Finance and development: Rethinking the role of financial transparency

ARTICLE IN PRESS

JID: JBF

[m5G;December 20, 2019;12:40]

Journal of Banking and Finance xxx (xxxx) xxx

Contents lists available at ScienceDirect

Journal of Banking and Finance journal homepage: www.elsevier.com/locate/jbf

Finance and development: Rethinking the role of financial transparencyR,RR Burak R. Uras Tilburg University, CentER, European Banking Center, The Netherlands

a r t i c l e

i n f o

Article history: Received 19 June 2019 Accepted 10 December 2019 Available online xxx JEL classification: E44 G2 O16 O47 Keywords: Financial development Transparency Adverse selection Under-investment

a b s t r a c t Over the last decade many developing countries strengthened their transparency standards with the objective of improving asset market allocations and macroeconomic outcomes. This paper develops a general equilibrium model and argues that in a financially underdeveloped economy - with uninsurable consumption risk and stochastic-investment - enforcing financial transparency might be counterproductive. The framework builds upon a standard property that illiquid asset markets cause under-investment in assets that pay in the long-run, because individually rational agents hoard cash to exploit sales of underpriced long-term assets. First, I show that in this environment private revelation of news about investment-returns could give a chance to sell low-quality assets and then characterize the conditions under which the lack of financial transparency reduces under-investment and improves macroeconomic development. An empirical analysis reveals that the theoretical predictions of the model is in line with cross-country data.

1. Introduction Asset markets in developing countries lack substantial liquidity compared to their counterparts in high-income countries. Developing country markets are also characterized by limited disclosure of information between sellers and buyers, which gives asset originators the chance to hide information related to business fundamentals and causes adverse selection. Based on these two stylized facts, the literature on finance and development perceives adverse selection as an imperfection that needs to be corrected if the objective is to sustain an efficient allocation of capital across its users.1 This paper develops a novel theory and argues the opposite: adverse selection could improve allocation efficiency in economies with illiquid asset markets. Dating back to the conclusions of Stiglitz and Weiss (1981) development economists have long associated adverse selection with

R

Previously circulated as “Efficient Lemons”. I would like thank Wolf Wagner for many valuable conversations and recommendations on this paper. I also would like to thank two anonymous referees, Thorsten Beck, Thanasis Geramichalos, Cyril Monnet and seminar participants at Tilburg University and Central Bank of Hungary for comments and suggestions. All remaining errors are mine. E-mail address: [email protected] 1 For example, Feldman and Kumar (1995), Singh (1997), Rajan and Zingales (20 01, 20 03) and Kaminsky and Schmukler (2008) argue that better financial transparency standards are likely to create aggregate welfare gains. RR

© 2019 Elsevier B.V. All rights reserved.

the concepts of “credit rationing”, “market breakdowns”, “inefficient allocations” and eventually with “low macroeconomic development”. Cross-country data indeed exhibits a positive correlation between macro-development and financial transparency: Business Extent of Disclosure Index of World Bank, which measures a country’s enforced public disclosure of business financial information, correlates with real per-capita income.2 Policy makers throughout the world - and especially those in developing countries - continuously implement regulatory changes in order to lower the opacity of asset market transactions and to alleviate informational deficiencies. For instance, World Bank’s Doing-Business Report reveals that emerging-market economies and developing countries such as Albania, Azerbaijan, Georgia, Kazakhstan, Macedonia, Sri Lanka, Ukraine and Uzbekistan underwent a series of regulatory changes over the period of 2005–2017 and substantially improved the degree of information dissemination in their asset markets (World Bank Group (2005-2017).) There is a subtle property of adverse selection that largely got overlooked in the past literature: whether a long-term project is a dead-end lemon investment might not be known to the initial fi-

2 In Section 6 a formal regression analysis will be presented to document this pattern. Additionally it is important to note that using a relatively narrower sample of countries - compared to the World Bank data - Porta et al. (1998) also show that accounting standards exhibits a positive correlation with macroeconomic development.

https://doi.org/10.1016/j.jbankfin.2019.105721 0378-4266/© 2019 Elsevier B.V. All rights reserved.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

