included the evidence discussed so far.

3 Theory

I first analyze the case in which transaction costs are driven by frictions outside the

control of traders (innovators), and then I turn to characterize the scenario in which they

are endogenously determined by the mix of the dispersion in the traders’ valuations and

either the original owners’ market power or their privileged information (the likelihood of a

more productive technology and the incompleteness of contracts). Finally, I document that

the key model implications survive when the economy becomes “political” and thus someone

is excluded from the institutional design (see section 3.3.1) and when the original owners

can decide whether to produce or invest in their property (see section 3.3.2).

3.1 Property Rights and Exogenous Transaction Costs

Following Guerriero (2016a), I consider a society composed by a mass one of original

owners and a mass one of potential buyers, all having linear utility over a good x (see

footnote 14 for the risk aversion case). While the original owners value x at v, the potential

buyers have a valuation λ uniformly distributed over [ λ,λ ]

with l ≡ λ−λ, λ > v > λ, and

8

λm ≡ ( λ + λ

) /2. I employ the uniform distribution to obtain closed form solutions, but in

the appendix I show that the results hold under more generic density functions.

Original owners are randomly matched to potential buyers by an intermediation technol-

ogy allowing the latter to either obtain the good via consensual transfer by paying the former

v and bearing positive transaction costs α or expropriate it at no cost. The assumption that

expropriation is costless does not bear any loss of generality.7 α has no social value and

gages inefficiencies like the costs borne by potential buyers to borrow v or those necessary to

legalize the transfer and due to bribery, bargaining costs, excessive regulation, or a foreign

intermediary’s mark-up.8 I posit that α < min { v,λ−v

} , and I discuss this assumption in

footnotes 11, 12, and 13. An expropriated x is returned back with odds γ, which thus sum-

marizes the legal protection of the original owner’s property rights or the odds with which

the original owner is protected by a property rule as opposed to 1 −γ, which is the chance

with which the potential buyer is shielded by a liability rule (Calabresi and Melamed, 1972).

One glaring example of the type of expropriation just discussed is that of an intermediary

(agent) selling to a buyer in good-faith a good stolen (embezzled) from an original owner (his

principal) at a low price. As happens in the model, then the good is possibly given back to

its original owner and the strength of property rights reduces the probability that the buyer

consumes. More generally, when the potential buyer directly steals x, only with probability

γ the legal system forces her to hand the good back to its original owner. All in all, then

γ will be larger the longer the buyer needs to wait before acquiring ownership by adverse

possession, the stronger are the remedies in the original owner’s hands, and the more effective

is public enforcement (Dari-Mattiacci and Guerriero, 2015). The model however applies to

a large array of legal instances. First, the good can be envisaged as an input producing a

fixed value when transformed via the “old” technology in the hands of the original owners

and an uncertain one when the “new” technology is applied by the potential buyers. Second,

expropriation can assume the form of the squatting of either a piece of land or a building

(Brueckner and Selod, 2009).9 Third, the conflict between an original owner and a potential

7The model implications survive when expropriation entails either an expensive effort or a punishment and if the liability rule prescribes positive damages, provided that these cost are not too large (Guerriero, 2016a).

8Notably, the setup I consider can be interpreted as an efficient market maker assigning high-valuation potential buyers to original owners in exchange for the wasteful payment α.

9The good can be conceived as having a fixed value for the original owner and an uncertain one for the state

9

buyer can be reinterpreted as the one involving the creator of an intellectual property and an

infringer with 1−γ being the chance of either exhaustion of the related rights or compulsory

licensing (Ghosh, 2014; Bond and Saggi, 2017). Fourth, the same tension is isomorphic to

that between a creditor/minority shareholder and a majority stockholder who can tunnel

resources out of the firm through a complaisant management (Johnson et al., 2000). Finally,

the very same interaction is strategically similar to that setting an agent against a principal

in a segmented labor, financial or housing market with 1−γ gaging the principal’s capacity

to lawfully operate in the secondary segment (Piazzesi, Schneider, and Stroebel, 2017).

At time t0, γ is chosen to maximize the social welfare, which is the sum of the original

owners’ and potential buyers’ utilities. At time t1, individuals learn who they are and,

thus, their valuation. At time t2, individuals are matched by the intermediation technology.

Finally at time t3, any expropriated x is given back to its original owner with probability γ.

In evaluating the foregoing, two remarks should be stressed. First, an alternative in-

strument in the hands of society is to decrease α. This policy reduces market distortions

leading to complete property rights (see footnotes 11 and 12), but it is very costly to im-

plement because of political imperfections (Besley and Ghatak, 2010) and the opposition of

those agents who gain from larger transaction costs (see section 3.2 and Barry, Hatfield, and

Kominers, [2014]). Second, the model implication are unaffected when the original owners

have heterogeneous valuations and the traders can bargain over the price (Guerriero, 2016a).

3.1.1 Equilibrium

A potential buyer buys if her valuation λ net of the purchasing costs v + α is greater

than her expected payoff from expropriation (1 −γ) λ or λ ≥ λ̂ ≡ v+α γ

.10 When selecting

the optimal property rights level γ∗,11 society then maximizes the strictly concave function

∫ λ λ̂

λ−α l

dλ +

∫ λ̂ λ

(1 −γ) λ + γv l

dλ (1)

as in the instances regulated by “partial taking” law. Then, the state pays severance damages (1 −γ) v to the original owner to compensate a partial expropriation by the potential buyer (Sackman at al., 2013).

10Provided that only one good can be consumed, the results continue to stand even if those x that are purchased can be expropriated before consumption since all those willing to expropriate have already done it.

11The objective function in equation (1) is strictly concave for α < v. Should this inequality fail, γ∗ will be 1 (0)

for v < λm if the social welfare is larger at 1 (0) than it is at 0 (1) or whenever (v + α) 2−2

( αλ + vλ

) +λ2 >

(≤)0. This last inequality is more difficult to satisfy the larger α is because of the hypothesis v + α < λ.

10

for λ̂ < λ or γ∗ ∈ ( v+α λ , 1 ]

and the full expropriation social welfare (1 −γ) λm + γv ≡ WFE

whenever λ̂ ≥ λ and therefore all potential buyers prefer to expropriate.

Switching from complete to incomplete property rights—i.e., from γ∗ = 1 to γ∗ < 1—

has three effects: 1. it saves α at the cost of misallocating x with probability γ∗ for the

v+α ≤ λ < λ̂ matches; 2. it avoids misallocation with probability 1−γ∗ for the v ≤ λ < v+α

matches by expanding the consumption set of the potential buyers; 3. it misallocates x with

probability 1 −γ∗ for the λ < v matches. While this last effect is negative, the sum of the

first two is positive, provided that α is not too small.12 For λ̂ < λ, optimal property rights

are uniquely defined by the necessary and sufficient first-order condition

−2 dλ̂

dγ

( γ∗λ̂−γ∗v −α

) − ( λ̂2 −λ2

) + 2v

( λ̂−λ

) = 0 ↔ (γ∗)2 =

v2 −α2

λ (2v −λ) . (2)

Equation (2) implies that a rise in γ has a marginal effect, which is positive, and an infra-

marginal effect ∫ λ̂ λ v−λ l dλ = −(

λ̂−λ)(λ̂+λ−2v) 2l

, which can be negative only if v < λm. For

v ≥ λm then, optimal property rights are complete since also WFE rises with γ. For v < λm instead, WFE falls with γ and a γ∗ ≤ 1 is possible. It equals either the level defined by

equation (2) or 0 depending on which of the two maximizes the social welfare. The latter is

more likely the case the larger α is.13 Moreover, an interior γ∗ falls with α (see figure 2).

Intuitively, a rise in α has the infra-marginal effect of reducing the potential buyer’s

payoff for all the λ ≥ λ̂ matches and the marginal effect of raising λ̂ and thus misallocating

goods otherwise earmarked to high-valuation potential buyers with probability γ∗. Both

effects call for weaker property rights. Proposition 1 rephrases this idea:14

Proposition 1: Optimal property rights γ∗ weakly fall with the transaction costs α.

Proposition 1 is not only consistent with the preliminary evidence discussed in section 2

and the more in depth analysis in section 4, but it also sheds light on several institutionalized

cases of incomplete property rights. First, it rationalizes the post-war condoning of power

12Precisely, if α2 > (1 −γ) v2. Moreover, γ∗ < 1 whenever α > v −λ, which can be given my assumptions. 13The exact condition is γ∗

( λ−v

) > α, which is true for the lowest interior γ∗ = (v + α) λ

−1 if α < λ−v.

14If risk-averse, the traders who gain an expected utility lower than that prevailing under the certain scenario of full property rights incur also a loss u. Since all original owners (potential buyers) weakly prefer complete (incomplete) property rights, a rise in risk aversion is isomorphic to a fall in v and thus induces a weakly lower γ∗. Indeed, an increase in v has the infra-marginal effect of boosting the original owners’ payoff when property rights are protected and the marginal effect of raising λ̂. Both patterns imply a higher γ∗.

11

thefts by Indian farmers. In fact, despite its annual cost is around 1.5 percentage points of the

2012 GDP, local politicians have been strenuously defending it by asserting that collecting the

electricity invoices, which are mainly constituted by billing costs, would destroy subsistence

farming (Charnoz and Swain, 2012). Second, the cost of eviction or the inability to provide a

sufficient supply of housing because of regulatory requirements and speculative land-holding

are the most recurring justifications to the tolerance towards the roughly 40 percent share of

private lands invaded in developing countries and the two billion squatters estimated around

the world (Brueckner and Selod, 2009). Third, exhaustion of intellectual property rights

has been mainly implemented in high-transaction costs developing countries (Ghosh, 2014),

whereas the Article 31 of the TRIPS agreement allows the participants to impose compulsory

licensing if the commercial terms for a voluntary license are “unreasonable” (Bond and Saggi,

2017). Finally, Piazzesi, Schneider, and Stroebel (2017) conclude that financial costs explain

14 percent of the price gap between first and secondary housing market segments.

3.2 Property Rights and Endogenous Transaction Costs

Next, I evaluate several key instances of endogenously determined transaction costs.

3.2.1 Property Rights and Market Power

Following Guerriero (2016a), I consider the case in which the intermediation technology

is in the hands of the original owners and α is the mark-up on their valuation v. They select

it between t1 and t2 by maximizing the sum of the expected profits and the expected payoff

from consuming x when handed back, i.e., (v+α)(λ−λ̂)

l + γ∗v

(λ̂−λ) l

for v+α γ∗

= λ̂ < λ and γ∗v

otherwise. Then, α∗ can be positive only for λ̂ < λ when it equals α∗ = γ∗(λ+v)

2 − v and

rises with the strength of property rights, which in turn increases the original owners’ payoff

regardless of whether transfers are consensual. A rise in γ∗ reduces instead the potential

buyers’ payoff from expropriating and, through its positive impact on α∗, their utility from

buying x. Because of the linearity of preferences, the two effects cancel out and dλ̂ dγ∗

= 0 (see

the Appendix). Since α is now a transfer, γ∗ < 1 if the distortions in the potential buyers’

demand are sizable. Being λ̂ = λ+v 2 < λ, society maximizes the linear function

∫ λ λ̂

λ

l dλ +

∫ λ̂ λ

(1 −γ) λ + γv l

dλ, (3)

12

whose derivative with respect to γ for v < λm is negative for α ∗ > (2v −λ) γ∗ − v or

λ + 2λ > 3v and positive otherwise. Then, γ∗ possibly jumps from 0 to 1 as α∗ becomes

sufficiently small. For γ∗ = 1, α∗ = (λ−v)

2 and the mark-up increases with the difference

between high-valuation potential buyers’ and original owners’ valuation. Again, optimal

property rights must be complete if v ≥ λm. Proposition 2 takes stock of this analysis:

Proposition 2: The optimal property rights level γ∗ weakly decreases with the mark-up

α∗, which in turn is larger the higher the dispersion in the traders’ valuations λ−v is.

