Trust and delegation: Theory and evidence

Accepted Manuscript

Trust and Delegation: Theory and Evidence Nurullah Gur , Christian Bjørnskov PII: DOI: Reference:

S0147-5967(16)00018-4 10.1016/j.jce.2016.02.002 YJCEC 2514

To appear in:

Journal of Comparative Economics

Received date: Revised date: Accepted date:

6 October 2015 14 February 2016 14 February 2016

Please cite this article as: Nurullah Gur , Christian Bjørnskov , Trust and Delegation: Theory and Evidence, Journal of Comparative Economics (2016), doi: 10.1016/j.jce.2016.02.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

   

ACCEPTED MANUSCRIPT This paper investigates the effect of trust on delegation. We first develop a simple model that yields directly testable implications. Our OLS and 2SLS results show that trust has positive effect on delegation. The effect of trust is increasing in economic sophistication.

AC

CE

PT

ED

M

AN US

CR IP T

Highlights

1

ACCEPTED Trust and Delegation: MANUSCRIPT Theory and Evidence1

Nurullah Gur2 School of Business and Management Sciences, Istanbul Medipol University, Kavacık Mahallesi Ekinciler Caddesi No: 19 34810 Beykoz-Istanbul Turkey. e-mail: [email protected] Tel: +902166815100 Fax: +902125212377

CR IP T

Christian Bjørnskov Department of Economics and Business, Aarhus University, Fuglesangs Allé 5, DK-8200 Aarhus V, Denmark; e-mail: [email protected] Tel: +4587164819

Abstract: Social trust is associated with good economic performance, but little is known about the transmission

AN US

mechanisms connecting trust and performance. We explore the effect of trust on delegation decisions. In a theoretical framework, we note that delegation is a low-cost option when management decisions can be implemented without monitoring. This option is, however, risky and more likely to be profitable in higher-trust environments. In a set of cross-country regressions, we show a strong association between trust and delegation,

ED

M

which is increasing in economic sophistication.

Keywords: Social trust, economic development, delegation

PT

JEL Codes: L22, O40, Z13, 1. Introduction

CE

Productivity is one of the key drivers of economic growth. As Bloom and Van Reenen (2010a) highlight, poor

AC

management practices reduce productivity. Therefore, if one considers productivity increases desirable, it seems well motivated to try to identify ways to reform such practices. Decentralized decision-making has been seen as an important way to improve management in firms, and thus increase firms’ performance because (a) it reduces the cost of information transfers and communication, (b) it increases firms’ speed of response to market changes, and (c) it increases job satisfaction (Bloom and Van Reenen, 2010b). 1

We would first of all like to thank the editor (Daniel Berkowitz) and the reviewer for the kind and very helpful comments on our paper. We thank Niclas Berggren for insightful comments on an earlier version. We also thank participants at 14 th EBES Conference (2014, Barcelona) and the Istanbul Technical University Department of Economics seminar series for their feedback. All remaining errors are of course entirely ours. 2 Corresponding author.

2

ACCEPTED MANUSCRIPT Yet, although delegation might provide productivity, flexibility and job satisfaction, it will also pose risks such as principal-agent problems. When the decision-making authority is delegated to an agent, there is no guarantee that decisions are made in line with the interests of the principal. In other words, the CEO might act in his/her own interests rather than those of the shareholders or the subordinates might act in their own interests rather than those of the CEO and shareholders.3 In response to the principal-agent problem, transaction cost economics (TCE) suggests certain control mechanisms for successful delegation. Contracting institutions

CR IP T

represent potential remedies, but are often insufficient because of contractual incompleteness: It is impossible to package all possible issues and specify all future contingencies in advance, which provides room for problems associated with opportunist behavior and bounded rationality (Williamson, 1971; 1975; 1981). As Cingano and Pinotti (2012) underline, while contracting institutions may only be partly effective cures to

AN US

problems and promote cooperation in market transactions between firms and its suppliers, the delegation of decision-making authority mostly remains outside the shadow of the law. Due to contractual incompleteness, more recent studies extend the TCE framework by featuring trust as an alternative governance device (Bromiley and Cummings, 1995; Nooteboom, 2000; Bromiley and Harris, 2006; Aghion et al., 2013). We

M

follow this line of research by investigating the potential role of social trust in delegation of authority.

ED

Following in the footsteps of the pioneering work of Banfield (1958), Putnam (1993) and Fukuyama (1995), a growing empirical literature shows that trust is positively related to long-run growth and development

PT

(Knack and Keefer, 1997; Zak and Knack, 2001, Algan and Cahuc, 2010, Horváth, 2013). Within this literature, a smaller number of more recent papers focus on the mechanisms through which trust affects growth and

CE

development. Some of the identified mechanisms are financial development (Guiso et al., 2008a, 2008b),

AC

human capital (Bjørnskov, 2009; Papagapitos and Riley, 2009; Dearmon and Grier, 2011; Bjørnskov and Méon, 2013), public expenditures (Ponzetto and Troiano, 2012; Bergh and Bjørnskov, 2011), lighter regulatory burdens (Aghion et al. 2010; Pinotti, 2012) and inclusive economic and judicial institutions (La Porta et al., 1997; Tabellini, 2008; Bjørnskov, 2010).

3

In family firms, due to principal-agent problem, most of the firm owners do not employ professional CEOs. They may also tend to

be unwilling to give any responsibility to their managers, preferring to control all activities from recruitment to marketing.

3

ACCEPTED When focusing, as we do in this paper, on whetherMANUSCRIPT social trust affects delegation, a first question to ask is how trust might theoretically affect delegation decisions? In general, the literature shows that trust facilitates cooperation among anonymous persons and reduces the necessity of monitoring (Fukuyama, 1995; La Porta et al. 1997; Bjørnskov, 2009). As is well-known, principle-agent problems are harmful for corporate governance. High trust might be a remedy for these problems, and thus improve corporate governance, as monitoring becomes relatively cheaper (e.g. Zak and Knack, 2001). If the level of trust is high, the shift of decision-making authority from superiors to subordinates might be also more likely. Fukuyama (1995: 31) summarizes this issue

CR IP T

as follows: “A high-trust society can organize its workplace on a more flexible and group-oriented basis, with more responsibility delegated to lower levels of the organization. Low-trust societies, by contrast, must fence in and isolate their workers with a series of bureaucratic rules.” We therefore hypothesize that in high-trust

AN US

societies, CEOs might give more responsibility to their managers and managers might be more open to give responsibility to their subordinates. On the other hand, delegation may arguably be lower in low-trust societies. The effect of trust on delegation may also depend on the level of economic sophistication. As the economy develops, its increasing scale usually outstrips the capabilities of centralized management because

M

firms will start to be late in responding to changing market conditions, and information processing becomes

ED

increasingly costly. As economic structure thus becomes complicated, the cost of monitoring employees doing increasingly complex jobs in particular gradually increases, making direct centralized control steadily less

PT

effective. Therefore, the importance of cooperation and coordination, but also the likelihood of principal-agent problems, increases with the level of economic sophistication that makes skilled tasks less monitorable. On this

CE

basis, we hypothesize that the effect of trust on delegation decisions is increasing in the level of economic

AC

(technological) sophistication, as trust potentially limits the cost increase. Our paper is related to the newly emerging literature on trust and delegation. In particular, two recent studies deal with micro-level evidence of trust effects on delegation, which we briefly outline before turning to a macro-level theory of how trust might affect the use of delegation in society as a whole. Bloom et al. (2012) focus on the effect of trust, measured either as the level of trust in region where central headquarter of the plant is located or as bilateral trust levels between a multinational’s country of origin and its subsidiaries’ locations, on decentralized decision making process within firms. In their model, a CEO has two options to solve 4