JID: JBF 2

ARTICLE IN PRESS

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

nanciers (originators) by the time of project initiation. Intuitively, this uncertainty would make long-term investment costly and undermine macroeconomic development if in expectation investment opportunities are of positive net present value. Private revelation of asset-return news could then give a chance to originators to sell low-quality assets before realization of returns. This mechanism could potentially raise incentives to finance long-term investment ex-ante. The important question that arises is what conditions would characterize an economic environment that would induce adverse selection to be welfare-improving. This paper shows that a consumption-based early asset-sale motive in a liquidity-constrained market could assign an aggregate welfare-improving role to adverse selection. Specifically, I develop a general equilibrium model to study the interactions between consumption risk, investment uncertainty and adverse selection in the presence of liquidity constrained spot markets. In the model agents can invest in a stochastic long-term asset or hold cash, where the risk associated with long-term investment is uninsurable ex-ante. A speculative motive for holding cash emerges because the spot market where assets are traded for cash is liquidity constrained and asset originators who are subject to early consumption shocks are willing to sell underpriced assets. While individually rational, the cash-holdings reduce aggregate long-term investment and lower macroeconomic output (and welfare), since in expected terms investment projects are of positive net present value. Given this economic environment, I study the type of adverse selection that arises because asset originators receive private news about future asset returns. The quality news arrive at an interim date after the asset origination has taken place and before the realization of stochastic investment returns. Importantly, early consumption shocks also arrive privately and hence the spot market cannot distinguish whether the trading motive of an asset seller is information based or liquidity-need based. The model thus exhibits three key features of developing country financial markets: (i) imperfect insurance against consumption (liquidity) shocks, (ii) investment-return uncertainty and (iii) privately arriving news on liquidity shocks and investment returns. The general equilibrium analysis reveals that while (i) is always counterproductive, in the presence of (i), (ii) and (iii) could improve macroeconomic welfare. The intuition for this highlighted result is as follows: when the spot market is developed enough to fully absorb the liquidity risk, assets are priced based on fundamentals and then investment uncertainty and adverse selection do not have any effect on expected gains of investors and liquidity providers - leaving the aggregate investment unchanged. This is not the case when constrained liquidity in the spot market leads to Allen and Gale (1994) style cashin-the-market pricing, where asset prices reflect liquidity shortages. With scarce liquidity, if investment returns are uncertain, asset prices in the spot market do not fully adjust when the average quality of assets supplied deteriorates due to adverse selection, which gives for a patient asset originator a chance to swap a low-quality asset with an asset of average market quality. This opportunistic market participation by originators with low quality assets increases the gains from long-term investment relative to cash hoarding and leads to a reallocation of resources towards long-term asset origination. While I demonstrate the theoretical structure in the most tractable benchmark setup, I show that the analysis and the key result is extendable to a wide-spectrum of alternative model specifications. For instance, I also examine asset type choice (private information arrival vs public information arrival) by asset originators and then endogenous information acquisition by suppliers of liquidity and show the robustness of the key model insight with respect to endogenizing the information structure of the economy.

In another important extension, I evaluate the case where liquidity shortages in the spot market do not occur in every aggregate state of economy. This extension allows to observe that as spot markets develop - and become increasingly more liquid - the welfareimproving effects of adverse selection on allocation efficiency start to diminish, highlighting the relevance of the key theoretical mechanism for underdeveloped - liquidity-constrained - markets. Finally, I provide an empirical analysis and show that crosscountry data is in line with the novel theoretical prediction from the paper: in a spectrum of cross-country regressions with real GDP per capita as a left-hand-side variable the interaction between “disclosure standards & illiquid market dummy” is negative and significant while “disclosure standards” by itself has a positive and significant coefficient estimate. This suggestive empirical evidence thus supports that the theoretical results from this paper may be relevant in the context of disclosure regulations conducted in developing countries over the last decade. Related literature. This paper contributes to several important lines of research. First and foremost, the paper is the most relevant for the literature on counterproductive consequences of financial development. In this literature, Pagano and Jappelli (1994) develop a model to show that borrowing limits and economic growth could display a negative correlation. Bencivenga et al. (1995, 1996) and Bencivenga et al. (20 0 0) investigate the non-monotone consequences of financial market trading in dynamic general equilibrium economies. Castro et al. (2004) isolate supply and demandside effects of rising investor protection using an overlappinggenerations model and show that the supply-side effect can induce investor protection to be welfare-reducing in economies with restrictions on capital-flows. More recently Philippon (2010) and Philippon (2015) studies the macroeconomic costs of an oversized financial sector. The current paper contributes to this literature by introducing information frictions in a model of limited consumption insurance and investment uncertainty and uncovers the caveats associated with strict transparency regulations in economies with under-developed asset markets. Second, the paper relates to the literature on transparency, finance and economic development. In this literature, Feldman and Kumar (1995) and (Singh, 1997) propose the argument that limited transparency could be a barrier to stock market development and economic growth.3 In line with this theoretical argument, in their seminal paper (Porta et al., 1998) collect accounting standards data from 44 countries and document a positive correlation between per capita income and the quality of accounting standards, which as the authors also state “to a large extent is a consequence of disclosure rules”. Utilizing the data from Porta et al. (1998), Chinn and Ito (2006) show that the quality of accounting standards do not necessarily predict a growth in stock markets in both developed and developing countries. Chinn and Ito (2006)’s result implies that transparency, asset markets and economic development do not necessarily move together in the long-run. My paper also relates to this line of research by studying the theoretical interactions between illiquid asset markets, financial transparency and economic development. I formalize that there could be a potential counterproductive interaction between market illiquidity and transparency enforcement that policy might need to pay attention to. I also provide cross-country empirical evidence that supports this theoretical prediction, in which I use a newer proxy that directly measures enforcement of information disclosure at the aggregate level. Third, the key findings of the paper are relevant for the studies on social returns of information in the context of financial markets. 3 Rajan and Zingales (2001) and Rajan and Zingales (2003) reach a similar conclusion concerning the interaction between limited transparency and the overall development of the financial sector.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