Proposition 2 not only rationalizes some of the results discussed in sections 2 and 4, but

it also helps make sense of growing evidence on the weakness of property rights on renewable

resources in the presence of market power. Common access to the fishing harvest together

with individual transferable quotas indeed is more often observed in Nova Scotia and New

Zealand where the fishermen’s market power is the strongest (Croutzet and Lasserre, 2017).15

3.2.2 Property Rights and Lemons-type Distortions

As noticed by Hasen and McAdams (1997), “theft may avoid the “lemons” problem.” Fol-

lowing Guerriero (2016a), I maintain here that the original owners have private information

on v, which is drawn from an uniform distribution with support [ 0,λ ]

and correlated with

the valuation of the potential buyers. In particular, ∆/2 of them value x at θv, 1 − ∆ > 0

of them have valuation αv, and the remainder gain from consuming x a payoff v/θ with

θ > 2 > α > 1. Here, θ gages the polarization of the potential buyers’ preferences, α

covers the role that exogenous transaction costs have in the basic setup and represents a

measure of the difference between the payoff of the middle-valuation potential buyers and

the valuation of the original owners, and a rise in ∆ constitutes a mean-preserving spread of

the λ distribution. These three parameters emphasize the impact of lemon-type distortions

on allocative efficiency in the regime L. To illustrate, middle and low-valuation potential

buyers expropriate (do not consume) if γ∗ < (=) 1 since the expected value of x is pL/2 at

the exogenous price pL. Original owners indeed sell only if v ≤ pL. High-valuation potential

buyers buy (expropriate) if γ∗ ≥ (<) 2 θ , and society maximizes the linear function

15Focusing on the fishing industry, Croutzet and Lasserre (2017) also show that the strength of the property rights on renewable resources should fall with the elasticity of output to effort and rise with the price elasticity of demand, the number of firms, and the difference between input and output values.

13

θλm ∆

2 + (1 − ∆) [(1 −γ) αλm + γλm] +

[ (1 −γ)

λm θ

+ γλm

] ∆

2 (4)

for γ∗ ≥ 2 θ

and (1 −γ) [ (1 − ∆) α + θ

2+1 θ

∆ 2

] λm + γλm otherwise. Hence, the derivative of

society’s problem with respect to γ is positive for γ∗ ≥ 2 θ

and both 1 − ∆ and α sufficiently

small and it is negative otherwise. Intuitively, as the share of middle-valuation potential

buyers for which x will be misallocated under complete property rights becomes less im-

portant because of a rise in heterogeneity, γ∗ jumps from 0 to 1 for α small. Similarly, a

sufficiently large rise in lemons-type distortions α induces a decrease in the optimal property

rights level from 1 to 0 for ∆ small. Proposition 3 recaps the analysis of this section:

Proposition 3: Optimal property rights γ∗ weakly fall with the lemon-type distortions

α, which in turn constitute a measure of the dispersion in the traders’ valuations.

Proposition 3 not only clears up some of the evidence illustrated in sections 2 and 4, but

it also helps explain recent stylized facts about sharing economies. In particular, Lee (2016)

builds on 2008 data on BitTorrent private-network file sharing activity and album sales to

conclude that the former has no impact (a positive effect) on top(mid)-tier artists’ sales for

which asymmetric information on perceived talent is the least (most) detrimental.

3.2.3 Property Rights and Incomplete Contracting Costs

Setup.—Consider now a mass one of upstream firms and a mass one of downstream firms.

The latter can employ two technologies to produce x. The “old” one does not necessitate

any input from the upstream firm, and it produces one unit of output of value δλ with δ < 1,

λ uniformly distributed over [ λ,λ ] , l ≡ λ−λ, λ > v > λ, and λm ≡

( λ + λ

) /2. The “new”

technology instead delivers one unit of output of value λ and requires first an upfront payment

v by the downstream firm, and then the provision by the upstream firm of an input whose

cost c is realized after the preliminary phase and equals 0 with probability 0 < α < 1 and

(1 − δ) λ otherwise.16 While both costs and the firms’ payoffs are observable, unverifiable,

and ex ante non contractible, only the cost of the input is ex post contractible. Without loss

of generality in particular, I maintain that after the preliminary phase the upstream firm has

16Should the low cost be µ < (1 − δ) λ, the social welfare will equal α ∫λ λ̂ λ−v l dλ+(1 −α)

∫λ λ̂ λ−v−µ

l dλ+

∫ λ̂ λ δλ l dλ

with λ̂ ≡ v+γα 1−δ ≥ λ̃ and the whole argument of this section will go through unchanged.

14

all the bargaining power. Hence, the cost uncertainty creates with odds α an “appropriable

quasi-rent” (1 − δ) λ, which is larger the more productive the new technology is relative to

the old one, and in turn the incentive for the upstream firm to always claim that c = (1 − δ) λ

(Barzel, 1989).17 Only with probability γ yet, the upstream firm is legally allowed to charge

the high cost. Hence, γ captures the strength of the upstream firm’s property rights on

his input relative to the power of the downstream firms’ property rights on x. When the

input is an idea or know-how for instance, v might be seen as the expenses supported by

the downstream firm to let the upstream firm understand how to incorporate his input in

the old production process, c is the cost of providing the input, and γ can be interpreted as

the probability that, in a lawsuit for breach of contract launched by the downstream firm, a

court allows the upstream firm to charge for the input provision (1 − δ) λ instead of 0. This

setup squares with the idea that courts display personal biases and arbitrarily evaluate the

evidence when certain states are hard to verify and so subject to interpretation (Gennaioli,

2013). Then, the parameter γ can also be seen as the share of pro-upstream firm courts.

The timing of the institutional and economic activities is as follows. At time t0, γ is

chosen to maximize the social welfare, which is the sum of upstream and downstream firms’

payoffs. At time t1, each downstream firm selects her preferred technology. If the old one

is employed, production is immediate, otherwise the downstream firm is matched to an

upstream one and the two firms sign a contract establishing that in time t2 the latter has

residual rights on his input and all bargaining power at renegotiation (Grossman and Hart,

1986). At time t2, first the preliminary phase is concluded, then the upstream firm makes a

take-it-or-leave-it request to the downstream firm. Next, the latter can either accept, reject

and turn to the old technology, or reject and exploit the available legal remedies. In this last

case, the downstream and upstream firms’ payoffs are determined by the prevailing γ.

Interpretation.—The parameter α should be seen as a general measure of incomplete

contracting costs. To elaborate, the conflict between upstream and downstream firms re-

sembles again that between a creditor/minority shareholder and a majority stockholder who

can tunnel value out of the firm through a complaisant management asserting that investing

17Under the usual assumption of Nash bargaining indeed, the upstream firm will require for c = 0 an input price equal to (1 − δ) λ

2 . Then, the new technology will be adopted for all λ larger than 2v

α(1−δ) , which is

inefficiently higher than λ̃, and the gist of the whole analysis will continue to hold true.

15

in a new activity needs resources that are ex ante non contractible. Similarly, the contrast

between the two firms is strategically equal to the interaction between a principal and an

agent who can breach their contract by claiming that the ex ante non contractible cost of

using a new technology is (1 − δ) λ instead of being 0 (Ganglmair, 2017).

Equilibrium.—For γ∗ = 1, the downstream firm obtains δλ−v with the new technology

and δλ with the old one. She prefers the latter and the social welfare is ∫ λ λ δλ l dλ, which is

inefficiently lower than that under the ex ante efficient equilibrium prescribing innovation

and a downstream’s (upstream’s) expected payoff of αλ + (1 −α) δλ (0) for λ ≥ λ̃ ≡ v α(1−δ) .

Incomplete property rights solve this hold-up failure since they allow the downstream firm

to obtain from adopting the new technology λ−v with probability 1−γ and δλ−v otherwise

and the upstream firm to get from the input provision the expected loss (1 −α) (1 − δ) λ

with probability 1 −γ and the expected gain α (1 − δ) λ otherwise. Hence, the downstream

firm will prefer the new technology for λ ≥ λ̂ ≡ v (1−δ)(1−γ) , and the upstream firm will produce

only if property rights are sufficiently strong or γ ∗

(1−γ∗) ≥ 1−α α ↔ γ∗ ≥ 1 −α. For γ∗ ≥ 1 −α

and λ̂ < λ, society maximizes the following strictly concave function

α

∫ λ λ̂

λ−v l

dλ + (1 −α) ∫ λ λ̂

δλ−v l

dλ +

∫ λ̂ λ

δλ

l dλ, (5)

which decreases with γ for −dλ̂ dγ

[ αλ̂ + (1 −α) δλ̂−v − δλ̂

] = − v

2(α−1+γ) (1−δ)(1−γ)3

≤ 0 or if γ ≥ 1−α.

To elaborate, a rise in γ ≥ 1−α has the welfare decreasing marginal effect of shrinking the set

of matches for which the downstream firm adopts the new technology. Yet, a γ∗ weakly larger

than 1 −α is necessary to push the upstream firm to participate in the production process.

Therefore, γ∗ = 1−α. In other words, the larger the odds α of a positive appropriable quasi-

rent are and thus the more severe asset specificities are, the lower γ∗ should be to convince

the downstream firm to adopt the new technology and the upstream firm to provide his

input.18 Crucially, the latter is indifferent between charging as input price the expected cost

and holding-up the downstream firm given γ∗. By assuming that if indifferent he selects the

former strategy, then the setup produces no hold-up in equilibrium. For α large moreover, the

18This incentive will be unnecessary, should the cost be certain. In this case, full property rights will contem- poraneously solve the hold-up failure and assure the upstream firm’s participation constraint.

16

welfare under incomplete property rights and innovation is larger than that with γ∗ = 1 and

without innovation. The latter also describes the scenario in which λ̂ ≥ λ ↔ v ≥ α (1 − δ) λ.

Proposition 4 summarizes the main conclusions of this section:

Proposition 4: Optimal property rights γ∗ fall with the incomplete contracting costs α,

which in turn measure the likelihood that the new technology is more productive.

Recognizing that both parties should be incentivized to foster technological diffusion,

proposition 4 entails that, differently from Grossman and Hart (1986), Hart and Moore

(1990), and Williamson (2010), asset allocation is optimized by alternatively leaving residual

claimant each party. This result has key ramifications for the theory of the firm.19

More generally, the findings of this section are not only coherent with some of the es-

timates discussed in sections 2 and 4, but they also help make sense of anecdotal evidence

on the positive link between the strength of the downstream firm’s intellectual property

rights—a high 1 − γ∗—and the severity of asset specificities. In particular, Burk and Mc-

Donnel (2007) document that stronger and easier to obtain downstream firm’s intellectual

property rights—i.e., trade secrecy and copyright instead of patents—are typically granted

whenever asset specificities are particularly severe, i.e., technological businesses, such as the

software industry, and the entertainment industries.20 Furthermore, these remedies allow the

downstream firm to impose on her employees and upstream partners non-disclosure agree-

ments, work-for-hire provisions, and non-compete clauses reducing their ability to exploit

proprietary information, i.e., a low γ∗ (Burk and McDonnel, 2007). “From the standpoint of

employee incentive [indeed], trade secrecy [and copyright are the] most expensive method of

protection [since they include] forms of intellectual capital most likely to become commingled

[. . . ] from the skills or knowledge of an employee” [Burk and McDonnel 2007, p. 609].

3.3 Robustness Checks

In this section, I document the robustness of the main model implications to two key

alternative assumptions, i.e., the possibility that part of the population is excluded from

19Guerriero (2017) shows that the model prediction survives if the firms can choose whether to vertically inte- grate and that this happens more often when asset specificities (property rights) are more severe (stronger).

20First, trade secrecy and copyright arise spontaneously, upon fixation of a creative work in a tangible medium of expression, whereas patent are granted only upon the disclosure of the claimed invention (Burk and McDonnel, 2007). Second, they receive a longer protection, i.e., respectively perpetual monopoly and the author’s life plus 50 to 100 years rather than the 20 years of patent protection (Burk and McDonnel, 2007).

17

institutional design and the possibility for the original owners to decide whether to produce

the good or to invest on it before trading (see the appendix for the relative proofs).

3.3.1 The Political Economy of Property Rights Protection

Thus far, I have examined the design of property rights on exchangeable value under a

perfect veil of ignorance, behind which everybody is identical, and the choice of property

rights on innovation by downstream and upstream firms with equal size and in turn political

power. Reality is however much less ideal. To evaluate the positive side of property rights

protection, I consider a situation in which the group of agents selecting γ either knows his

future role in the economy or is endowed with important inputs and so can exclude the rest

of the population from the social welfare maximization (Alesina, Aghion, and Trebbi, 2004).