ACCEPTED production problems. He can address these problems MANUSCRIPT directly or delegate decisions to a plant manager, but choosing a delegation process requires trust. The CEO might not trust the plant manager’s decisions because of misaligned incentives such as those arising from corruption. When the CEO does not trust plant managers, there will be less decentralization and more direct interventions from the CEO and she will become time constrained due to a large number of decisions. The theoretical model in Bloom et al. (2012) shows that firm size is thereby reduced when the CEO’s trust in the plant manager is low because he spends more time on solving problems

CR IP T

instead of delegating decisions. Trust helps firms to become more decentralized, and decentralization in turn might improve productivity by supporting a large equilibrium firm size. Collecting data on the decentralization of investment, hiring, production, and sales decisions from corporate headquarters to local plant managers in almost 4000 firms in the United States, Europe, and Asia, Bloom et al. (2012) test their model’s propositions.

AN US

Their results show that trust is positively associated with decentralization and firm size.

In another recent paper, Cingano and Pinotti (2012) test these theoretical implications by following Rajan and Zingales’s (1998) difference-in-difference methodology to estimate the effect of trust on the structure of production. Cingano and Pinotti (2012) argue that if trust solves principle-agent problems, and thus increases

M

decentralization, it will affect high-delegation industries (such as manufacture of machinery and equipment)

ED

more than low-delegation industries (such as leather, leather products and footwear). Using micro level data from both Italian regions and European countries, they test whether trust increases value added, export and firm

PT

size relatively more in delegation-intensive industries. Their results show that trust increases value added,

industries.

CE

exports and firm size more in high delegation-intensive industries relative to low delegation-intensive

AC

As a contribution to this literature, the aim of this paper to test whether social trust affects delegation by using a cross-country approach. In other words, we ask whether the recent findings generalize to a national setting. To do so, we first develop a simple model that yields directly testable implications. We then test the predictions of this simple model by using a macro level dataset. In order to provide a causal relation, we run two-stage least-squares (2SLS) regressions with predetermined instruments to assess the impact of exogenous variations in trust. To the best of our knowledge, this is the first paper that investigates this topic at the macro level. 5

ACCEPTED The rest of the paper is organized as follows:MANUSCRIPT Section 2 outlines a simple model illustrating the relationship between trust and delegation to frame the empirical analysis. The data used in this paper and the estimation strategy are presented in section 3. Section 4 reports our findings. Finally, section 5 concludes.

2. Theoretical Model To model the choice of delegation, we rely on these insights and start with a standard principal-agent

CR IP T

framework. A principal, for example a CEO, has the choice between either hiring unskilled labour or hiring skilled labour that is more productive but also difficult to monitor. If the principal hires skilled labour, he must decide between delegating responsibilities directly to a sub-unit, represented by a hired agent, or not delegating directly but employing a supervisor to monitor the sub-unit. In the latter case, substantially less responsibility is

AN US

delegated as an extra level of management is created. Figure 1 illustrates the trade-off faced by the principal. This game provides the background for an optimization decision to be taken by the management: how many unskilled and skilled workers to employ, and how much to delegate directly to employees instead of highpaid supervisors. Unskilled workers earn wu, skilled workers earn ws and supervisors earn wm. The productivity

M

loss if skilled workers are not supervised and decide to shirk is a-b, and the a priori risk of shirking is 1-µ.

ED

Shirking brings skilled labor a benefit e in case it is the shirking type. Figure 1 approximately here

PT

As such, we introduce social trust as one of two elements of the model. First, one might imagine that lowtrust shirkers are more likely to benefit from shirking as they feel less shame by behaving dishonestly and their

CE

benefit e thus does not include a cognitive cost (e.g. Banfield, 1975; Bjørnskov, 2011). Second, a standard way

AC

of modelling trust is to assume that low-trust individuals, regardless whether they are skilled / educated or not, are more likely to be potential shirkers such that the risk 1-µ negatively reflects the overall trust level of society when firms cannot ex ante monitor the type of employees. The firm optimizes a standard production function, including a management cost associated with supervisors monitoring employees. Wages are taken as given by each firm and the interest rate r is set exogenously by a central bank. Profits, π, are maximized by setting employment lu and ls and the level of delegation (supervisors), ζ. This yields the management’s decision problem: 6

Max π = θ [nuβ ns1-β ]α k1-α - lu wu - ACCEPTED ls ws – ζ wm – r MANUSCRIPT k

(1)

Where nu = lu

(2)

ns = (ζ a + (1- ζ) (µ a + (1 - µ) b)) ls

(3)

As such, production takes place with a technology that varies in its level of sophistication. A higher β thus implies that the technology makes unskilled labor relatively more productive (on the margin) while a lower β

CR IP T

signifies technological sophistication that requires skilled labor. This set-up yields the following first-order conditions, where f is the production function in 1): d π / d k = 0  (1-α) f = r k

(4)

d π / d lu = 0  α β f = wu lu

(5)

d π / d ls = 0  α (1-β) f = ws ls

AN US

(6)

d π / d ζ = 0  α (1- β) (1 - µ) (a - b) f = wm (ζ a + (1- ζ) (µ a + (1 - µ) b)) (7) At the firm level, at which wages are taking as given, these four conditions yield an optimum skilled labor force of: 𝜇𝑎+(1−𝜇)𝑏

M

𝑙𝑠 = 𝑤𝑚 (1−𝜇)(𝑎−𝑏)

(8)

if the shirking risk µ is low:

𝑑𝑙𝑢 𝑑𝜇

𝑤

𝑎

𝑚 = 𝑤 (1−𝜇) [2 − (1−𝜇)(𝑎−𝑏)] > 0 𝑠

𝛽

𝑤 𝑑𝑙𝑠

= 1−𝛽 𝑤 𝑠

𝑑𝜇

(9)

PT

𝑑𝜇

>0

(10)

CE

𝑑𝑙𝑠

ED

As such, it is easy to observe that skilled and unskilled labor is used more intensively if trust is high, i.e.

𝑢

AC

Since labor is a limited resource, the increased demand for ‘regular’ labor therefore implies a reduced demand for management. The positive trust effect on the demand for regular skilled and unskilled labor thus implies a lower demand for management, ζ, and thus expanded use of delegation. Even without providing a general equilibrium (with a mathematically ugly analytical solution) noting that management must intuitively be hired from the same educated pool as skilled labor provides us with additional insights. First, with less delegation, i.e. a higher management share ζ, a smaller share of the population will work in skilled jobs, ls. Second, for any given wu, the returns to management are decreasing in trust. Combined, 7

this must necessarily decrease wu andACCEPTED – given that MANUSCRIPT wm > ws such that there is an incentive to work in management, increase wage inequality. As such, the modelling framework generates a particular result wellknown from the trust literature that trust and income inequality are negatively correlated (Uslaner, 2002; Bergh and Bjørnskov, 2014). Overall, the model leads us to two directly testable hypotheses: H1: Delegation is increasing in trust Proof: Note that (9) and (10) imply that ζ must logically be decreasing in trust.

CR IP T

H2: The effect of trust on delegation decisions is increasing in the level of technological (economic) sophistication.