JID: JBF

ARTICLE IN PRESS

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

In this literature, building upon the insights by Hirshleifer (1971), Andolfatto et al. (2014), and Dang et al. (2017) study the optimality of aggregate information disclosure on asset returns in models of risk-shifting. Also related to this strand of research, Monnet and Quintin (2013) demonstrate that when secondary markets are shallow, more information can reduce the expected payoff of stakeholders who need to liquidate their positions early. My paper also argues for the optimality of information non-disclosure. However, the current paper differs from the existing studies in several important dimensions. First, in this paper it is not the economy-wide ignorance that promotes economic well-being, but private information - and the adverse selection induced by it - that is essential for macro-development. Second, my findings reveal the optimality of asset swaps allowed by opacity, which turn out to improve aggregate allocation efficiency. I show that this mechanism is the most relevant in under-developed (liquidity-constrained) financial markets. Third, the results hold even for the case of low risk aversion and as a matter of fact with risk-neutral agents as I analyze in the benchmark specification - unlike for the models analyzed by Andolfatto et al. (2014) and Dang et al. (2017). The remainder of the paper is organized as follows. Section 2 presents the benchmark analysis. Section 3 incorporates an endogenous asset structure choice to the benchmark. Section 4 discusses further extensions and robustness checks. Section 5 provides the cross-country empirical analysis. Section 6 concludes.

2. Benchmark analysis I investigate an economy with three dates (0,1, and 2). There is a continuum of risk-neutral financiers (agents) with unit measure. On date 0, each agent is allocated with 1 unit of a physical endowment, called cash. Cash can be consumed on any date and stored costlessly (with a one-to-one rate of return) in-between any two dates. Agents are subject to a random liquidity demand for consumption on date 1, which is uninsurable ex-ante. The consumption demand is assumed to arrive with probability 21 at the beginning of date 1. Those agents who are hit by the liquidity shock value consumption on date 1 only. The remaining agents value date-1 and date-2 consumption equally - though I will concentrate on allocations at which they consume on date-2 only. For the rest of the paper, the date-1 consumers will be referred as “early consumers” or “impatient agents” whereas the date-2 consumers will be called “late consumers” or “patient agents”. The date-1 demand for liquidity is private information. Neither impatient nor patient agents value consumption on date 0. The expected life-time utility of an agent is thus given by

V =

1 [c1 + c2 ], 2

(1)

where c1 and c2 denote the consumption on date 1 and 2, respectively.4 On date 0, each agent can invest a fraction of his cash holding into an (agent-specific) asset. For each unit of cash invested on date 0, the asset pays R on date 2 with probability q, and otherwise zero, where q ∈ (0, 1). The return realizations are uncorrelated across assets and investment risk cannot be insured ex-ante. I assume that assets are of positive net present value: qR > 1. The assets that will pay out on date 2 will be called high quality assets