It seems natural to think of these “insiders” as the original owners and the potential buyers

with moderate valuations. This approach incorporates into the model the idea put forward

by a growing literature on endogenous lobbying that the groups actively participating in

policy-making and shaping reforms are those most affected by them (Felli and Merlo, 2006;

Guerriero, 2016b). Here, these are original owners and middle-valuation potential buyers.

In the case of exogenously determined transaction costs then, γ∗ maximizes ∫ λ λ̂ λ−α l dλ +∫ λ̂

λ+� (1−γ)λ+γv

l dλ + γv�

l for λ̂ < λ and WFE − (1 −γ) �

2+2�λ 2l

otherwise when the excluded

potential buyers have valuation lower than λ+� and � not too large. Comparing this objective

function with that in equation (1) suggests that γ∗ still falls with α for � not too large, but

it is set inefficiently high, e.g., Zamindari system of taxation allowing Indian landowners to

evict evading tenants who were often more productive (Besley and Ghatak, 2010). If instead

to be excluded are the potential buyers whose valuation is higher than λ − �, γ∗ decreases

with α for � small and equals (is higher than) the γ∗ found in the basic setup if interior

(otherwise). When instead the transaction costs are driven by market power, α∗ continues

to be determined as in section 3.2.1 and thus a rise in γ still has only an infra-marginal

effect on society’s objective function. As a result, equation (3) implies that the analysis is

unchanged when high-valuation potential buyers are excluded from the institutional design

and that γ∗ is set too high and falls with α∗ for � small when instead low-valuation potential

buyers are kept out. Similarly, in the scenario of transaction costs determined by lemon-

18

type distortions, a gaze at equation (4) suggests that equilibrium property rights are again

inefficiently high and decrease with α provided that � is not too large. Finally, a glance

at equation (5) clarifies that excluding a sufficiently small group of either high- or low-λ

downstream firms from the institutional design does not affect at all the equilibrium when

the transaction costs are shaped by incomplete contracting inefficiencies.

3.3.2 The Disincentive to Effort Effect of Weak Property Rights

Production.—As highlighted in section 1, all the other contributions documenting that

weak property rights can be optimal focus on endowment economies. Here, I prove that the

mechanisms illustrated in sections 3.1, 3.2.1, and 3.2.2 survive when production is introduced

in the theoretical framework. This time, original owners decide between t1 and t2 whether

to produce x at the cost κ < v. Therefore, there is no production for γ∗ = 0 and λ̂ ≥ λ, but

there can be when the original owners’ expected utility is positive for γ∗ > 0 and so λ̂ < λ.

To elaborate, the original owners’ expected utility increases with γ∗, and therefore there

is a γ̃ such that x is produced only if γ∗ ≥ γ̃. Since production creates value also for the

potential buyers, society always selects the maximum γ̂ between γ̃ and γ∗ for κ not too large.

In the most interesting case of endogenous transaction costs finally, γ̃ weakly decreases with

α for λ sufficiently large and v not too small compared to the transaction costs α. This last

comparative statics leaves qualitatively unaffected the main model testable implication.

Investment.—The standard “security” argument for strong property rights claims that

expropriation induces a disincentive to invest (Besley and Ghatak, 2010). To understand

how the intuition affects the basic results, I analyze a regime I envisioning an investment

possibly implemented by the original owners between t1 and t2 at the cost ζ < v and raising

their valuation v and those of the potential buyers λ to respectively v (1 + ρ) and λ (1 + ρ)

with ρ > 0.21 Accordingly, this setup is similar to the instance of a production economy.

When transaction costs are exogenous, original owners invest only if γ∗I > 0 and their

expected utility v (1 + ρ) λ−λ̂I l

+ γ∗Iv (1 + ρ) λ̂I−λ l − ζ is weakly positive. Both γ∗I and λ̂I are

21When investment is continuous, the algebra becomes so tangled to be uninformative about the model ro- bustness except in the λ large and v small case when the basic analysis stands intact being dρ

dγ∗ small. The

core results will also survive, should potential buyers decide whether to invest. Then, γ∗ will be optimally weakened as in the case of “tracing” by an original owner of an asset substituted to the early property by an intermediary and possibly acquired by a potential buyer in good faith or of its proceeds (Smith, 1997).

19

as in section 3.1.1 with α/ (1 + ρ) in place of α, and thus γ∗I > γ ∗ and λ̂I < λ̂. Hence,

investment inducement weakly strengthens property rights protection. Since the original

owners’ expected payoff rises with γ∗I , there is a γ̃I such that investment goes through only if

optimal property rights are larger than γ̃I. For γ ∗ I > γ̃I, society picks γ

∗ I instead of γ

∗ if the

social welfare is larger at γ∗I with investment than it is at γ ∗ without. For γ∗I ≤ γ̃I instead, γ̃I

is preferred to γ∗ if the social welfare is larger at γ̃I with investment than it is at γ ∗ without.

To elaborate, investment always goes through for α sufficiently small compared to v and for

ρ (ζ) not too small (large). For v ≥ λm, then γ∗ = 1 and investment materializes.

In the case of transaction costs determined by the existence of market power, the fact that

the original owners’ expected utility is multiplied by 1 + ρ implies that α∗I = α ∗ (1 + ρ) and

so γ∗I = γ ∗ and λ̂I = λ̂. Since the original owners’ expected payoff rises with γ

∗ I , investment

realizes only if γ∗I is larger than γ̃I. As in the production case, γ̃I falls with α when the

mark-up is sufficiently small compared to v. Under the same condition and for ρ (ζ) not too

small (large), society picks the maximum γ̂I between γ ∗ I and γI since investment is socially

beneficial. Finally, γ∗I = 1 and thus investment is always successful for v ≥ λm.

For what concerns the scenario of transaction costs driven by lemon-type distortions, a

glance at equation (4) suggests that society’s objective function is now multiplied by 1 + ρ

and the analysis of the basic setup is unchanged provided that α < 2 1+ρ

. High-valuation

potential buyers buy for γ∗ ≥ 2 θ(1+ρ)

and expropriate otherwise. Once again, the original

owners’ expected utility rises with γ∗I , and so investment prevails only if optimal property

rights are larger than γ̃I. Society then selects γ̂I if investment is welfare enhancing and, in

particular, if its fixed cost is sufficiently small and its return is sufficiently large.

4 Evidence

The main model implication is that incomplete property rights are efficient when transac-

tion costs impede both trade and innovation whether or not constitution writers are benev-

olent and the disincentive to effort effect of weak property rights is taken into account. As

a consequence, the main model testable prediction can be stated as follows:

Prediction: The strength of the original owners’ (upstream firms’) property rights falls

with the severity of market frictions and failures (incomplete contracting costs), which rises

20

with the dispersion in the traders’ valuations (likelihood of a more productive technology).

The negative (positive) links between the measures of the protection of personal, in-

tellectual, and financial property and the proxies for market frictions and failures (asset

specificities) documented in section 2 are consistent with such a prediction. Yet, they may

be capturing reverse causality, they may be driven by the confounding effect of omitted

variables, or they may be attenuated by the error in the measurement of transaction costs.

4.1 Identifying Causal Relationships

I pursue several strategies to evaluate whether the correlations illustrated in section 2 are

causal. First, I control for the other main drivers of property rights discussed by the extant

literature. Second, I use selection on observables to assess the likelihood that the estimates

are driven by unobservables. Finally, I devise a 2SLS approach based on the aforementioned

positive dependence of market friction and failures (incomplete contracting costs) on the

dispersion in the traders’ valuations (likelihood of a more productive technology).22

4.1.1 Controlling for Observables

The key variables potentially omitted from the analysis of section 2 are those shaping

property rights through channels other than the trade-off between inefficient exclusion from

trade/innovation and expropriation and associated with transaction costs. They are the de-

velopment level, the inclusiveness of political institutions, the level of state capacity, internal

conflicts, and the extent of human capital accumulation. Next, I illustrate them in turn.

First, transaction costs could affect property rights through their adverse impact on

economic development, which in turn is related to the protection of private rights through

the modernization effect (Williamson, 2010). Accordingly, I consider the natural logarithm

of the output-side real GDP at chained PPPs in 2011 US dollar per capita, i.e., Income.

Second, less inclusive political institutions obstruct the distribution of the rents from

innovation and trade among non-elite (North, Wallis, and Weingast, 2009), while easing

the expropriation of the property of private citizens by both politicians and elite members

(Acemoglu and Johnson, 2005). To assess the importance of such a mechanism, I consider the

22For the data aggregated at the cross-sectional level moreover, I can confirm the core of the analysis and I cannot reject that transaction costs are exogenous and that the overidentifying restrictions hold.

21

constraints on the executive authority score from the POLITY IV dataset, i.e., Democracy.23

Third, low levels of non-produced output worsening the disincentive to effort effect, the

incidence of external wars, and German and Scandinavian legal origins are stable predictors

of a state capacity to exchange/innovate and protect private rights (see Besley and Persson,

[2009]; but also Guerriero, [2016b]). While I take into account the last factor through the

country fixed effects, I assess the role of the first two by controlling for the crude oil proved

reserves in barrels per capita—i.e., Reserves —and the share of previous half-century in which

the country was involved in external military conflicts, i.e., Conflict-External.

Fourth, inter-groups conflicts are related to both large transaction costs (Seitz, Tarasov,

and Zakharenko, 2015) and insecure private rights (Ashraf and Galor, 2013). I consider the

role of these connections by incorporating into the analysis the share of previous half-century

in which the country was involved in internal military conflicts, i.e., Conflict-Internal.

Fifth, more limited transaction costs provide incentives for investment in human capital

(Foss, 2011), which in turn helps shift labor from traditional to modern sectors and so

curb inequality and favor institutions that protect private property (Cervellati, Fortunato,

and Sunde, 2008). To evaluate these patterns, I control for the ratio of the total tertiary

enrollment regardless of age to the population of the relative age group, i.e., Human-Capital.

Finally, a culture of morality and the efficiency of public enforcement alter the incentives

of intermediaries and thus the balance between protecting property and enhancing the re-

liance on contracts (Dari-Mattiacci and Guerriero, 2015).24 Moreover, they may shape the

intensity of both exchange and innovation (Seitz, Tarasov, and Zakharenko, 2015). Esti-

mates reported in the Internet appendix reveal that considering these factors greatly lowers

the number of available observations but leaves essentially intact the gist of the analysis.

Including all these controls in the specification also factors in the heterogeneity in pref-

erence/productivity, which reduces the contribution to the social welfare of the utilities of

those middle-valuation(productivity) potential buyers (downstream firms) whose exclusion

23To illustrate, the score ranges between one and seven and assumes higher values when the holder of the executive power is accountable to the citizens and/or the government is constrained by checks and balances.

24While I proxy a culture of morality with the first principal component extracted from the level of generalized trust and the importance of respect self-reported to either the World Value Survey or the European Value Study, I measure the efficiency of public enforcement with the first principal component extracted from the number of police personnel and that of professional judges per 100,000 inhabitants as collected from the United Nations’ surveys of crime trends and the operations of criminal justice systems.

22

from trade (innovation) is more socially costly (Guerriero, 2016a). Since transaction costs

might be correlated with heterogeneity, this aspect of the identification strategy is key.

Panels (A) to (C) of table 4 report the estimates relative to the specifications with

dependent variables Property-Rights, Intellectual-Property, and Shareholders-Protection re-

spectively and considering all together the extra controls. The key observations are that

the coefficients on these regressors are generally jointly insignificant, whereas those on the

proxies for transaction costs are negative and significant at 5 percent or better and display a

magnitude almost equal to their counterparts in table 3. In the same spirit, results available

upon request reveal that the impacts of the proxies for transaction costs remain similar when

considered together and that the extra controls contribute only marginally to the total R2.25

4.1.2 Using Selection on Observables to Assess the Bias from Unobservables

Despite my attempts to control for relevant observable factors, OLS estimates may still

be biased by unobservables. To evaluate this issue, I calculate the index proposed by Altonji,

Elder, and Taber (2005) to measure how much stronger selection on unobservables, relative

to selection on observables, must be to explain away the full estimated effect.26 To see how

the index is calculated, consider a regression with a restricted set of control variables and one

with a full set of controls. Next, denote the estimate of the coefficient attached to a proxy for

transaction costs from the first regression βR, where R stands for “restricted,” and that from

the second regression βF , where F stands for “full.” Then, the index is the absolute value

of βF/(βR − βF ). The intuition behind the formula is as follows. The lower the absolute

value of (βR −βF ) is, the less the estimate of the relevant coefficient is affected by selection

on observables, and the stronger selection on unobservables needs to be to explain away the

entire effect. Moreover, the higher the absolute value of βF is, the greater is the effect that

needs to be explained away by selection on unobservables, and thus the higher is the index.