Proof: Note that β is a measure of technological simplicity and that the derivative of (10) is decreasing in β. The second derivate of ζ must thus be increasing in β, i.e. when simple production technology favors

β, and thus decreasing in economic sophistication.

M

3. Data

AN US

unskilled tasks. This implies that the marginal effect of trust on the demand for unskilled labor is increasing in

ED

Our measure of delegation is from the Executive Opinion Survey of the Global Competitiveness Report (see Schwab and Sala-i Martin, 2013) and is therefore a large-scale survey-based index. The main aim of this survey

PT

is to gather useful information on a broad set of factors affecting a country’s competitiveness and sustainable development. 13,638 firms were surveyed from 144 countries. While the average sample size per country is

CE

94.7, the median is 86, and firm level answers were aggregated at the country level based on the assumption

AC

that the survey composition is approximately representative for each country. In order to take into account the structure of a country’s economy, sector weights were nevertheless used when averages were computed.4 Our delegation index is constructed from the answers of CEOs and other members of companies’ senior management to the following question: “In your country would you assess the willingness to delegate the authority to subordinates?” and ranges from 1 to 7. Higher values indicate high willingness to delegate among

4

See chapter 1.3 of The Global Competitiveness Report for more details on survey and sampling (Schwab and Sala-i-Martin, 2013).

8

ACCEPTED managers with the data in the present sample distributedMANUSCRIPT from 2 (Burkina Faso) to 6 (Denmark) and an average of 3.8. Following the previous literature, trust is measured through to the standard question “In general, do you think most people can be trusted?” To maximize the number of observations, we follow Bjørnskov and Méon (2013) in compiling several sources: the five waves of the World Values Survey between 1981 and 2005 (Inglehart et al. 2004), data from the 1995 and 2003 LatinoBarometro, the 2001–2004 Asian and East Asian

CR IP T

Barometers, the 2001–2007 AfroBarometer, and the 2002–2004 Danish Social Capital Project. Despite the apparent vagueness of the trust question, a series of studies since Knack and Keefer (1997) has documented its validity (e.g. Bjørnskov, 2011; Sapienza et al., 2013).

A main challenge is the absence of a standard set of determinants of delegation decisions. We have

AN US

therefore selected a set of different control variables from different sources, based on two premises. First, we include variables that are known to be associated with social trust and thus could confound the relation between trust and delegation. Second, we include other variables relating to our theoretical considerations in section 2. We first include a post-communist dummy variable (Post-Communist), as the formerly communist

M

countries are known to have lower trust levels but also a history of substantial centralization; this variable is

ED

from Bjørnskov and Méon (2013). The other major difference in most trust studies is the Nordic countries, which are the most trusting in the world and known to have a strong culture of delegation of decisions and flat

PT

corporate hierarchies. We therefore include a dummy variable for these countries (Nordic). As a first institutional variable, we use a democracy index (Democracy) from the Polity IV dataset, which

CE

is widely used in the literature (Marshall and Jaggers, 2010). The Democracy index reflects the regulation and

AC

competitiveness of political participation, competitiveness and openness of political executive recruitment, and constitutional constraints on the chief executive. The original democracy index ranges from -10 (full autocracy) to 10 (full democracy), which we include as more institutionally developed countries may be able to apply more sophisticated technology (Aghion et al., 2007). Following our theoretical considerations, we alternatively use the share of the population aged 25 and over that has completed secondary education (SecEdu) as a proxy for education. This variable is from Barro and Lee (2010), and comes to proxy for the skilled share of the population in the model framework, ls + ζ. 9

ACCEPTED MANUSCRIPT In order to ensure that such decisions are freely made, we include an index of freedom in life (FreedomLife) from the Gallup World Poll (2013). It is the percentage of respondents declaring themselves satisfied when asked the question, “Are you satisfied or dissatisfied with your freedom to choose what you do with your life?” As a second institutional measure, we also include a standard index of business regulatory freedom obtained from the Fraser Institute’s Economic Freedom of the World reports (Gwartney et al., 2013). This index is distributed between 0 and 10 where higher ratings indicate greater freedom from business

CR IP T

regulations in credit, product and labor markets, and is included for two separate reasons. First, Aghion et al. (2010) show that the index is related to social trust. Second, Young and Lawson (2014) demonstrate that heavier regulations in particular are associated with lower labor shares of national income (in the model, this would be the relative wage share pertaining to lu and ls). Similarly, we also include a measure of the overall

AN US

quality of institutions, measured by the index of rule of law from the Worldwide Governance Indicators constructed by Kaufmann et al. (2009). We additionally use an index of state antiquity that shows the presence of state government, spanning 39 half centuries from 1 AD to 1950 AD, which is taken from Bockstette et al. (2002). This index was constructed by using the following questions: (1) Is there government above the tribal

M

level?; (2) Is government foreign or locally based?; and (3) How much of the territory of the modern country

ED

was ruled by this government? This index of state history lies between 0 and 1 with higher values indicating high state antiquity and thereby a different institutional history.

PT

Oil and gas income (OilGas) denotes the value of a country’s oil and gas production (constant 2000 US dollars) divided by its midyear population and is included since resource-rich nations often use different

CE

production technologies for a sizeable share of national income (which would skew the β of our model). These

AC

data are from Ross (2012) who calculates oil rents by taking the total value of each country’s annual oil and natural gas production, and subtracting the country-specific extraction costs, including the cost of capital. Likewise, we use the standard measure of trade openness, as the sum of exports and imports of goods and services as a share of GDP, as well as the urbanization rate (the percentage of the total population living in urban areas). We take trade openness and urbanization rate from the World Bank Development Indicators (World Bank, 2014), as both could affect technology.

10

ACCEPTED To provide evidence suggesting the existence of aMANUSCRIPT causal effect of trust, we follow a recent strand in the trust literature by using a particular set of instruments for social trust suggested in Tabellini (2008) and Bergh and Bjørnskov (2011). Following Bergh and Bjørnskov (2011), we instrument the level of trust by a dummy variable capturing whether the predominant language of a country exhibits the pronoun drop characteristic, the average temperature in the coldest month of the year and a dummy for the existence of monarchical institutions, such as countries being ‘ruled’ by a king, emperor or other sovereign. All data are summarized in Table 1.

CR IP T

Table 1 approximately here

4. A Descriptive Look

Our main hypothesis states that social trust positively affects delegation decisions. A first look in Figure 2 at the

AN US

data we just outlined suggests that this might well be so. The figure clearly illustrates a strong, overall association between delegation and trust, consistent with H1. However, it also shows a number of strong outliers, not least Vietnam and Burundi (VIE and BUR in the figure) that appear to have either too little delegation or excessive trust scores. At the other end, delegation appears stronger than hypothesized in

M

Malaysia and the Philippines (MAY and PHI) that are low-trust societies.