4 In Section 5 I relax the assumption of risk-neutrality and perfect substitutability across periods.

3

and the non-paying assets will be called low quality assets. Assets cannot be scrapped on date 1.5 , 6 At the core of my analysis is the revelation of asset quality information. Specifically, at the beginning of date 1, information about the date-2 return for each asset arrives. I investigate two different information structures: (i) Opaque assets - with private information: only the originator of the asset receives the information about asset quality; and, (ii) Transparent assets - with public information: the information about asset quality is public to all agents in the economy. In this section’s analysis the information structure is exogenously imposed in order to evaluate the aggregate implications of adverse selection borne by opaque assets. In Sections 3 and 4, I incorporate endogenous transparency and information acquisition decisions and study their macroeconomic implications. 2.1. Efficient allocation I start by characterizing the socially efficient allocation. Specifically, let us consider a social planner who manages the aggregate endowment of cash as well as the aggregate stock of assets in the economy and decides on the allocation of consumption between date 1 and date 2. Let us denote with η the (aggregate) share of funds kept in cash between dates 0 and 1, such that 1 − η then becomes the aggregate asset origination. I use variables c˜1 and c˜2 to denote the aggregate consumption to be made available on date 1 and date 2, respectively. Since on date 1 only impatient agents have a strict demand for consumption, all consumption made available on date 1 will go to impatient consumers, while date-2 consumption will go to patient consumers - as long as date-1 consumption does not exceed date2 consumption. Recalling that the consumption levels of impatient and patient investors are c1 and c2 , we have:

c1 = 2 c1 and c2 = 2 c2 . The social planner solves:

max

c˜1 ,c˜2 ,η

1 [c˜1 + c˜2 ], 2

(2)

s.t. c˜1 ≤ η,

(3)

c˜2 ≤ (1 − η )qR,

(4)

c1 = 2 c1 , c2 = 2 c2 , c1 ≤ c2 .

(5)

Given that I have assumed qR > 1, the solution to the planer’s problem is to invest all available cash on date-0 in productive assets (η = 0) and to set  c1 = 0 and  c2 = qR. This implies

c1 = 0, c2 = 2qR, and the macroeconomic output and consumption in the economy is qR. Given linear utility, macroeconomic welfare is then also given by

Vsoc = qR.

(6)

5

In Appendix F I relax this assumption. I would like to highlight that the assumptions of “no date-2 return for low quality assets” and “50% chance of being hit with a date-1 liquidity need” are made without loss of generality, because the key properties of the baseline specification remain unchanged in expense of notational burden if I were to assume general parametrizations for the value of low-state return and for the distribution of liquidity shocks. 6

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF 4

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

Proposition 1. In equilibrium: 1. the aggregate portfolio contains assets of positive measure, i.e. η < 1, 2. the aggregate portfolio contains cash holdings of positive measure, i.e. η > 0, 3. there is cash-in-the-market pricing, i.e. ptr < R, 4. date-0 portfolios satisfy a no-arbitrage condition between the expected return to originate an asset (EA ) and the expected return to move forward with cash (EC ), i.e. EA = EC ,7 5. the market clearing condition on date-1 is given by Timeline 1. Benchmark.

1 tr 1 p q (1 − η ) = η , 2 2 characterizing the equilibrium price of a high-quality asset:

2.2. Equilibrium with transparent assets I now consider the market equilibrium, allowing agents to trade assets in a spot market on date 1. To represent a developing country situation, the spot market is assumed to be liquidity constrained, such that it provides an incomplete insurance against early-consumption risk. In this section I characterize the equilibrium with transparent assets, which I extend to the case of opaque assets with private information revelation in Section 2.3. I start with defining the equilibrium concept, followed by the general properties of the equilibrium. Both the concept of the equilibrium as well as its general properties resemble those of the cash-in-the-market pricing equilibrium of Allen and Gale (1994). Building upon these equilibrium properties, I will solve for the market clearing and the optimal portfolio choice of the agents and then characterize the macroeconomic performance with transparent assets. A. Equilibrium concept On date-0, each agent decides how much cash to invest in originating an asset (1 − η) and how much cash to carry forward to date-1 (η). On date-1 the agent learns whether he is hit with a liquidity shock, i.e. whether he is an early consumer or not. If he is an early consumer he visits an asset market and supplies there his entire asset holding in exchange of cash and consumes his endof-date-1 cash holding. If he is a late consumer, he can supply cash in the market in exchange of assets and consume the realized portfolio return on date-2. The timing of events is summarized in Timeline 1. The date-1 asset market opens for transactions after the realization of liquidity shocks and the arrival of asset-specific public news about quality. Given the public nature of the arriving information, high quality assets trade in the market at the unit price of ptr . On date-0, agents make a portfolio choice to maximize the expected utility by taking as given the deterministic date-1 asset price ptr in addition to expected returns to holding cash and originating assets. The equilibrium is thus characterized by the following conditions as in Allen and Gale (1994): (E1) ptr clears the asset market on date-1, where - given public revelation of asset quality information on date-1 - only high quality assets could be supplied by asset originators. (E2) In the asset market, investors’ portfolios maximize expected utilities as of date-0. B. Properties of the equilibrium Before I characterize the equilibrium portfolio choices and the implied aggregate welfare in the following proposition I provide some general properties of the equilibrium, which also largely follow from Allen and Gale (1994).

ptr =

η . ( 1 − η )q

Proof. See Appendix A.



From property 5 we can observe the cash-in-the-market pricing feature of the asset price: the price is determined by the ratio of available cash and the quantity of assets offered for sale, i.e. η/(1 − η )q. Importantly, this situation arises because there is no outside source on date-1 which could supplement the spot market on date-1 with additional liquidity. This is a characteristic feature of an isolated asset market, which is highly relevant in the context of developing countries with underdeveloped financial markets. C. Portfolio choice Having determined the spot market clearing, I can now characterize the date-0 equilibrium portfolio choice. Asset origination allows the impatient agent to earn ptr on date 1 with probability q. The return to a patient asset originator is R which also gets realized with probability q. Therefore, the expected life-time value of originating one unit of asset as of date 0 is expressed as

EA =

 1  tr qp + qR , 2

which after substituting in the asset price becomes

EA =





1 η + qR . 2 1−η

(7)

As for moving forward with cash on date-0; the unit return on cash equals 1 if the agent turns out to be impatient. If the agent is patient, one unit of cash can buy p1tr units of the asset on date 1, giving a return of

R ptr

on date 2. Hence, the expected life-time re-

turn from holding onto one unit of cash until date 1 is expressed as

EC =





1 R 1 + tr . 2 p

Substituting the market-clearing price ptr this expression becomes

EC =





1 1−η 1+ qR . 2 η

(8)

On date 0, agents choose their portfolio allocation based on the expected returns from holding onto cash and originating an asset. In the competitive equilibrium Property 4 holds and expected returns from holding cash and originating an asset must be equalized. Solving (7) and (8) together yields the equilibrium quantity of cash that gets carried over and the price of transparent assets:

ηtr =

1 , 2

7 Properties 1–4 are valid also with a concave utility specification as well as under general distributions of liquidity shocks and arbitrary specifications of the values of high-and-low state asset returns.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

ptr =

1 . q

I obtain for (the macroeconomic output and) the expected life-time utility8 : tr Vmarket = EA = EC =

1 [qR + 1]. 2

(9)

Comparing the expected life-time value of an agent with transparent markets (9) against the expected life-time value under a social planner’s regime (6), the following result is derived. tr Proposition 2. Vsoc > Vmarket .

Proposition 2 shows that the allocation implied by the transparent market is sub-optimal. The reason is that the opportunity to buy assets at prices below fundamental value (ptr < R) induces agents to hold cash (ηtr > 0) and invest less on date-0. Holding cash however is inefficient due to the assumption of qR > 1, hence a lower level of macro-welfare results. To observe this let us consider a modified environment with risk-free asset investment, where asset returns are not subject to idiosyncratic shocks. We can note that with risk-free (rf) investment, given properties 1–3, expected returns to originate assets and to hold cash and the market clearing conditions are expressed as the following:

 1 rf p + qR , 2   1 qR EC = 1 + rf , 2 p EA =

1 rf 1 p (1 − η ) = η , 2 2 using which together with Property 4 rf Vmarket =

1 [qR + 1] 2

follows: reducing investment uncertainty does not have any consequence on macroeconomic output. To the contrary, as we see next, investment uncertainty is needed to have adverse selection play a welfare-enhancing role in the presence of liquidity constrained asset markets. I would like to conclude with the following remark. The linear model structure, which is a deviation from the standard Diamond and Dybvig (1983) way of modelling banking, helps with obtaining tractable model properties. One such property is that transparent asset markets are characterized by under-investment and not by under-provision of liquidity as in Diamond and Dybvig (1983). Although, it is important for clean closed-form solutions, as I discuss in Section 4 an extended set-up, resembling Diamond and Dybvig (1983) framework more closely, does not necessarily revert the theoretical properties I highlight in the paper.