In table 5, I consider the specifications conditioning only for country and year fixed effects

and reported in table 3 as the restricted regressions and those controlling for all observables

in table 4 as the full regressions. The ratios calculated when the dependent variable is either

25According to the Shapley value decomposition, their contribution amounts on average only to the 13 percent, whereas the share explained by transaction costs (country and year fixed effects) is the 35 (52) percent.

26I use the version developed by Bellows and Miguel (2009) for possibly endogenous continuous variables.

23

Property-Rights, Intellectual-Property, or Shareholders-Protection are reported respectively

in columns (1) to (3). None of them is less than one, and the median and average ratios

are 29.2 and 60. Therefore, to attribute the entire OLS estimates of the coefficients on the

proxies for transaction costs to selection effects, selection on unobservables would have to be

on average sixty times greater than selection on all observables, which seems unlikely.

4.1.3 2SLS Estimates: Property Rights and Endogenous Transaction Costs

My final strategy is to use instrumental variables. This approach requires instruments

that are correlated with transaction costs but uncorrelated with any other dimension affecting

property rights. To achieve this goal, I build on the positive dependence of market failures

(incomplete contracting costs) on the dispersion in the traders’ valuations (likelihood of

a more productive technology), and I use as excluded instruments for transaction costs

two one to seven indexes. The first one captures the availability for firms of the latest

technologies—i.e., Technology-Availability —and the other measures the quality of math and

science education, i.e., Math-Science. This choice is consistent with Añón Higón et al.

(2017), who build on over one million firm-level data for 23 EU states between 2003 and

2014 to document that more labor skilled firms are closer to the technology frontier and this

pattern is more accentuated in service sectors. Such results have four crucial implications

for my 2SLS estimation approach. First, countries endowed with a human capital more apt

to absorb ideas and knowledge are those more likely to display not only smaller differences

in firms’ productivity but also a smaller dispersion in the payoffs of their users and thus

a smaller extent of market power and more limited lemons-type distortions. Second, these

are also the countries in which the likelihood of a more productive technology will be larger

(Acemoglu, Aghion, and Zilibotti, 2006) and so will be the severity of asset specificities.

Third, it reasonable to think that the very same drivers of market failures determine market

frictions. Accordingly, Acemoglu, Aghion, and Zilibotti (2006) suggest that credit market

imperfections are typical of countries further away from the technology frontier as opposed

to those close to approach it. Finally, it seems unlikely that a society’s inclination to spread

and adopt more effectively innovation is systematically related to the protection of property

rights conditional on the inclusiveness of political institutions, the level of state capacity,

24

internal conflicts, fixed effects, and especially the level of development and human capital.

Table 6 reports in panels (A) to (C) the 2SLS estimates of the specifications with

dependent variables respectively Property-Rights, Intellectual-Property, and Shareholders-

Protection and controlling for all observables. Starting from the first stages, both Technology-

Availability and Math-Science have always a negative (positive) and generally statistically

significant relationship with the proxies for market friction and failures (incomplete contract-

ing costs).27 Accordingly, I can always reject that the estimated equation is underidentified

at 1 percent. Turning to the second stages, the proxies for market friction and failures

(measure of incomplete contracting costs) have (has) always negative (positive) and strongly

significant—at 1 percent—effects (impact) on the measures of property rights protection and

the attached coefficients (coefficient) display(s) a magnitude very similar to those in table

4. Finally, I cannot reject that the overidentifying restrictions hold at a level nowhere lower

than 5 percent and I document in the Internet appendix that the excluded instruments have

no direct impact on property rights in the semi-reduced form regressions. Overall, these

patterns support the inference made above and confirm the validity of the model prediction.

5 Conclusions

This paper has developed and tested a model clarifying how the protection of property

rights is optimally weakened in the face of sizable costs of trading/innovating, and in particu-

lar market frictions and failures, and how the latter are more severe whenever the dispersion

in the traders’ valuations and the likelihood of a more productive technology are larger.

To characterize the general trade-off between inefficient exclusion from trade/innovation

and expropriation guiding property rights selection, I study both the possibly consensual ex-

change of economic value between its original owner and a potential buyer and a downstream

firm’s choice of whether to produce in-house through an old technology or to adopt a new

one necessitating an upstream firm’s input. In the former case, fully protecting the original

owners’ property implies that some high-valuation potential buyers inefficiently refuse to buy

it because of transaction costs. When instead property rights are weak, low-valuation po-

27Consistent with this evidence, Dari-Mattiacci and Guerriero (2015) document that a culture of self-reliance, which fosters innovation, is linked to stronger original owner’s property rights mainly through its positive impact on the intermediaries’ morality and its negative effect on the strength of law enforcement.

25

tential buyers inefficiently expropriate the original owners’ property. The trade-off between

these two misallocations entails that the strength of property rights, and thus the size of the

market, will be more limited the larger transaction costs are regardless of whether the latter

are driven by frictions outside the control of traders/innovators or determined by the mix of

the dispersion in traders’ valuations and either the original owners’ market power or their

privileged information. A similar conclusion holds true for the upstream firm’s property

rights on his input. Being the relative cost random and ex ante non contractible, a strong

protection of the upstream firms’ property discourages the downstream firms from innovat-

ing because of the risk of being held-up. When instead the upstream firms are only weakly

protected, low-productivity downstream firms inefficiently exploit the input. Balancing these

two misallocations entails that the protection of the upstream firms’ property rights (size

of the market) will be weaker (larger) the greater incomplete contracting costs are, i.e., the

higher is the probability of low-cost realization. Crucially, these implications survive when a

group of traders/innovators has a larger political influence on institutional design and when

the disincentive to effort effect of weak property rights is taken into account.

To evaluate the central model predictions, I focus on a panel of 135 countries spanning

the 2006-2015 period. OLS estimates suggest that the protection of the original owners’

(downstream firms’) property rights is the weakest (strongest) where market frictions—i.e.,

excessive regulation and financial inefficiencies—and failures—i.e., lack of the competitive-

ness of corporate activity and lemons-type distortions (incomplete contracting costs due to

asset specificities)—are the largest. To determine if these relationships are indeed causal,

I pursue three strategies. First, I control not only for country and year fixed effects, but

also for the development level, the inclusiveness of political institutions, the strength of state

capacity, internal conflicts, and the level of human capital. Considering these observables

together leaves the results almost intact. Second, I calculate that the influence of unob-

servables would have to be on average sixty times greater than that of all observables to

explain away the negative links between property rights and transaction costs. Finally, I

use as excluded instruments proxies for both the availability of the latest technologies and

the quality of math and science education. Conditional on all observables, the 2SLS esti-

mates are strongly consistent with the OLS ones. In addition, the validity of the exclusion

26

restriction is vindicated by the canonical under-identification and over-identification tests.

I close by highlighting how two central results of my analysis open key avenues for future

research. First, the tendency of property rights toward optimality does not imply that the

existing legal variation is irrelevant and thus does not warrant reforms. On the contrary,

the model reveals that special interests can distort the design of property rights away from

optimality when the political process is less than perfect. Second, weak property rights are

society’s response to the existence of sizable transaction costs, which in turn are driven by the

dispersion in the traders’ valuations and the likelihood of a more productive technology, and

therefore their negative correlations with economic outcomes are—at least partly—spurious.

As a consequence, further research on the relationships among endogenous property rights,

endogenous transaction costs, and economic outcomes is needed.

27

Appendix

Property Rights and Market Power

The first-order condition of society’s problem is −λ̂−λ 2l

( λ̂ + λ− 2v

) = 0, whose left-hand

side is the infra-marginal effect of a rise in γ. The latter can be negative only for v < λm. �

Property Rights and Lemons-type Distortions

The high-valuation potential buyers’ expected payoff from buying θpL 2 − pL is weakly

greater than her expected payoff from expropriation (1 −γ) θpL 2

for γ∗ higher than 2 θ <

1 being θ > 2. The middle(low)-valuation potential buyers instead do not buy because

αpL 2 −pL(pL2θ −pL) < 0. For γ

∗ ≤ 2 θ , the derivative of society’s objective function with respect

to γ is − [ (1 − ∆) (α− 1) + (θ−1)

2

θ ∆ 2

] λm, which is negative. For γ

∗ > 2 θ

instead, it equals

− [ (1 − ∆) (α− 1) − θ−1

θ ∆ 2

] λm, which falls with α, rises with ∆, is negative for ∆ → 0, and

positive for ∆ → 1. γ∗ possibly jumps from 0 to 1 for α (∆) sufficiently small (large). �

Considering a Generic Probability Density Function of the Potential Buyers’ Valuation

In the case of exogenous transaction costs, potential buyers value x at λ ∈ [ λ,λ ]

dis-

tributed according to the log-concave probability density function f with cumulative distri-

bution function F. Then, γ∗ maximizes ∫ λ λ̂

(λ−α) dF (λ) + ∫ λ̂ λ

[(1 −γ) λ + γv] dF (λ) for

λ̂ < λ and WFE otherwise. For λ̂ < λ, γ∗ > 0 is defined by 1−γ ∗

γ∗ vλ̂f

( λ̂ ) − ( λ̂−v

) F ( λ̂ )

+∫ λ̂ λ F (λ) dλ = 0 and society’s objective function is sub-modular in γ and α when

f′(λ̂) f(λ̂)

<

α v(v+α)

γ∗

1−γ∗ . While the right-hand side of this inequality is greater than α

v(λ−v−α) , its left-hand

side is lower than f ′

f (v + α) since log-concavity of f implies a decreasing f

′

f (Dharmadhikari

and Joag-Dev, 1988). Hence, the inequality is true and thus dγ ∗

dα ≤ 0 for f

′

f (v + α) < α

v(λ−v−α)

or equally if v is sufficiently large (see section 3.1 for the corresponding restriction in the basic

setup). To understand this last remark, notice that the right-hand side of the sufficient condi-

tion increases with v for 2v > λ−α, whereas its left-hand side is negative for v+α larger than

the mode of λ, being every log-concave density function defined on a real support unimodal

(Dharmadhikari and Joag-Dev, 1988). For λ̂ < λ, the infra-marginal effect of a rise in γ is[ v −E

( λ|λ ≤ λ̂

)] F ( λ̂ )

, which can be negative only if v < λm = E (λ) ≥ E ( λ|λ ≤ λ̂

) .

For λ̂ < λ and v ≥ λm therefore, γ∗ = 1. For λ̂ < λ and v < λm instead, γ∗ can jump from

0 to the interior solution whenever ∫ λ λ̂

(λ−α) dF (λ) + ∫ λ̂ λ

[(1 −γ∗) λ + γ∗v] dF (λ) > λm.

28

This last condition is more difficult to satisfy the larger the transaction costs are since the

derivative of its left-hand side with respect to α equals −1−γ ∗

γ∗ vf ( λ̂ ) −F

( λ )

+ F ( λ̂ ) < 0.

Next, I show that the results produced by the other setups survive to considering any f.

When the original owners have market power, two observations are key. First, the mark-

up α∗ maximizes now the original owners’ expected payoff ∫ λ λ̂

(v + α) dF (λ) + ∫ λ̂ λ γ∗vdF (λ),

whose derivative with respect to α is 1 − F ( λ̂ ) − ( λ̂−v

) f ( λ̂ )

. Hence, dα ∗

dγ = λ̂ and

thus dλ̂ dγ

= ( dα∗

dγ γ −v −α∗

) 1 γ2

= 0. Again, the effect of a rise in the strength of optimal

property rights on society’s objective function ∫ λ λ̂ λdF (λ) +

∫ λ̂ λ

[(1 −γ) λ + γv] dF (λ) equals

−λ̂−λ 2

( λ̂ + λ− 2v

) f ( λ̂ )

, which is negative for v+α ∗

γ∗ + λ − 2v > 0 and positive otherwise.