ED

Figure 2 approximately here Yet, our second hypothesis states that although we would expect that trust is positively correlated with

PT

delegation, the theoretical considerations lead us to expect a substantially stronger association in relatively more economically and technologically developed societies, and a correspondingly weak association in poor

CE

countries. Indeed, when we divide our sample into two parts (80 developed and developing countries versus 40

AC

poor countries), there is no positive correlation between delegation and trust in least developed countries, as illustrated by the trend line through the black rectangles (the poor country observations). A consequence of the better fit is also that Vietnam seizes to be an outlier. The trend lines are therefore in line with the interpretation that high trust countries tend to have more decentralized decision-making processes and thus consistent with our theoretical priors (H2). In the following, we therefore explore whether these simple associations are evidence of an actual, economically meaningful relation. 11

ACCEPTED MANUSCRIPT 5. Methodology and Results In this section, we estimate the two types of linear regression models to investigate how trust is related to delegation: 𝐷𝑒𝑙𝑒𝑔𝑎𝑡𝑖𝑜𝑛𝑖 = 𝜋 + γTrust i + θ𝑍𝑖 + δXi + εi

(11)

𝐷𝑒𝑙𝑒𝑔𝑎𝑡𝑖𝑜𝑛𝑖 = 𝜋 + γTrust i + θZi + ϑTrust i Zi + δXi + εi

(12)

CR IP T

where Delegation is our index of delegation, Trust is the level of generalized trust, X is our vector of control variables, Z is a vector of potentially moderating control variables (testing H2), and ε is an unobserved error term. All standard errors are robust to heteroskedasticity.

AN US

5.1. Main results

Table 2 reports our OLS results testing our first hypothesis. As can be seen in column (1), the estimated coefficient of trust is positive and statistically significant at 1%. It is also economically significant, as a one standard deviation in trust is associated with approximately a 60 percent increase in the delegation measure. In

M

the next column, we include a post-communist dummy, ass several papers highlight that socialist regimes are

ED

known to suffer from centralized rule, have a lack of voluntary cooperation, and low trust levels (Kornai et al., 2003). Given the persistence of institutions and culture, we would consequently expect the post-communist

PT

dummy to bear a negative coefficient. As expected, our post-communist dummy variable is negatively and significantly associated with delegation. As noted above, in contrast to socialist regimes, Nordic countries are

CE

known to have decentralized rule, substantial voluntary cooperation and very high trust levels. Taking account

AC

of this in column (3) shows that Nordic countries do indeed have higher delegation. However, social trust remains highly influential even after controlling for post-communist and Nordic countries. Column (4) includes our measure of freedom in life. Freedom should increase individual responsibility, and may thus also affect decentralization (cf. Pitlik et al., 2015). Allowing for the effect of freedom is important since it might be strongly correlated with both trust and delegation. Consistent with such considerations, we find that freedom in life is positively associated with delegation and statistically significant at 1%. One might also expect weakly democratic institutions to limit cooperation and thus delegation, as well as individuals’ 12

MANUSCRIPT attitudes and corporate governance. ToACCEPTED control for such effects, we add a measure of democracy, which in column (5) turns out to have a positive and statistically significant association with delegation. Lastly, we include human capital level because the decision to delegate might both be easier and more profitable when there is a significant stock of human capital in the working population. As seen in column (6), the estimated coefficient of human capital is positive and statistically significant. However, even though the coefficient on trust in column (6) is reduced by about half, perhaps suggesting that human capital is a transmission channel for

Table 2 approximately here

5.2. Causality tests

CR IP T

some of the trust effect, it remains significant after including all of our control variables simultaneously.

AN US

Overall, our OLS results therefore suggest that there is a positive and statistically significant relationship between social trust and delegation. Our main finding is also robust even after controlling for a set of related control variables, and the results in Table 2 are in line with our expectations with regard to most of the

thus not indicate a causal association.

M

covariates. Yet, the OLS estimates in the table may arguably suffer from endogeneity or simultaneity bias and

ED

First, skeptics might argue that OLS results suffer from endogeneity problems in the sense that delegation could cause higher trust through, e.g., reputation effects (e.g. Hardin, 1992). Second, there might be omitted

PT

determinants of delegation that will be correlated with trust. Third, our estimates might suffer from a measurement error problem. Given the difficulties in collecting data on trust, there might be particular validity

CE

concerns about measurement error in trust in relatively poor societies (e.g. Holm and Danielson, 2005). We

AC

alleviate these three problems by using two-stage least square estimation (2SLS) with three different instrumental variables for trust. Justifications for the variables being utilized as instruments are as follows. Tabellini (2008) argues that languages that allow the personal pronoun to be dropped tend to give less emphasis to individual rights, which in turn reflects a culture of mistrust. It has similarly been argued that people have historically been more dependent on strangers for survival in relatively colder climates, an idea that dates back to Aristotle and Hippocrates. This factor would make high trust an evolutionary dominant strategy in colder climates 13

(Bjørnskov and Méon, 2013). Lastly, ACCEPTED Bjørnskov and MANUSCRIPT Méon (2013) finds and argues that people living in monarchies are more trusting because having a monarch family might provide social stability and represent a symbol of unity. Another potential reason for the relevance of monarchy as an instrument is that citizens may be more willing to accept monarchy if they are confident that monarchs will not cooperate with rent-seeking groups in society, i.e. that their high trust extends to the monarch. İn this situation, monarchy proxies for some stable underlying factor that creates trust, yet may not be properly exogenous. However, results in Bjørnskov

leaving a monarchy on average lose seven percentage points trust.

CR IP T

and Svendsen (2013) suggest that the association between monarchy and social trust is causal as emigrants

All three instruments have the benefit that they are clearly predetermined, as they derive from previous centuries. Using predetermined instruments for trust, we estimate the causal effect of trust on delegation. Our

AN US

first-stage regression equation is:

𝑇𝑟𝑢𝑠𝑡𝑖 = 𝜈 + θMonarchyi + λPronoundropi + ηMintempi + φXi + ξi

(13)

where Monarchy and Pronoundrop are dummies for monarchies and pronoun-drop characteristic and Mintemp is the average temperature in the coldest month. This strategy will be valid as long as our instruments

M

are uncorrelated with the error term εi in equation (1) – that is, if our instrumental variables have no effect on

ED

delegation other than through their impact on trust. Our 2SLS results are shown in Table 3, which shows that the first-stage F-statistics are above the critical value of 10, suggesting that we can reject the hypothesis that our

PT

instruments are weak at the 1% significance level.5 First-stage results are shown in Appendix Table 1. Sargan test results for the over-identifying restrictions also indicate that the instruments meet the exclusion restrictions.

CE

As expected, while pronoun-drop characteristic and average temperature is negatively related with trust, the

AC

legacy of monarchy is positively associated with trust. The 2SLS results show that social trust still has a positive and highly significant impact on delegation. The coefficient of trust in column (1) implies that a one standard deviation in trust is associated with approximately an 87 percent increase in delegation. Our 2SLS estimate of trust is thus larger than our OLS estimate, as is often the case with instrumented estimates. Although the instrumented estimate is not significantly different from the OLS estimates, the difference may perhaps suggest, as in other studies, that 5

The a priori relevance of our set of instruments is also confirmed by a high Shea R 2. Results are available upon request.

14

ACCEPTED MANUSCRIPT measurement bias in the trust variable creates a negative attenuation bias larger than any potentially positive biases from reverse causality and omitted variables.6 Columns (2) to (6) furthermore show that the effect of trust is robust to the inclusion of additional control variables. As shown in the appendix, where we employ an alternative instrumental approach in a somewhat smaller sample, these tests are not specific to our choice of instrumental variables.