As in the baseline case of a uniform distribution f, γ∗ jumps from 0 to 1 as α∗, which is

either 0 or its maximum value α = min { v,λ−v

} , becomes sufficiently small.

Turning to the scenario of transaction costs driven by lemon-type distortions, it follows

from society’s objective function that only the mean of the potential buyers’ valuation dis-

tribution λm matters. This remark highlights the full generality of the analysis.

For what finally concerns the case of transaction costs determined by incomplete con-

tracting inefficiencies, society maximizes α ∫ λ λ̂

(λ−v) dF (λ) + (1 −α) ∫ λ λ̂

(δλ−v) dF (λ) +∫ λ̂ λ δλdF (λ), which strictly increases with γ when −dλ̂

dγ

[ αλ̂ + (1 −α) δλ̂−v − δλ̂

] f ( λ̂ )

=

− v 2(α−1+γ)

(1−δ)(1−γ)3 f ( λ̂ ) > 0 or if γ < 1 − α and weakly decreases otherwise. Hence, the unique

and global solution is γ∗ = 1−α as in the baseline scenario of a uniform distribution f. �

The Political Economy of Property Rights Protection

In the case of exogenous transaction costs and excluded low-valuation potential buyers,

the derivative of the objective function is v 2−α2 (γ∗)2

+ (λ + �) 2 −2vλ = 0 for λ̂ < λ and v−λm +

2�λ+�2

2l otherwise, the infra-marginal effect of a rise in γ for λ̂ < λ is −(

λ̂−λ−�)(λ̂+λ+�−2v) 2l

+ 2v� 2l

,

and the second-order conditions are as in the basic setup. Thus, γ∗ is set inefficiently high,

falls with α provided that (λ + �) 2−2vλ < 0 or � not too large, and possibly jumps from 0 to

the interior solution for α sufficiently small under the same condition discussed in the basic

setup. When instead to be excluded are high-valuation potential buyers, society maximizes

the objective function ∫ λ−� λ̂

λ−α l dλ +

∫ λ̂ λ

(1−γ)λ+γv l

dλ for λ̂ < λ and WFE −(1 −γ) 2�λ−� 2

2l oth-

erwise. Both the infra-marginal effect of a rise in γ and the first and second-order conditions

for a positive solution are as in the baseline analysis, whereas for λ̂ ≥ λ the first-order con-

29

dition is v−λm + 2�λ−� 2

2l and thus leads again to an inefficiently large γ∗. The condition such

that γ∗ possibly jumps from 0 to the positive solution is easier to satisfy the smaller α is for

α < γ∗ ( λ−v − �

) and thus under the basic model restrictions on α for � small.

When original owners have market power, the only relevant case is λ < λ̂ and society max-

imizes ∫ λ λ̂ λ l dλ +

∫ λ̂ λ+�

(1−γ)λ+γv l

dλ, whose derivative with respect to γ is −( λ̂−λ−�)(λ̂+λ+�−2v)

2l +

2v� 2l

. Hence, equilibrium property rights jump from 0 to 1 for α∗ and � sufficiently small.

Turning to the case of transaction costs driven by lemon-type distortions and low(high)-

valuation potential buyers not participating in the institutional design, society’s objec-

tive function is that of the basic setup with (1 −γ) λm θ

( ∆ 2 − � )

in place of (1 −γ) λm θ

∆ 2

((1 −γ) θλm (

∆ 2 − � )

in place of (1 −γ) θλm ∆2 for γ ∗ ≤ 2

θ and θλm

( ∆ 2 − � )

in place of

θλm ∆ 2

otherwise). While for γ∗ ≤ 2 θ

the first-order condition of society’s problem equals

− [ (1 − ∆) (α− 1) + (θ−1)

2

θ ∆ 2 − �

θ

] λm (−

[ (1 − ∆) (α− 1) + (θ−1)

2

θ ∆ 2 − �θ

] λm) and so it is

negative for � not too large, for γ∗ > 2 θ

it equals − [ (1 − ∆) (α− 1) − θ−1

θ ∆ 2 − �

θ

] λm and

thus is negative for ∆ and � small and α large (the baseline case analysis applies).

For what finally concerns the case of transaction costs determined by incomplete con-

tracting inefficiencies, the equilibrium is the same whether or not a sufficiently small group

of either high- or low-λ downstream firms is kept out of the institutional design.

Crucially, the political economy case can be generalized to the consideration of generic

distributions of λ and/or the inclusion of production and investment activities. �

Production

In the case of exogenous transaction costs, the original owners’ expected utility equals v(λ−λ̂)

l +

γ∗v(λ̂−λ) l

−κ and thus γ̃ is implicitly defined by v[λ−(v+α)γ̃−1]

l +

γ̃v[(v+α)γ̃−1−λ] l

= κ,

whose left-hand side rises with γ̃ because −dλ̂ dγ̃

v l

(1 − γ̃) + v(λ̂−λ)

l = vλ̂

γ̃l (1 − γ̃) +

v(λ̂−λ) l ≥

0. Thus, production realizes for γ∗ ≥ γ̃. When the original owners have market power,

their expected utility is (v+α∗)(λ−λ̂)

l +

γ∗v(λ̂−λ) l

−κ, and γ̃ is defined by (v+α∗)[λ−(v+α∗)γ̃−1]

l +

γ̃v[(v+α∗)γ̃−1−λ] l

= κ, whose left-hand side rises with γ̃ because −dλ̂ dγ̃

v(1−γ̃)+α∗ l

+ v(λ̂−λ)

l =

λ̂ γ̃l

[v (1 − γ̃) + α∗]+ v(λ̂−λ)

l ≥ 0 and rises with α if −dλ̂

dα v(1−γ̃)+α∗

l + λ−λ̂

l = −v(1−γ̃)+α

∗

γ̃l + λ−λ̂

l ≥ 0

or for λ large and v not too small compared to α∗ and so γ∗ → 1. Hence, production realizes

for γ∗ ≥ γ̃, and dγ̃ dα ≤ 0. For what finally concerns the scenario of transaction costs driven

by lemon-type distortions, the original owners’ expected utility is ∆ 2 pL +

( 1 − ∆

2

) γ∗λm −κ

30

and γ̃ is independent of α being equal to 2κ−∆pL (2−∆)λm

. Once again, x is produced if γ∗ ≥ γ̃. �

Investment

With exogenous transaction costs, potential buyers buy if λ (1 + ρ) − v (1 + ρ) − α ≥

(1 −γI) λ or λ ≥ λ̂I ≡ v+α(1+ρ)−1

γI and thus society’s objective function is

∫ λ λ̂I

λ(1+ρ)−α l

dλ +

(1 + ρ) ∫ λ̂I λ

(1−γ)λ+γv l

dλ, γ∗I = v2−( α1+ρ)

2

λ(2v−λ) , and the original owners’ expected payoff from in-

vestment equals v (1 + ρ) (λ−λ̂I)

l + γ∗Iv (1 + ρ)

λ̂I−λ l − ζ, whose derivative with respect to γ∗I

is −dλ̂I dγ∗ I v (1 + ρ)

(1−γ∗I ) l

+ v (1 + ρ) λ̂I−λ l ≥ 0. For γ∗I > γ̃I, society picks γ

∗ I if the social

welfare is larger at γ∗I with investment than it is at γ ∗ without or whenever the inequality∫ λ

λ̂I

λ(1+ρ)−α l

dλ− ∫ λ λ̂ λ−α l dλ + (1 + ρ)

∫ λ̂I λ

(1−γ∗I )λ+γ ∗ I v

l dλ−

∫ λ̂ λ

(1−γ∗)λ+γ∗v l

dλ−ζ ≥ 0 holds. The

left hand side of this condition is larger than (1 + ρ) ( γ∗λ̂2 −γ∗I λ̂

2 I

) + (γ∗I −γ

∗) λ2 (1 + ρ) +

2v [ (1 + ρ) γ∗I

( λ̂I −λ

) −γ∗

( λ̂−λ

)] −2α

( λ̂− λ̂I

) −ζl, which is positive for α sufficiently

small compared to v and so γ∗I → γ ∗ and for ρ (ζ) not too small (large). A similar analysis

applies to the γ∗I ≤ γ̃I scenario when society chooses γ̃I and not γ ∗ if the social welfare is

larger at γ̃I with investment than it is at γ ∗ without investment. In the former case, λ̂I is

evaluated at γ̃I. If v ≥ λm finally, then γ∗ = 1 and investment is certain.

When the original owners have market power, their expected utility rises with and falls

with α if −dλ̂ dα

v(1−γ̃I)(1+ρ)+α∗ l

+ λ−λ̂ l

= −v(1−γ̃I)+α ∗

γ̃I(1+ρ)l + λ−λ̂

l ≥ 0 or if v is not too small compared

to α∗ since then γ∗ → 1. Hence, γ̃I falls with α when the mark-up is sufficiently small

compared to v and society selects γ̂I if investment is welfare enhancing. For γ ∗ > γ̃I, this

is the case if ρ ∫ λ λ̂ λ l dλ + ρ

∫ λ̂ λ

(1−γ∗)λ+γ∗v l

dλ − ζ ≥ 0 and therefore for ρ (ζ) not too small

(large). For γ∗ ≤ γ̃I instead, the analysis is as in the exogenous transaction costs case and

γ̃I is selected for α small compared to v and so γ ∗ I → γ

∗ and for ρ (ζ) not too small (large).

For what finally concerns the scenario of transaction costs driven by lemon-type distor-

tions, the original owners’ expected utility is ∆ 2 pL (1 + ρ) +

( 1 − ∆

2

) γ∗λm (1 + ρ) − κ, and

γ̃I = 2κ−∆pL(1+ρ) (2−∆)λm(1+ρ)

, and γ∗I is as in the basic setup. Investment prevails only if optimal property

rights are larger than γ̃I, and society then selects γ̂I if investment is welfare enhancing. For

γ∗ > γ̃I, this is the case if θλm ∆ 2

+(1 − ∆) [(1 −γ∗) αλm + γ∗λm]+ [ (1 −γ∗) λm

θ + γ∗λm

] ∆ 2 ≥

ζρ−1 and so for ρ (ζ) not too small (large). For γ∗ ≤ γ̃I instead, society selects γ̃I if

ρθλm ∆ 2

+ [(1 + ρ) (1 − γ̃I) − (1 −γ∗)] [ (1 − ∆) αλm + λmθ

∆ 2

] + ( 1 − ∆

2

) (γ̃I −γ∗) λm − ζ ≥ 0

and so for α small relative to v and so γ∗I → γ ∗ and for ρ (ζ) not too small (large). �

31

References

Acemoglu, Daron, and Simon Johnson. 2005. “Unbundling Institutions.” Journal of Politi-

cal Economy, 113: 949-995.

Acemoglu, Daron, Philippe Aghion, and Fabrizio Zilibotti. 2006. “Distance to Frontier,

Selection, and Economic Growth.” Journal of the European Economic Association, 4:

37-74.

Aghion, Philippe, Yann Algan, Pierre Cahuc, and Andrei Shleifer. 2010. “Regulation and

Distrust.” Quarterly Journal of Economics, 125: 1015-1049.

Allen, Douglas W. 2000. “Transaction Costs.” In Encyclopedia of Law and Economics, Vol.

I. Boudewijn Bouckaert and Gerrit De Geest, eds. Chelthenham: Edward Elgar Press:

893-926.

Altonji, Joseph G., Todd E. Elder, and Christopher R. Taber. 2005. “Selection on Observed

and Unobserved Variables: Assessing the Effectiveness of Catholic Schools.” Journal of

Political Economy, 113: 151-184.

Añón Higón, M. Dolores, Juan A. Máñez, Maŕıa E. Rochina-Barrachina, Amparo Sanchis,

and Juan A. Sanchis. 2017. “The Determinants of the Distance to TFP Frontier among

European Firms.” Unpublished.

Arruñada, Benito, Giorgio Zanarone, and Nuno Garoupa. 2017. “Property Rights in Se-

quential Exchange.” Unpublished.

Ashraf, Quamrul, and Oded Galor. 2013. “Genetic Diversity and the Origins of Cultural

Fragmentation.” American Economic Review, 103: 528-533.

Bar-Gill, Oren, and Nicola Persico. 2016. “Exchange Efficiency with Weak Ownership

Rights.” American Economic Journal: Microeconomics, 8: 230-267.

Barry, Jordan M., John William Hatfield, and Scott Duke Kominers. 2014. “Coasean Keep-

Away: Voluntary Transaction Costs.” Unpublished.

32

Barzel, Yoram. 1989. Economic Analysis of Property Rights. Cambridge University Press.

Bellows, John, and Edward Miguel. 2009. “War and Local Collective Action in Sierra

Leone.” Journal of Public Economics, 93: 1144-1157.

Besley, Timothy, and Maitreesh Ghatak. 2010. “Property Rights and Economic Develop-

ment.” In Handbook of Development Economics, Vol. V. Mark Rosenzweig and Dani

Rodrik, eds. Amsterdam: Elsevier: 4525-4595.

Besley, Timothy, and Torsten Persson. 2009. “The Origins of State Capacity: Property

Rights, Taxation and Politics.” American Economic Review, 99: 1218-1244.

Bond, Eric W., and Kamal Saggi. 2017. “Compulsory Licensing and Patent Protection: a

North-South Perspective.” Forthcoming at the Economic Journal.

Bouckeart, Boudewijn, and Gerrit De Geest. 1995. “Private Takings, Private Taxes and

Private Compulsory Services: The Economic Doctrine of Quasi Contracts.” International

Review of Law and Economics, 15: 463-487.

Brueckner, Jan K., and Harris Selod. 2009. “A Theory of Urban Squatting and Land-Tenure

Formalization in Developing Countries.” American Economic Journal: Economic Policy,

1: 28-51.

Burk, Dan L., and Brett McDonnel. 2007. “The Goldilocks Hypothesis: Balancing Intellec-

tual Property Rights at the Boundary of the Firm.” University of Illinois Law Review,

2: 575-636.

Calabresi, Guido, and Douglas A. Melamed. 1972. “Property Rules, Liability Rules and

Inalienability: One View of the Cathedral.” Harvard Law Review, 85: 1089-1128.

Cervellati, Matteo, Piergiuseppe Fortunato, and Uwe Sunde. 2008. “Hobbes to Rousseau:

Inequality, Institutions, and Development.” Economic Journal, 118: 1354-1384.

Charnoz, Olivier, and Ashwini Swain. 2012. “In Pursuit of Energy Efficiency in India’s

Agriculture: Fighting “Free Power” or Working with it?” Unpublished.

33

Coase, Ronald H. 1960. “The Problem of Social Cost.” Journal of Law and Economics, 3:

1-44.

Cook, R. Dennis. 1977. “Detection of Influential Observations in Linear Regression.” Tech-

nometrics, 19: 15-18.

Croutzet, Alexandre, and Pierre Lasserre. 2017. “Optimal Completeness of Property Rights

on Renewable Resources in the Presence of Market Power.” Resource and Energy Eco-

nomics, 49: 16-32.

Dari-Mattiacci, Giuseppe. 2012. “Endogenous Transaction Costs.” Unpublished.

Dari-Mattiacci, Giuseppe, and Carmine Guerriero. 2015. “Law and Culture: A Theory of

Comparative Variation in Bona Fide Purchase Rules.” Oxford Journal of Legal Studies,

35: 543-574.

Dharmadhikari, Sudhakar, and Kumar Joag-Dev. 1988. Unimodality, Convexity, and Ap-

plications. Boston, MA: Academic Press.

Felli, Leonardo, and Antonio Merlo. 2006. “Endogenous Lobbying.” Journal of the European

Economic Association, 4: 180-215.

Foss, Nicolai J. 2011. “Human Capital and Transaction Cost Economics.” In the Oxford

Handbook of Human Capital, Alan BurtonJones and J.C. Spender, eds. Oxford: Oxford

University Press: 165-185.

Ganglmair, Bernhard. 2017. “Efficient Material Breach of Contract.” Journal of Law,

Economics, and Organization, 33: 507-540.

Gennaioli, Nicola. 2013. “Optimal Contracts with Enforcement Risk.” Journal of the Euro-

pean Economic Association, 11: 59-62.

Ghosh, Shubha. 2014. “The Implementation of Exhaustion Policies: Lessons from National

Experiences.” Unpublished.

34

Grossman, Sanford J., and Oliver D. Hart. 1986. “The Costs and Benefits of Ownership: A

Theory of Vertical and Lateral Integration.” Journal of Political Economy, 94: 691-719.

Guerriero, Carmine. 2016a. “Endogenous Property Rights.” Journal of Law and Economics,

59: 313-358.

Guerriero, Carmine. 2016b. “Endogenous Legal Traditions and Economic Outcomes.” Jour-

nal of Comparative Economics, 44: 416-433

Guerriero, Carmine. 2017. “Endogenous Property Rights and the Nature of the Firm.”

Unpublished.

Hart, Oliver, and John Moore. 1990. “Property Rights and the Nature of the Firm.” Journal

of Political Economy, 98: 1119-1158.

Hasen, Richard L., and Richard H. McAdams. 1997. “The Surprisingly Complex Case

Against Theft.” International Review of Law and Economics, 17: 367-378.

Johnson, Simon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer. 2000.

“Tunneling.” American Economic Review, 90: 22-27.

Jordan, Jim. 2006. “Pillage and Property.” Journal of Economic Theory, 131: 26-44.

North, Douglass C., John J. Wallis, Barry R. Weingast. 2009. Violence and Social Orders:

A Conceptual Framework for Interpreting Recorded Human History. Cambridge, UK:

Cambridge University Press.

Piazzesi, Monika, Martin Schneider, and Johannes Stroebel. 2017. “Segmented Housing

Search.” Unpublished.

Sackman, Julius, at al. 2013. Nichols on Eminent Domain. Albany, New York, NY: Matthew

Bender.

Segal, Ilya, and Michael Whinston. 2016. “Property Rights and the Efficiency of Bargain-

ing.” Journal of the European Economic Association, 14: 1287-1328.

35

Seitz, Michael, Alexander Tarasov, Roman Zakharenko. 2015. “Trade Costs, Conflicts, and

Defense Spending.” Journal of International Economics, 95: 305-318.

Smith, Adam. 2017. An Inquiry into the Nature and Causes of the Wealth of Nations.

Oxford: Oxford University Press.

Smith, Lionel D., 1997. The Law of Tracing. Oxford: Oxford University Press.

Williamson, Oliver E. 2010. “Transaction Cost Economics: The Natural Progression.”

American Economic Review, 100: 673-690.

World Economic Forum—WEF. 2015. The Global Competitiveness Report 2014-2015. Geneva:

WEF Press.

36

Tables and Figures

Table 1: Full Sample Albania; Algeria; Angola; Argentina; Armenia; Australia; Austria; Azerbaijan; Bahrain; Bangladesh; Belgium; Benin; Bhutan; Bolivia; Botswana; Brazil; Bulgaria; Burkina Faso; Burundi; Cambodia; Cameroon; Canada; Chad; Chile; China; Colombia; Costa Rica; Cote d’Ivoire; Croatia; Cyprus; Czech Republic; Denmark; Dominican Republic; Ecuador; Egypt; El Salvador; Estonia; Ethiopia; Finland; France; Gabon; Gambia; Germany; Ghana; Greece; Guatemala; Guinea; Honduras: Hungary; India; Indonesia; Iran; Ireland; Israel; Italy; Jamaica; Japan; Jordan; Kazakhstan; Kenya; Kuwait; Kyrgyz Republic; Lao; Latvia; Lebanon; Lesotho; Liberia; Lithuania; Luxembourg; Macedonia; Mada- gascar; Malawi; Malaysia; Mali; Mauritania; Mauritius; Mexico; Moldova; Mongolia; Montenegro; Morocco; Mozambique; Myanmar; Namibia; Nepal; Netherlands; New Zealand; Nicaragua; Nigeria; Norway; Oman; Pakistan; Panama; Paraguay; Peru; Philippines; Poland; Portugal; Qatar; Romania; Russia; Rwanda; Saudi Arabia; Senegal; Serbia; Sierra Leone; Singapore; Slovak Republic; Slovenia; South Africa; Spain; Sri Lanka; Suriname; Swaziland; Sweden; Switzerland; Syria; Taiwan; Tajikistan; Tanzania; Thailand; Trinidad & Tobago; Tunisia; Turkey; Uganda; Ukraine; United Arab Emirates; United Kingdom; United States; Uruguay; Venezuela; Vietnam; Yemen; Zambia; Zimbabwe.

Table 2: Summary of Variables Variable Definition and Sources Statistics

Property-Rights: Index ranging between one and seven and gaging the strength of generic property 4.388 rights. Source: 2006-2015 EOS, available at https://www.weforum.org/reports/ (1.034)

Property Intellectual-Property:

Index ranging between one and seven and gaging how strong are intellectual 3.713 rights: property rights. Source: 2006-2015 EOS. (1.124)

Shareholders-Protection: Index ranging between one and seven and gaging to what extent are the interests 4.305 of minority shareholders protected by the legal system. Source: 2006-2015 EOS. (0.750)

Over-Regulation: Index ranging between one and seven and gaging how burdensome is for firms to 3.708 comply with governmental administrative requirements. Source: 2006-2015 EOS. (0.658)

Unavailability-Financing: Index ranging between one and seven and gaging the financial sector difficulties 2.489 in providing products and services to businesses. Source: 2006-2015 EOS. (0.955)

Market-Dominance: Index ranging between one and seven and falling with the competitiveness of 3.168

Transaction corporate activity. Source: 2006-2015 EOS. (0.837) costs:

Trade-Barriers: Index ranging between one and seven and gaging how much non-tariff barriers curb 2.545 imported goods ability to compete with domestic ones. Source: 2006-2015 EOS. (0.685)

Asymmetric-Information: Index ranging between one and seven and falling with the extent of information used 3.447 by buyers to make purchasing decisions. Source: 2006-2015 EOS. (0.835)

Asset-Specificities: Index ranging between one and seven and gaging the competitive advantage of the 3.595 country’s companies in international markets. Source: 2006-2015 EOS. (1.015)

Income: Natural logarithm of the output-side real GDP at chained PPPs in 2011 US dollar 9.213 per capita. Source: Penn World Table 9.0, available at https://pwt.sas.upenn.edu/ (1.194)

Democracy: Polity IV constraints on the executive authority score ranging between one and 5.303 seven. Source: POLITY IV dataset, available at http://www.systemicpeace.org (1.875)

Reserves: Crude oil proved reserves in barrels per capita. Source: Energy Information 1895.327

Other Administration, available at http://www.eia.gov/ (9400.035) controls:

Conflict-External: Share of previous half-century in which the country was involved in external 0.014 military conflicts. Source: http://www.correlatesofwar.org/ (0.050)

Conflict-Internal: Share of previous half-century in which the country was involved in internal 0.064 military conflicts. Source: http://www.correlatesofwar.org/ (0.120)

Human-Capital: Ratio of total tertiary enrollment, regardless of age, to the population of the relative 33.758 age group. Source: 2006-2015 EOS. (25.843)

Technology-Availability: Index ranging between one and seven and gaging the availability of the latest 4.634

Excluded technologies. Source: 2006-2015 EOS. (1.033) instruments:

Math-Science: Index ranging between one and seven and gaging the quality of math and science 3.943 education. Source: 2006-2015 EOS. (0.975)

Note: 1. The last column reports the mean value and, in parentheses, the standard deviation of each variable. Both are computed for the 1350 observations used in tables 2 to 6.

37

Figure 1: Strength of Property Rights Protection

Note: 1. Here, I divide the range of each of the four variables, whose definitions and sources are listed in table 2, into four equal intervals.

38

Figure 2: Property Rights and Financial Inefficiencies

Note: 1. Residuals and fitted values lines are obtained from the sample employed in table 3.

Figure 3: Property Rights and Market Power

Note: 1. Residuals and fitted values lines are obtained from the sample employed in table 3.