CR IP T

Table 3 approximately here

5.3. Robustness tests

As a next step we further test the strength and robustness of our results by including other potential determinants of delegation in Table 4. A major alternative explanation of delegation is the quality of institutions

AN US

and regulations. First, including an index of business regulatory freedom, the positive parameter estimate for the business regulatory freedom in column (1) indicates that delegation is on average higher in countries with low burdens of business regulation. Second, when including a measure of rule of law known to be strongly associated with trust, results in column (2) show that the quality of legal institutions has a positive and

M

statistically significant effect on delegation. Reassuringly, the effect of social trust persists despite the inclusion

ED

of the institutional and regulatory environment. The emergence of early polities and state societies (state antiquity) is another important determinant of the quality of current institutions (Chanda and Putterman, 2007;

PT

Ang, 2013). State antiquity might also arguably affect corporate governance through affecting the quality of current institutions or proxying for remnants of colonial institutions. To control for this argument, we include

CE

an index of state antiquity in column (3). Yet, state antiquity has no significant effect on delegation while the

AC

effect of trust remains positive and statistically significant. Natural resources empower rulers as they might use the resource revenues to entrench themselves and their supporters, camouflage public budgets, impose restrictions on press freedom and block democratic and

6

Another source of the IV bias may be Dunning’s (2008) problem of heterogeneous identification of the instrumental variables.

Bjørnskov and Méon (2013) for example note that most standard instruments for social trust tend mainly to provide identification in societies above some level of economic development. If the effect is heterogeneous in development, as we show in the following, the instrumented average estimates will be biased in that direction.

15

MANUSCRIPT institutions reforms (Aidt, 2003; Ross, ACCEPTED 2012). In column (4), where we control for natural resources wealth, results show this inclusion nevertheless does not affect our main result. Second, countries’ trade policy might also force firms to change their organization structure. Local firms in an open trade regime are more exposed to competitive pressures, which will force them to decentralize its organization structure in order to be more productive and reap comparative advantage. Therefore, we augment our specification by including trade openness. As seen in column (5), trade openness has a positive effect on delegation, which becomes statistically

CR IP T

significant at 10%. In the next column, we control for urbanization to check the strength of the effect of trust. Urbanization is often observed as a consequence of economic complexity and development, and a structural transition away from agriculture. Technical and cultural changes that urbanization brings to our life might not only affect our political and institutional landscape, but also our ways of doing business. When we add the level

AN US

of urbanization into our analysis, we find that its effect on delegation is positive and statistically significant at 1%, as one would expect if urbanization is a proxy for technological development. Lastly, we take geography into account, as there is consensus that geography matters for the quality of institutions and culture. Therefore, one might argue that the effect of trust might just reflect geographical

M

differences. In column (6), we therefore take account of this argument by adding continent dummies into our

ED

analysis. Social trust nevertheless remains an important determinant of delegation even after controlling for geography. In sum, our results show that trust has a robust effect that is separate from institutions, regulation,

Table 4 approximately here

CE

PT

trade openness, urbanization and geography.

AC

5.4. Heterogeneous effects

As a final concern, our theoretical model shows that firms would not need to delegate authority in countries with non-complex economic structures where most production is undertaken with relatively simple techniques that allow for low-cost monitoring. In order to test the prediction in H2, we run a set of regression models with interaction terms as in equation (12). Our Z-vector of moderating variables proxying for level of economic complexity consists of GDP per capita, urbanization, level of democracy, and trade openness.

16

ACCEPTED MANUSCRIPT Our first variable for economic complexity is GDP per capita PPP (constant 2005 international $), which is the most immediately intuitive measure for the level of economic complexity. The Urbanization rate is our second variable for economic complexity, as it occurs when countries switch their economic structure from agriculture to industry. Therefore, urbanization is generally accepted as a product of modernization and industrialization. Due to this transition, there is often a strong relationship between urbanization and economic complexity (see Acemoglu et al., 2002; Henderson, 2003). Our third variable that proxies for the level of

CR IP T

economic complexity is democracy, as democratic political rights are argued to protect property rights, reduce entry barriers and support free enterprises, and may also offer superior environments in which to exploit advantages of new technologies (Acemoglu, 2008). In empirical work, Aghion et al. (2007) for example find that democracy affects productivity growth in different sectors differently. While democracy is conductive to

AN US

growth in more advanced sectors of an economy, it does not matter for growth in less advanced sector, suggesting that democratization or related institutional changes will make an economy more advanced and complex (e.g. Rode and Gwartney, 2012).

In Table 5, we include the full baseline specification from Table 2 although it is not shown. It is important

M

to note that with interaction terms, the single terms cannot be interpreted per se (cf. Brambor et al., 2006).

ED

Instead, for interactions that turn out to imply significantly heterogeneous effects, we plot the relation against trust in subsequent graphs, including 95 % confidence intervals.7 We first of all note in column (1) that the

PT

effect of trust on delegation is highly heterogeneous in economic development. The solid sloping line in the corresponding Figure 3 shows how the effect of trust changes with GDP per capita; 95% confidence intervals

CE

around this line allow us to determine the conditions under which trust has statistically significant effect on

AC

delegation. As can be seen in the figure, at low levels of GDP per capita, trust may even lower delegation, although negative estimates are never significant. Conversely, beyond a GDP per capita level of approximately 1100 $, the effect of trust becomes positive and significant above GDP per capita level of about 3300 $.8

7

To save space, we only provide the figure for the interaction term between trust and GDP per capita in the manuscript. Other figures

are available upon request. 8

In our sample we have 84 countries, which have GDP per capita (PPP) above 3300 $. Countries that are below per capita income

level of 3300 $ are: Liberia, Burundi, Ethiopia, Malawi, Sierra Leone, Mozambique, Rwanda, Madagascar, Mali, Uganda, Nepal,

17

ACCEPTED MANUSCRIPT Table 5 approximately here As shown in column 2, the effect of trust also becomes similarly positive as countries urbanize and turns significant at urbanization levels above 35% of the population. Consistently with institutional arguments, the results in column 3 of Table 6 also show that the effect of trust on delegation increases with the level of democracy and become significant at relatively low levels of political freedom around a Polity level of zero.9 Lastly, we find no evidence of any heterogeneity in openness to trade.

CR IP T

In summary, a set of relatively simple cross-sectional estimates, employing a new measure of delegation, show that social trust is a robust and statistically significant predictor of delegation. We observe approximately the same degree of heterogeneity across three indicators of sophistication, but not across openness, which may more allow countries to benefit from their comparative advantages, regardless of their

AN US

sophistication. With the significant heterogeneity in mind, we therefore proceed to discuss these results in the final section.

Figure 3 approximately here

M

6. Conclusion

ED

The literature on social trust entered the mainstream of economics, political economy and political science following the seminal contributions of Putnam (1993), Fukuyama (1995), and perhaps particularly Knack and

PT

Keefer (1997). It is now considered a determinant of long-run growth rates, institutional quality, and education.

CE

Recent studies also indicate support for Knack and Keefer’s (1997) preliminary suggestion that high-trust

Burkina Faso, Tanzania, Zambia, Bangladesh, Lesotho, Ghana, Kenya, Benin, Cambodia, Senegal, Kyrgyzstan, Nigeria, Pakistan,

AC

Vietnam, Yemen, India, Nicaragua, Moldova, Guyana, Cape Verde, Mongolia, Philippines, Indonesia and Honduras. We use GDP per capita PPP (constant 2005 international $) in 2005. Results are similar when we use GDP per capita PPP (constant 2011 international $) in 2005. Results are available upon request. 9

We note that a potential problem with the heterogeneity results in general is that they could have been driven by very uncertain trust

observations in poor and non-democratic countries (cf. Holm and Danielson, 2005). In further tests, we have therefore excluded all observations from countries with trust scores from either only one survey, or from several surveys where trust scores vary by more than 20 percent. These results, which are available upon request, do not change any inferences. We thank a reviewer for pointing out this problem.