Figure 4: Property Rights and Lemons-type Distortions

Note: 1. Residuals and fitted values lines are obtained from the sample employed in table 3.

39

Figure 5: Downstream Firm’s Property Rights and Incomplete Contracting Costs

Note: 1. Residuals and fitted values lines are obtained from the sample employed in table 3.

Figure 6: Property Rights and Transaction Costs

40

Table 3: Property Rights and Transaction Costs (1) (2) (3) (4) (5) (6)

Panel A. The dependent variable is Property-Rights

Over-Regulation - 0.445 (0.049)***

Unavailability-Financing - 0.190 (0.078)**

Market-Dominance - 0.370 (0.061)***

Trade-Barriers - 0.238 (0.047)***

Asymmetric-Information - 0.314 (0.044)***

Asset-Specificities 0.290 (0.059)***

Estimation OLS

R2 0.40 0.28 0.37 0.31 0.36 0.29 Number of observations 1350 1350 1350 1350 1350 1350

Panel B. The dependent variable is Intellectual-Property

Over-Regulation - 0.495 (0.063)***

Unavailability-Financing - 0.161 (0.080)**

Market-Dominance - 0.470 (0.065)***

Trade-Barriers - 0.181 (0.054)***

Asymmetric-Information - 0.425 (0.053)***

Asset-Specificities 0.492 (0.078)***

Estimation OLS

R2 0.24 0.06 0.26 0.07 0.25 0.18 Number of observations 1350 1350 1350 1350 1350 1350

Panel C. The dependent variable is Shareholders-Protection

Over-Regulation - 0.337 (0.060)***

Unavailability-Financing - 0.243 (0.102)**

Market-Dominance - 0.356 (0.067)***

Trade-Barriers - 0.329 (0.052)***

Asymmetric-Information - 0.317 (0.056)***

Asset-Specificities 0.367 (0.071)***

Estimation OLS

R2 0.25 0.23 0.28 0.28 0.28 0.24 Number of observations 1350 1350 1350 1350 1350 1350

Notes: 1. Robust standard errors allowing for clustering by country in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include country and year fixed effects.

Table 4: Property Rights and Transaction Costs — Controlling for Observables (1) (2) (3) (4) (5) (6)

Panel A. The dependent variable is Property-Rights

Over-Regulation - 0.442 (0.049)***

Unavailability-Financing - 0.202 (0.071)***

Market-Dominance - 0.357 (0.061)***

Trade-Barriers - 0.226 (0.049)***

Asymmetric-Information - 0.307 (0.048)***

Asset-Specificities 0.272 (0.058)***

P-value for all extra controls 0.08 0.07 0.33 0.31 0.56 0.27 Estimation OLS

R2 0.42 0.31 0.38 0.32 0.37 0.31 Number of observations 1350 1350 1350 1350 1350 1350

Panel B. The dependent variable is Intellectual-Property

Over-Regulation - 0.482 (0.063)***

Unavailability-Financing - 0.163 (0.077)**

Market-Dominance - 0.460 (0.065)***

Trade-Barriers - 0.174 (0.055)***

Asymmetric-Information - 0.414 (0.055)***

Asset-Specificities 0.469 (0.078)***

P-value for all extra controls 0.04 0.10 0.08 0.14 0.56 0.17 Estimation OLS

R2 0.26 0.10 0.28 0.11 0.26 0.20 Number of observations 1350 1350 1350 1350 1350 1350

Panel C. The dependent variable is Shareholders-Protection

Over-Regulation - 0.340 (0.059)***

Unavailability-Financing - 0.268 (0.108)**

Market-Dominance - 0.357 (0.063)***

Trade-Barriers - 0.340 (0.053)***

Asymmetric-Information - 0.329 (0.055)***

Asset-Specificities 0.352 (0.068)***

P-value for all extra controls 0.03 0.01 0.01 0.04 0.01 0.11 Estimation OLS

R2 0.27 0.25 0.30 0.30 0.30 0.25 Number of observations 1350 1350 1350 1350 1350 1350

Notes: 1. Robust standard errors allowing for clustering by country in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include country and year fixed effects, Income, Democracy, Reserves, Conflict-External, Conflict-Internal, and Human-Capital.

41

Table 5: Using Selection on Observables to Assess the Bias from Unobservables (1) (2) (3)

The dependent variable is Property-Rights Intellectual-Property Shareholders-Protection

The measure of transaction cost is:

Over-Regulation 147.33 37.08 113.33

Unavailability-Financing 16.83 81.50 10.72

Market-Dominance 27.46 46 357

Trade-Barriers 18.83 24.86 30.91

Asymmetric-Information 43.86 37.64 27.42

Asset-Specificities 15.11 20.39 23.47

Note: 1. Each cell reports an index constructed as explained in section 4.1.2 and based on the coefficients attached to the measure of transaction costs listed on the left and obtained from two regressions. In one, the covariates include only country and year fixed effects. In the other, the “full set” of covariates gathers country and year fixed effects, Income, Democracy, Reserves, Conflict-External, Conflict-Internal, and Human-Capital. The number of observations is 1350.

Table 6: Property Rights and Endogenous Transaction Costs (1) (2) (3) (4) (5) (6)

Panel A. The dependent variable is Property-Rights

Over-Regulation - 0.945 (0.166)***

Unavailability-Financing - 1.625 (0.396)***

Market-Dominance - 0.219 (0.047)***

Trade-Barriers - 1.185 (0.271)***

Asymmetric-Information - 0.776 (0.122)***

Asset-Specificities 2.043 (0.647)***

P-value for all extra controls 0.00 0.17 0.67 0.35 0.18 0.79 First Stage for the Transaction Costs Measure

Technology-Availability - 0.139 - 0.112 - 0.219 - 0.168 - 0.100 0.023 (0.051)*** (0.038)*** (0.047)*** (0.049)*** (0.055)* (0.034)

Math-Science - 0.294 - 0.126 - 0.379 - 0.137 - 0.411 0.162 (0.053)*** (0.057)** (0.058)*** (0.074)* (0.062)*** (0.051)***

P-value of underidentification test 0.00 0.00 0.00 0.00 0.00 0.00 P-value of overidentification test 0.49 0.15 0.24 0.07 0.65 0.46 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Panel B. The dependent variable is Intellectual-Property

Over-Regulation - 1.342 (0.206)***

Unavailability-Financing - 2.344 (0.613)***

Market-Dominance - 0.971 (0.103)***

Trade-Barriers - 1.725 (0.403)***

Asymmetric-Information - 1.088 (0.137)***

Asset-Specificities 2.846 (0.774)***

P-value for all extra controls 0.05 0.33 0.15 0.14 0.21 0.67 P-value of underidentification test 0.00 0.00 0.00 0.00 0.00 0.01 P-value of overidentification test 0.81 0.24 0.37 0.16 0.30 0.23 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Panel C. The dependent variable is Shareholders-Protection

Over-Regulation - 0.630 (0.163)***

Unavailability-Financing - 1.182 (0.296)***

Market-Dominance - 0.466 (0.108)***

Trade-Barriers - 0.902 (0.232)***

Asymmetric-Information - 0.481 (0.128)***

Asset-Specificities 1.217 (0.467)***

P-value for all extra controls 0.00 0.03 0.00 0.00 0.01 0.62 P-value of underidentification test 0.00 0.00 0.00 0.00 0.00 0.01 P-value of overidentification test 0.29 0.86 0.41 0.79 0.08 0.05 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Notes: 1. Robust standard errors allowing for clustering by country in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include country and year fixed effects, Income, Democracy, Reserves, Conflict-External, Conflict-Internal, and Human-Capital. The

endogenous variable in columns (1) to (6) is respectively Over-Regulation, Unavailability-Financing, Market-Dominance, Trade-Barriers, Asymmetric- Information, and Asset-Specificities, whereas the excluded instruments are Technology-Availability and Math-Science.

3. The null hypothesis of the Kleibergen-Paap test is that the excluded instruments are uncorrelated with the endogenous regressor. 4. The null hypothesis of the Hansen test of overidentifying restrictions is that the excluded instruments, as a group, are exogenous.

42

APPENDIX (NOT FOR PUBLICATION)

References

Dari-Mattiacci, Giuseppe, and Carmine Guerriero. 2017. “A Novel Dataset on Horizontal

Property Rights in 126 Jurisdictions.” Data in Brief, 11: 557-561.

Supplementary Tables

Table I: Summary of Variables Variable Definition and Sources Statistics

Years needed for adverse possession by any good-faith possessor of a movable good. 9.066 Adverse-Possession: Source: Dari-Mattiacci and Guerriero (2017). (10.570)

Property [53] rights: Index ranging between 1 and 10 and gaging the quality of both conflict of interest 5.101

Investors-Protection regulation and shareholder governance. Source: World Bank, Doing Business Project (1.575) 2006-2015, available at http://www.doingbusiness.org/data [1350] Index ranging between one and seven and rising with the extent to which making 2.902

Bribery: bribes in connection with their daily activities is common for firms. Source: 2006- (1.198) 2015 EOS. [1350] Index ranging between one and seven and rising with the difficulty to obtain a bank 3.995

Unavailability-Loans: loan with only a good business plan and no collateral. Source: 2006-2015 EOS. (0.874) [1350]

Index ranging between one and seven and rising with how difficult is it for companies 3.254 Transaction Unavailability-Equity: to raise money by issuing shares on the stock market. Source: 2006-2015 EOS. (1.105) costs: [1350]

Index ranging between one and seven and falling with the intensity of competition 2.165 Market-Power : in the local markets. Source: 2006-2015 EOS. (0.667)

[1350] Index ranging between one and seven and gaging the extent of production 3.792

Production-Sophistication: sophistication. A value of one (seven) suggests that production uses labor-intensive (1.064) (sophisticated and knowledge-intensive) processes. Source: 2006-2015 EOS. [1350] See text. Source: United Nations, Surveys of Crime Trends and the Operations of - 0.007

Enforcement: Criminal Justice Systems 2005-2015, available at https://www.unodc.org/ (1.012) Other [740] controls: See text. Sources: European Value Study and World Value Survey available at - 0.0004

Culture: http://www.europeanvaluesstudy.eu/ and http://www.worldvaluessurvey.org/ (1.030) [590]

Note: 1. The last column reports the mean value and, in parentheses, the standard deviation of each variable. Both are computed for the 1350 observations used in tables III to VI, except in the case of Adverse-Possession for which they are calculated for the 91 observations employed in table II.

43

Table II: Property Rights and Endogenous Transaction Costs — Cross-Section (1) (2) (3) (4) (5) (6)

Panel A. The dependent variable is Adverse-Possession

Over-Regulation - 14.204 (4.986)***

Unavailability-Financing - 6.370 (3.149)**

Market-Dominance - 11.314 (4.060)***

Trade-Barriers - 11.827 (6.805)*

Asymmetric-Information - 15.717 (6.552)**

Asset-Specificities 8.742 (2.988)***

P-value of underidentification test 0.20 0.00 0.00 0.01 0.00 0.00 P-value of overidentification test 0.82 0.19 0.81 0.21 0.82 0.97 P-value of endogeneity test 0.08 0.02 0.11 0.15 0.06 0.06 Estimation 2SLS Number of observations 53 53 53 53 53 53

Panel B. The dependent variable is Property-Rights

Over-Regulation - 1.880 (0.639)***

Unavailability-Financing - 1.135 (0.252)***

Market-Dominance - 1.514 (0.411)***

Trade-Barriers - 2.088 (0.722)***

Asymmetric-Information - 2.108 (0.705)***

Asset-Specificities 1.126 (0.277)***

P-value of underidentification test 0.20 0.00 0.00 0.01 0.00 0.00 P-value of overidentification test 0.17 0.32 0.18 0.40 0.14 0.09 P-value of endogeneity test 0.01 0.01 0.01 0.02 0.03 0.04 Estimation 2SLS Number of observations 53 53 53 53 53 53

Panel C. The dependent variable is Intellectual-Property

Over-Regulation - 2.297 (0.878)***

Unavailability-Financing - 1.232 (0.265)***

Market-Dominance - 1.841 (0.382)***

Trade-Barriers - 2.273 (0.827)***

Asymmetric-Information - 2.561 (0.717)***

Asset-Specificities 1.392 (0.288)***

P-value of underidentification test 0.20 0.00 0.00 0.01 0.00 0.00 P-value of overidentification test 0.57 0.07 0.45 0.18 0.56 0.20 P-value of endogeneity test 0.00 0.13 0.00 0.09 0.00 0.01 Estimation 2SLS Number of observations 53 53 53 53 53 53

Panel D. The dependent variable is Shareholders-Protection

Over-Regulation - 1.465 (0.682)**

Unavailability-Financing - 0.959 (0.141)***

Market-Dominance - 1.184 (0.267)***

Trade-Barriers - 1.761 (0.543)***

Asymmetric-Information - 1.650 (0.411)***

Asset-Specificities 0.870 (0.273)***

P-value of underidentification test 0.20 0.00 0.00 0.01 0.00 0.00 P-value of overidentification test 0.11 0.45 0.08 0.65 0.03 0.04 P-value of endogeneity test 0.01 0.01 0.03 0.00 0.06 0.05 Estimation 2SLS Number of observations 53 53 53 53 53 53

Notes: 1. Robust standard errors in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include Income, Democracy, Reserves, Conflict-External, Conflict-Internal, Human-Capital, Enforcement, and Culture. The en-

dogenous variable in columns (1) to (6) is respectively Over-Regulation, Unavailability-Financing, Market-Dominance, Trade-Barriers, Asymmetric- Information, and Asset-Specificities, whereas the excluded instruments are Technology-Availability and Math-Science.