18

MANUSCRIPT countries are more productive, all otherACCEPTED things being equal. However, much is still unknown about the likely transmission mechanisms between a culture of trust and productivity and economic performance. In this paper, we have revisited one of the original suggestions from the early literature: that social trust might affect the way work and production processes are organized. We outlined a simple model in which firms decide how much unskilled labor, skilled labor and management to hire. The decisions rest on the problem that costly monitoring is needed to ensure that skilled labor exerts a proper effort. Each firm thus makes a trade-off

CR IP T

between delegating decisions and risking that skilled labor does not exert optimal effort when not being closely monitored, or not delegating responsibility and thus incurring an additional management cost on the firm. We argue that in countries with higher trust levels, the monitoring problem is less important, implying that the optimal delegation choice is to delegate more responsibility than in low-trust countries. However, the

AN US

theoretical section also outlines how this choice is more important in countries with a more sophisticated production technology, as it rewards skills, but also exacerbates principal-agent problems as more skilled labor is difficult and costly to monitor.

In a large cross-section of countries, we find that countries exhibiting relatively higher trust levels indeed

M

do delegate more responsibility and more decisions. As far as instrumental variables techniques can inform

ED

about causal relations, the association between delegation and social trust appears causal. We also find support for our second hypothesis that this association is substantially stronger in countries with more developed and

PT

sophisticated production technologies: the pay-off of social trust in terms of delegation options is larger in more developed countries. In least developed countries, in which technologically complex production including non-

CE

monitorable work processes is uncommon, we find no evidence of trust effects on delegation.

AC

At the end of the day, delegation is likely to be an organizational option, which both allows firms to minimize their management costs and – in real life outside our simple modelling framework – allows firms to take advantage of local and potentially tacit knowledge in production decisions (Ikeda, 2008). However, this option rests on the risk of shirking when labor is not monitored. Our results thus points to a transmission mechanism connecting social trust and overall productivity, as high-trust countries have more access to lowcost delegation. Appendix 19

ACCEPTED MANUSCRIPT Table A1 approximately here As an alternative robustness check on our causality tests, we use trust levels of Americans whose grandparents came from different countries as an instrumental variable for the level of trust in their grandparents’ home countries. Empirical evidence shows that the level of trust is quite stable over time, and although this does not mean that the level of trust is completely unchanging, a substantial part of trust is transmitted from one generation to the next. Using US General Social Survey (GSS) in which in individuals are

CR IP T

asked about their parents’ country of origin, Rice and Feldman (1997) and Uslaner (2008) show that Americans whose grandparents came to the US from countries that have high levels of trust are trusting more than other Americans. Based on this stable pattern, the newly emerging literature examines the effect of culture on economic outcomes by using information on immigrants as a form of historically based identification (see

AN US

Fernandez and Fogli, 2005; Alesina and Giuliano, 2010; Algan and Cahuc, 2010; Bjørnskov and Svendsen, 2013). Following this literature, we here use ancestral trust as an instrument for our original trust variable. Ancestral trust is the current levels of trust of third-generation immigrants to the US. The original source of the data is the US General Social Survey, as reported in Bjørnskov and Svendsen (2013). 2SLS results with using

M

ancestral trust as instrument are shown in Table 5. These results also confirm that trust has positive effect on

ED

delegation.

Table A2 approximately here

PT

References

Acemoglu, Daron, 2008, Oligarchic versus democratic societies. Journal of the European Economic

CE

Association 6, 1-44.

AC

Acemoglu, Daron, Johnson, Simon, Robinson, James A., 2002. Reversal of fortune: Geography and institutions in the making of the modern world income distribution. Quarterly Journal of Economics 117, 1231-1294. Aghion, Philippe, Alesina, Alberto, Trebbi, Francesco, 2007. Democracy, technology, and growth. NBER working paper 13180. Aghion, Philippe, Algan, Yann, Cahuc, Pierre, Shleifer, Andrei, 2010. Regulation and distrust. Quarterly Journal of Economics 125, 1015-1049. Aghion, Philippe, Bloom, Nick, Van Reen, John, 2013. Incomplete contracts and the internal organization of 20

ACCEPTED firms. Journal of Law, Economics & Organization 30,MANUSCRIPT i37-i63. Aidt, Toke, 2003. Economic analysis of corruption: a survey. Economic Journal 113, F632-F652. Alesina, Alberto F., Giuliano, Paola, 2010. The power of the family. Journal of Economic Growth 15, 93-125. Algan, Yann, Cahuc, Pierre, 2010. Inherited trust and growth. American Economic Review, 100, 2060-92. Ang, James, 2013. Institutions and long-run impact of early development. Journal of Development Economics 108, 1-18.

CR IP T

Banfield, Edward C., 1958. The Moral Basis of a Backward Society. Free Press, Glencoe, IL. Banfield, Edward C., 1975. Corruption as a feature of governmental organization. Journal of Law and Economics 18 ,587–605.

Barro, Robert, Lee, Jong-Wha, 2010. A new data set of educational attainment in the world, 1950–2010. NBER

AN US

working paper 15902.

Bergh, Andreas, Bjørnskov, Christian, 2011. Historical trust levels predict the current size of the welfare state. Kyklos 64, 1-19.

Bergh, Andreas, Bjørnskov, Christian, 2014. Trust, welfare states, and income inequality: sorting out the

M

causality. European Journal of Political Economy 35, 183-199.

ED

Bjørnskov, Christian, 2009. Social trust and the growth of schooling. Economics of Education Review 28, 249257.

PT

Bjørnskov, Christian, 2010. How does social trust lead to better governance? An attempt to separate electoral and bureaucratic mechanisms. Public Choice 144, 323-346.

CE

Bjørnskov, Christian, 2011. Combating corruption: on the interplay between institutional quality and social

AC

trust. Journal of Law and Economics 54, 135-159. Bjørnskov, Christian, Méon, Pierre-Guillaume, 2013. Is trust the missing root of institutions, education, and development? Public Choice 157, 641-669. Bjørnskov, Christian, Svendsen, Gert, 2013. Does social trust determine the size of the welfare state? Evidence using historical identification. Public Choice 157, 269–286. Bloom, Nicholas, Sadun, Raffaella, Van Reenen, John, 2012. The organization of firms across countries. Quarterly Journal of Economics 127, 1663-1705. 21

ACCEPTED MANUSCRIPT Bloom, Nicholas, Van Reenen, John, 2010a. Why do management practices differ across firms and countries? Journal of Economic Perspectives 24, 203-224. Bloom, Nicholas and Van Reenen, John, 2010b. Human resource management and productivity. NBER Working Papers No. 16019. Bockstette, Valerie, Chanda, Areendam, Putterman, Louis, 2002. States and markets: the advantage of an early start. Journal of Economic Growth 7, 347-369.

CR IP T

Brambor, Thomas, Clark, Williams R., Golder, Matt, 2006. Understanding interaction models: improving empirical analyses. Political Analysis 14,: 63–82.

Bromiley, Philip, Cummings, Larry L., 1995. Transaction costs in organizations with trust. In: Bies, Robert J., Lewicki, Roy J., Sheeppad, Blair H. (Eds.), Research on Negotiations in Organizations. JAI Press,

AN US

Greenwich, pp.219–247.

Bromiley, Philip, Harris, Jared, 2006. Trust, transaction cost economics, and mechanisms. In: Bachmann, Reinhard, Zaheer, Akbar (Eds.), Handbook of Trust Research. Edward Elgar, Cheltenham, pp. 124–143. Chanda, Areendom, Putterman, Louis, 2007. Early starts, reversals and catch-up in the process of economic

M

development. Scandinavian Journal of Economics 109, 387-413.

ED

Cingano, Federico, Pinotti, Paolo, 2012. Trust, firm organization and the structure of production. Working Papers 053, Carlo F. Dondena Centre for Research on Social Dynamics (DONDENA), Università

PT

Commerciale Luigi Bocconi.

Dearmon Jacob, Grier, Robin, 2011. Trust and the accumulation of physical and human capital. European

CE

Journal of Political Economy 27, 507-519.

AC

Dunning, Thad, 2008. Model specification and instrumental variables regressions. Political Analysis 16, 290302.

Fernandez, Raquel, Fogli, Alessandra, 2005. Culture: an empirical investigation of beliefs, work, and fertility. American Economic Journal: Macroeconomics 1, 146-177. Fukuyama, Francis, 1995. Trust: The Social Virtues and the Creation of Prosperity. Free Press, Glencoe, IL. Gallup

World

Poll,

2013.

Gallup

world

poll.

Online

http://www.gallup.com/consulting/worldpoll/24046/About.aspx (accessed December, 2013). 22

database

at

MANUSCRIPT Guiso, Luigi, Sapienza, Paola, Zingales,ACCEPTED Luigi, 2008a. Trusting the stock market. Journal of Finance 63, 25572600. Guiso, Luigi, Sapienza, Paola, Zingales, Luigi, 2008b. Alfred Marshall Lecture: Social capital as good culture. Journal of the European Economic Association 6, 295-320. Hardin, Russell, 1992. The street-level epistemology of trust. Analyse & Kritik 14, 152-176. Henderson, J. Vernon, 2003. Urbanization and economic development. Annals of Economics and Finance 4,

CR IP T

274-341. Holm, Håkan J., Danielson, Andreas, 2005. Tropic trust versus Nordic trust: experimental evidence from Tanzania and Sweden. Economic Journal 115, 505-532.

Horváth, Roman, 2013. Does trust promote growth? Journal of Comparative Economics 41, 777-788.

AN US

Ikeda, Sanford, 2008. The meaning of “social capital” as it relates to the market process. Review of Austrian Economics 21, 167-182.

Kaufmann, Daniel, Kraay, Aaart, Mastruzzi, Massimo, 2009. Governance matters VIII: aggregate and individual governance indicators, 1996-2008. World Bank policy research working paper 4978, the World

M

Bank, Washington DC.

ED

Knack, Stephen, Keefer, Philip, 1997. Does social capital have an economic pay-off? A cross-country investigation. Quarterly Journal of Economics 112, 1251–1288.

PT

Kornai, János, Rothstein, Bo, Rose-Ackerman, Susan, 2003. Creating Social Trust in Post-Socialist Transition (Political Evolution and Institutional Change). Palgrave Macmillan, London.

CE

La Porta, Rafael, Lopez-de-Silanes, Florencio, Shleifer, Andrei, Vishny, Robert V., 1997. Trust in large

AC

organizations. American Economic Review Papers and Proceedings 87, 333-338. Marshall, Monty G., Gurr, Ted R., Jaggers, Keith, 2010. Polity IV project: political regime characteristics and transitions, 1800-2012. Codebook and data available at http://www.systemicpeace.org/inscr/inscr.htm (accessed December 2013). Nooteboom, Bart, 2000. Trust as governance device. In: Mark, Casson, Godley, Andrew (Eds.), Cultural Factors in Economic Growth. Springer, Berlin, pp.44-68.

23

ACCEPTED MANUSCRIPT Papagapitos, Agapitos, Riley, Robert, 2009. Social capital and human capital formation. Economics Letters 102, 158-160. Pinotti, Paolo, 2012. Trust, regulation and market failures. Review of Economics and Statistics 9, 650-658. Pitlik, Hans, Redín, Dulce M., Rode, Martin, 2015. Economic freedom, individual perceptions of life control, and life satisfaction. In: Gwartney, James, Lawson, Robert, Hall, Joshua (Eds.), Economic Freedom of the World: 2015 Annual Report. Fraser Institute: Vancouver, pp. 185-202.

Economics working paper 1307, Universitat Pompeu Fabra.

CR IP T

Ponzetto, Giacomo A.M., Troiano, Ugo, 2012. Social capital, government expenditures, and growth.

Putnam, Robert, 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton University Press, Princeton, NJ.

AN US

Rajan, Raghuram G., Zingales, Luigi, 1998. Financial dependence and growth. American Economic Review 88, 559-586.

Rice, Tom W., Feldman, Jan L., 1997. Civic culture and democracy from Europe to America. Journal of Politics 59, 1143-1172.

ED

Journal of Political Economy 28, 607-619.

M

Rode, Martin, Gwartney, James D., 2012. Does democratization facilitate economic liberalization? European

Ross, Michael L., 2012. The Oil Curse How Petroleum Wealth Shapes the Development of Nations. Princeton

PT

University Press, Princeton, NJ.

1313-1332.

CE

Sapienza, Paola, Toldra-Simats, Ana, Zingales, Luigi, 2013. Understanding trust. The Economic Journal 123,

AC

Schwab, Klaus, Sala-i Martin, Xavier, 2013. The Global Competitiveness Report 2013-2014. World Economic Forum, Geneva.

Tabellini, Guido, 2008. The scope of cooperation: values and incentives. Quarterly Journal of Economics 123, 905-950. Uslaner, Eric M., 2002. The Moral Foundations of Trust. Cambridge University Press, Cambridge, MA. Uslaner, Eric M., 2008. Where you stand depends upon where your grandparents sat: The inheritability of generalized trust, Public Opinion Quarterly 72, 725-740. 24

ACCEPTED MANUSCRIPT Williamson, Oliver E., 1971. The vertical integration of production: Market failure considerations. American Economic Review 61, 112-123. Williamson, Oliver E., 1975. Markets and Hierarchies: Analysis and Antitrust Implications. Free Press, New York, NY. Williamson, Oliver E., 1985. The Economic Institutions of Capitalism. Free Press, New York, NY. World Bank, 2014. World Development Indicators. Online database, the World Bank, Washington DC.

Figure 1 A P-A game

CR IP T

Zak, Paul and Stephen Knack, 2001. Trust and growth. Economic Journal 111, 295-321.

a-wm

ED

1

M

AN US

ws-e

b ws-e µ 1-µ a ws

AC

CE

PT

wu

25

Figure 2 Social trust and delegation

ACCEPTED MANUSCRIPT

6 5.5 MAY PHI

4.5 4 3.5

VIE

2.5 BRU

2 0

10

20

30 40 Social trust

Developed countries

60

70

CE

PT

ED

M

AN US

Poor countries

50

CR IP T

3

AC

Delegation

5

26

ACCEPTED MANUSCRIPT Figure 3 Heterogeneity of trust effects in economic development 0.06 0.05 0.04

Trust effect

0.03 0.02 0.01 0

CR IP T

-0.01 -0.02 -0.03 -0.04

M

AN US

GDP per capita

Mean

Standard deviation

122 122 122 122 118 120 109 115 122 112 118 118 122 119 122 122 122 57

3.82 25.15 0.20 0.04 67.01 3.07 23.09 5.79 0.11 0.46 2.55 92.43 57.37 8.83 0.15 0.73 10.69 33.52