3. The null hypothesis of the Kleibergen-Paap test is that the excluded instruments are uncorrelated with the endogenous regressor. 4. The null hypothesis of the Hansen test of overidentifying restrictions is that the excluded instruments, as a group, are exogenous. 5. The null hypothesis of the endogeneity test is that the endogenous regressor can be treated as exogenous.

Table III: Investors Protection and Endogenous Transaction Costs (1) (2) (3) (4) (5) (6)

The dependent variable is Investors-Protection

Over-Regulation - 0.241 (0.150)*

Unavailability-Financing - 0.469 (0.287)*

Market-Dominance - 0.180 (0.110)*

Trade-Barriers - 0.364 (0.213)*

Asymmetric-Information - 0.178 (0.127)

Asset-Specificities 0.442 (0.344)

P-value of underidentification test 0.00 0.00 0.00 0.00 0.00 0.00 P-value of overidentification test 0.49 0.75 0.56 0.90 0.33 0.28 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Notes: 1. Robust standard errors in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include country and year fixed effects, Income, Democracy, Reserves, Conflict-External, Conflict-Internal, and Human-Capital. The

endogenous variable in columns (1) to (6) is respectively Over-Regulation, Unavailability-Financing, Market-Dominance, Trade-Barriers, Asymmetric- Information, and Asset-Specificities, whereas the excluded instruments are Technology-Availability and Math-Science.

3. The null hypothesis of the Kleibergen-Paap test is that the excluded instruments are uncorrelated with the endogenous regressor. 4. The null hypothesis of the Hansen test of overidentifying restrictions is that the excluded instruments, as a group, are exogenous.

44

Table IV: Property Rights and Other Measures of Endogenous Transaction Costs (1) (2) (3) (4) (5)

Panel A. The dependent variable is Adverse-Possession

Bribery - 1.427 (0.321)***

Unavailability-Loans - 0.763 (0.141)***

Unavailability-Equity - 0.617 (0.175)***

Market-Power - 0.622 (0.144)***

Production-Sophistication 0.817 (0.127)***

P-value of underidentification test 0.00 0.00 0.00 0.00 0.00 P-value of overidentification test 0.60 0.07 0.00 0.00 0.53 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350

Panel B. The dependent variable is Intellectual-Property

Bribery - 2.027 (0.454)***

Unavailability-Loans - 1.051 (0.174)***

Unavailability-Equity - 0.802 (0.202)***

Market-Power - 0.937 (0.173)***

Production-Sophistication 1.159 (0.123)***

P-value of underidentification test 0.00 0.00 0.00 0.00 0.00 P-value of overidentification test 0.82 0.02 0.00 0.00 0.81 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350

Panel C. The dependent variable is Shareholders-Protection

Bribery - 0.953 (0.258)***

Unavailability-Loans - 0.422 (0.135)***

Unavailability-Equity - 0.216 (0.164)

Market-Power - 0.557 (0.154)***

Production-Sophistication 0.543 (0.139)***

P-value of underidentification test 0.00 0.00 0.00 0.00 0.00 P-value of overidentification test 0.28 0.01 0.00 0.08 0.28 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350

Notes: 1. Robust standard errors allowing for clustering by country in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include country and year fixed effects, Income, Democracy, Reserves, Conflict-External, Conflict-Internal, and Human-Capital.

The endogenous variables in columns (1) to (5) is respectively Bribery, Unavailability-Loans, Unavailability-Equity, Market-Power, and Production- Sophistication, whereas the excluded instruments are Technology-Availability and Math-Science.

3. The null hypothesis of the Kleibergen-Paap test is that the excluded instruments are uncorrelated with the endogenous regressor. 4. The null hypothesis of the Hansen test of overidentifying restrictions is that the excluded instruments, as a group, are exogenous.

Table V: Controlling for the Intermediation Technology (1) (2) (3) (4) (5) (6)

Panel A. The dependent variable is Property-Rights

Over-Regulation - 0.906 (0.376)**

Unavailability-Financing - 1.591 (0.454)***

Market-Dominance - 1.013 (0.293)***

Trade-Barriers - 1.340 (0.610)**

Asymmetric-Information - 0.753 (0.307)**

Asset-Specificities - 8.719 (16.529)

P-value of underidentification test 0.06 0.07 0.04 0.06 0.10 0.86 P-value of overidentification test 0.11 0.29 0.02 0.83 0.06 0.81 Estimation 2SLS Number of observations 310 310 310 310 310 310

Panel B. The dependent variable is Intellectual-Property

Over-Regulation - 1.452 (0.401)***

Unavailability-Financing - 2.150 (0.600)***

Market-Dominance - 1.665 (0.417)***

Trade-Barriers - 1.891 (0.780)**

Asymmetric-Information - 1.291 (0.423)***

Asset-Specificities - 14.025 (27.297)

P-value of underidentification test 0.06 0.07 0.04 0.06 0.10 0.86 P-value of overidentification test 0.26 0.12 0.07 0.68 0.06 0.88 Estimation 2SLS Number of observations 310 310 310 310 310 310

Panel C. The dependent variable is Shareholders-Protection

Over-Regulation - 0.620 (0.393)

Unavailability-Financing - 1.681 (0.487)***

Market-Dominance - 0.635 (0.438)

Trade-Barriers - 1.298 (0.479)***

Asymmetric-Information - 0.391 (0.421)

Asset-Specificities - 5.901 (11.821)

P-value of underidentification test 0.06 0.07 0.04 0.06 0.10 0.86 P-value of overidentification test 0.07 0.93 0.03 0.65 0.04 0.60 Estimation 2SLS Number of observations 310 310 310 310 310 310

Notes: 1. Robust standard errors allowing for clustering by country in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include country and year fixed effects, Income, Democracy, Reserves, Conflict-External, Conflict-Internal, Human-Capital, Enforce-

ment, and Culture. The endogenous variable in columns (1) to (6) is respectively Over-Regulation, Unavailability-Financing, Market-Dominance, Trade-Barriers, Asymmetric-Information, and Asset-Specificities, whereas the excluded instruments are Technology-Availability and Math-Science.

3. The null hypothesis of the Kleibergen-Paap test is that the excluded instruments are uncorrelated with the endogenous regressor. 4. The null hypothesis of the Hansen test of overidentifying restrictions is that the excluded instruments, as a group, are exogenous.

45

Table VI: Property Rights and Endogenous Transaction Costs — Semi-reduced Forms (1) (2) (3) (4) (5) (6)

Panel A. The dependent variable is Property-Rights

Over-Regulation - 1.031 (0.200)***

Unavailability-Financing - 2.403 (1.008)**

Market-Dominance - 0.800 (0.125)***

Trade-Barriers - 2.212 (1.127)**

Asymmetric-Information - 0.738 (0.140)***

Asset-Specificities 1.870 (0.601)***

Technology-Availability - 0.041 - 0.168 - 0.073 - 0.269 0.029 0.059 (0.062) (0.156) (0.064) (0.254) (0.062) (0.075)

P-value of underidentification test 0.00 0.03 0.00 0.07 0.00 0.00 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Panel B. The dependent variable is Intellectual-Property

Over-Regulation - 1.383 (0.253)***

Unavailability-Financing - 3.224 (1.433)**

Market-Dominance - 1.073 (0.161)***

Trade-Barriers - 2.967 (1.669)*

Asymmetric-Information - 0.990 (0.142)***

Asset-Specificities 2.508 (0.707)***

Technology-Availability - 0.020 - 0.190 - 0.062 - 0.326 0.074 0.114 (0.085) (0.210) (0.073) (0.367) (0.067) (0.084)

P-value of underidentification test 0.00 0.03 0.00 0.07 0.00 0.00 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Panel C. The dependent variable is Shareholders-Protection

Over-Regulation - 0.474 (0.219)**

Unavailability-Financing - 1.104 (0.570)**

Market-Dominance - 0.368 (0.160)**

Trade-Barriers - 1.017 (0.520)**

Asymmetric-Information - 0.339 (0.152)**

Asset-Specificities 0.859 (0.417)**

Technology-Availability 0.075 0.017 0.060 - 0.030 0.107 0.121 (0.069) (0.096) (0.072) (0.122) (0.059)* (0.056)**

P-value of underidentification test 0.00 0.03 0.00 0.07 0.00 0.00 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Panel D. The dependent variable is Property-Rights

Over-Regulation - 0.735 (0.331)**

Unavailability-Financing - 0.909 (0.430)**

Market-Dominance - 0.467 (0.203)**

Trade-Barriers - 0.609 (0.301)**

Asymmetric-Information - 1.024 (0.660)

Asset-Specificities 4.388 (6.399)

Technology-Availability 0.087 0.189 0.126 0.220 - 0.117 - 0.409 (0.115) (0.092)** (0.100) (0.081)*** (0.299) (1.067)

P-value of underidentification test 0.01 0.01 0.00 0.00 0.07 0.49 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Panel E. The dependent variable is Intellectual-Property

Over-Regulation - 1.241 (0.470)***

Unavailability-Financing - 1.535 (0.570)***

Market-Dominance - 0.788 (0.213)***

Trade-Barriers - 1.028 (0.399)***

Asymmetric-Information - 1.729 (0.930)*

Asset-Specificities 7.409 (10.093)

Technology-Availability 0.042 0.213 0.108 0.266 - 0.303 - 0.795 (0.166) (0.127)* (0.110) (0.127)** (0.414) (1.707)

P-value of underidentification test 0.01 0.01 0.00 0.00 0.07 0.49 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Panel F. The dependent variable is Shareholders-Protection

Over-Regulation - 1.012 (0.453)**

Unavailability-Financing - 1.253 (0.470)***

Market-Dominance - 0.643 (0.249)***

Trade-Barriers - 0.838 (0.338)**

Asymmetric-Information - 1.410 (0.813)*

Asset-Specificities 6.045 (8.391)

Technology-Availability - 0.158 - 0.019 - 0.105 0.024 - 0.440 - 0.841 (0.169) (0.112) (0.129) (0.091) (0.365) (1.395)

P-value of underidentification test 0.01 0.01 0.00 0.00 0.07 0.49 Estimation 2SLS Number of observations 1350 1350 1350 1350 1350 1350

Notes: 1. Robust standard errors allowing for clustering by country in parentheses. *** significant at the 1% confidence level; **, 5%; *, 10%. 2. All specifications include country and year fixed effects, Income, Democracy, Reserves, Conflict-External, Conflict-Internal, and Human-Capital. The

endogenous variables in columns (1) to (6) is respectively Over-Regulation, Unavailability-Financing, Market-Dominance, Trade-Barriers, Asymmetric- Information, and Asset-Specificities, whereas the excluded instruments are Math-Science in panels A to C and Technology-Availability otherwise.

3. The null hypothesis of the Kleibergen-Paap test is that the excluded instruments are uncorrelated with the endogenous regressor.

46