0.74 13.44 0.40 0.19 16.68 5.50 15.14 1.14 0.97 0.23 3.00 58.91 23.29 1.28 0.36 0.45 10.42 13.98

PT

Observations

CE

AC

Delegation Trust Post-Com Nordic FreedomLife Democracy SecEdu BusReg Rule StateAntq OilGas Open Urban GDPpc Monarchy Pronoundrop Mintemp Ancestral Trust

ED

Table 1 Descriptive statistics

27

ACCEPTED MANUSCRIPT Table 2 OLS results

Post-Com

(2) 0.032*** (0.005)

(3) 0.024*** (0.006)

(4) 0.022*** (0.005)

(5) 0.020*** (0.005)

(6) 0.015*** (0.005)

-0.321*** (0.102)

-0.302*** (0.102)

0.121 (0.117)

0.114 (0.114)

-0.176 (0.165)

0.964*** (0.269)

0.609*** (0.209)

0.550*** (0.191)

0.646*** (0.143)

0.022*** (0.002)

0.019*** (0.002)

0.019*** (0.003)

0.023** (0.010)

0.014 (0.010)

Nordic

FreedomLife

Democracy

CR IP T

(1) 0.033*** (0.005)

Trust

0.010*** (0.003)

SecEdu

118 117 105 0.592 0.619 0.653 174.12 139.57 128.91 parentheses. ***, ** and * denote statistically significant at the 1

AC

CE

PT

ED

M

AN US

No of Obs. 122 122 122 R2 0.355 0.385 0.431 F stat 38.80 25.58 244.42 Note: Robust (Heteroscedasticity-adjusted) standard errors are in percent, 5 percent and 10 percent, respectively.

28

ACCEPTED MANUSCRIPT Table 3 2SLS results

Post-Com

(2) 0.050*** (0.006)

(3) 0.050*** (0.008)

(4) 0.038*** (0.007)

(5) 0.034*** (0.006)

(6) 0.031*** (0.008)

-0.244** (0.121)

-0.245*** (0.119)

0.146 (0.122)

0.134 (0.116)

-0.081 (0.164)

-0.003 (0.269)

0.036 (0.224)

0.082 (0.213)

0.167 (0.254)

0.021*** (0.003)

0.019*** (0.003)

0.019*** (0.003)

0.019* (0.010)

0.010 (0.011)

Nordic

FreedomLife

Democracy

CR IP T

(1) 0.048*** (0.006)

Trust

0.007* (0.004)

SecEdu

118 15.76 0.411 0.533 215.20

117 14.24 0.831 0.578 199.51

AN US

No of Obs. 122 122 122 F stat (first stage) 13.91 20.52 14.50 Sargan Test 0.282 0.390 0.388 R2 0.277 0.272 0.273 F stat (second 51.43 73.36 261.07 stage) Note: Robust (Heteroscedasticity-adjusted) standard errors are in percent, 5 percent and 10 percent, respectively

105 10.41 0.897 0.606 161.89

AC

CE

PT

ED

M

parentheses. ***, ** and * denote statistically significant at the 1

29

ACCEPTED MANUSCRIPT

Table 4 Further results Trust

(1) 0.022*** (0.008)

BusReg

0.172*** (0.052)

(2) 0.023** (0.010)

(3) 0.031*** (0.000)

(4) 0.029*** (0.009)

(5) 0.034*** (0.008)

(6) 0.026*** (0.008)

(7) 0.026** (0.013)

0.173* (0.88)

Rule

StateAntq

-0.153 (0.239) 0.016 (0.017) 0.001* (0.000)

Open

CR IP T

OilGas

0.006*** (0.002)

Urban

No

No

No

No

No

All Controls

Yes

Yes

Yes

Yes

Yes

No

Yes

Yes

Yes

AN US

Continents

AC

CE

PT

ED

M

No of Obs. 102 105 100 103 102 105 105 F stat (first 12.49 6.27 8.37 9.18 10.54 9.14 4.08 stage) Sargan Test 0.880 0.817 0.879 0.898 0.744 0.255 0.629 R2 0.683 0.667 0.595 0.609 0.609 0.656 0.651 F stat (second 112.77 133.63 137.37 84.72 144.79 114.76 95.41 stage) Note: Robust (Heteroscedasticity-adjusted) standard errors are in parentheses. ***, ** and * denote statistically significant at the 1 percent, 5 percent and 10 percent, respectively

30

ACCEPTED MANUSCRIPT

Table 5 Interaction effects

(2)

(3)

(4)

Trust

-0.069*** (0.020)

-0.005 (0.006)

0.009* (0.004)

0.017 (0.011)

Log GDPpc

-0.034 (0.071)

Trust x GDPpc

0.009*** (0.002)

Urban

-0.001 (0.004)

Trust x Urban

0.000*** (0.000)

Democracy

-0.028 (0.017)

Trust x Democracy

0.001*** (0.000)

AN US

Open

Trust x Open

Controls

Yes

CR IP T

(1)

Yes

Yes

0.002 (0.03)

-0.000 (0.000) Yes

AC

CE

PT

ED

M

No. of Obs 103 105 102 105 R2 0.740 0.705 0.677 .675 F stat 99.25 67.14 95.24 151.92 Note: Robust (Heteroscedasticity-adjusted) standard errors are in parentheses. ***, ** and * denote statistically significant at the 1 percent, 5 percent and 10 percent, respectively. In each specification, dummies for post-communist and Nordic countries, freedom in life, democracy and education are included, but not shown.

APPENDIX Appendix Table A1 First-stage results 31

(1)

ACCEPTED MANUSCRIPT (2)

Monarchy

13.744*** (2.527)

Pronoundrop

-5.994** (2.527)

Mintemp

-0.324*** (0.103) 0.772*** (0.113)

Ancestral Trust

AC

CE

PT

ED

M

AN US

CR IP T

No of obs. 122 57 R2 0.331 0.470 F stat 13.91 46.11 Note: Results in column (1) represent first-stage results of column (1) of Table 3. Results in column (2) represent first-stage results of column (1) of Table (5). Robust (Heteroscedasticity-adjusted) standard errors are in parentheses. ***, ** and * denote statistically significant at the 1 percent, 5 percent and 10 percent, respectively.

32

ACCEPTED MANUSCRIPT Appendix Table A2 Results with alternative instrument (1) 0.046*** (0.007)

Trust

Post-Com

(2) 0.050*** (0.006)

(3) 0.049*** (0.009)

(4) 0.041*** (0.008)

(5) 0.040*** (0.011)

(6) 0.052** (0.020)

-0.355** (0.203)

-0.355* (0.201)

0.035 (0.160)

0.032 (0.159)

-0.095 (0.244)

0.024 (0.337)

-0.108 (0.276)

-0.095 (0.355)

-0.424 (0.601)

0.023*** (0.004)

0.023*** (0.005)

0.023*** (0.006)

0.001 (0.021)

-0.030 (0.036)

Nordic

FreedomLife

CR IP T

Democracy

SecEdu

0.005 (0.006)

AC

CE

PT

ED

M

AN US

No of Obs. 57 57 57 57 56 50 F stat (first 46.11 68.82 51.09 42.93 21.73 10.04 stage) R2 0.402 0.395 0.398 0.593 0.597 0.529 F stat (second 39.50 51.53 150.23 121.83 97.07 68.47 stage) Note: Robust (Heteroscedasticity-adjusted) standard errors are in parentheses. ***, ** and * denote statistically significant at the 1 percent, 5 percent and 10 percent, respectively

33