5

Plantin (2009). When assets are opaque, the market cannot distinguish between low-quality and high-quality assets on date1, and there is thus a single spot market for both asset-quality types where adverse selection prevails. The recursive solution with opaque assets - though resembles the general properties 1–4 listed above - distinguishes itself from the case of transparent assets by one important dimension, which I will delineate in this section. I again first state the equilibrium concept and then delineate on the characterization of the competitive equilibrium. A. Equilibrium concept The equilibrium concept utilized for the case of transparent assets largely carries over into the case of opaque assets - with Timeline 1 continuing to describe the sequence of events. Since low quality and high quality assets co-exist on the supply side of the date-1 spot market that are indistinguishable from each other, one price continues to clear the market, which I denote with pop . The co-existence of low quality and high quality assets in date-1 market will yield the additional equilibrium property below that patient sellers of low-quality would re-inject liquidity (obtained from asset sales) back into the market. This can be sustained in equilibrium, if sellers and buyers are anonymous and trading takes place until there is no outstanding demand left to trade further. Given this trading structure the following revised features characterize the equilibrium. (E1) pop clears the asset market on date-1, where - given private revelation of asset quality information on date-1 - both high quality and low quality assets could be supplied by asset originators. (E2) Sellers of assets are allowed to re-purchase other assets, such that trade continues until there is no excess supply or demand in the market. (E3) In the asset market, investors’ portfolios maximize date-0 expected utilities. B. Properties of the equilibrium The first four of the equilibrium properties that I derived with transparent assets fully remain into the case of opaque assets. In addition, with opaque assets the equilibrium exhibits the following properties summarized in the next proposition. Proposition 3. With opaque asset markets: 1. Patient sellers of low quality assets, who obtain additional liquidity from having sold assets in the market, re-inject these funds obtained from asset sales into the market to purchase assets.9 2. The market clearing is stated as

1 op 1 1 p (1 − η )(2 − q ) = η + pop (1 − η )(1 − q ), 2 2 2



 Liquidity Demand

Liquidity Supply

solving for the price of an opaque asset 2.3. Equilibrium with opaque assets Suppose now that all assets originated on date-0 are of opaque nature. This means originators of assets receive private news regarding future payoffs before they enter the spot market on date1, in the spirit of “learning by holding” property analyzed in 8 I would like to note that an identical price for transparent assets and date-0 valuations would be reached if one were to define transparency as “fully absence of information-arrival” at originator and purchaser sides of the assets. Specifically, in the case of no-information arrival on date-1 with respect to asset quality, the price of the transparent asset in equilibrium would get pinned down by ptr = 1−η η , in which case all impatient holders of assets sell in the date-1 market. Then, the implied values associated with both originating assets and holding cash as of datetr = EA = EC = 12 [qR + 1]. 0 would still be characterized in equilibrium as Vmarket

pop =

η

1−η

.

Proof. See Appendix A.



C. Portfolio Choice Since fractions of high quality assets (supplied by only impatient originators) as well as all low quality assets (supplied by both impatient and patient originators) are sold in the market, the likelihood of purchasing a high quality asset in the market turns out

9 This property also goes through with a concave utility specification and general distributional assumptions on liquidity shocks and the specification of the values of high-and-low state asset returns.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF 6

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx 1q 2 1 + 1 (1−q ) 2 2

to be

=

q 2−q .

Therefore, the expected life-time utility gen-

erated from asset origination as of date 0 is





1 op q R p + qR + (1 − q ) pop , 2 2 − q pop   1 op 3 − 2q = p + qR . 2 2−q

EA =

3. Transparency choice in equilibrium

Substituting in the asset price pop this yields:

 η 1

EA =

2 1−η

+ qR

 3 − 2q

.

2−q

(10)

Remark. For a given aggregate quantity of cash carried over to the date-1 asset market, asset opacity raises the expected benefits from originating an asset. This important property of the model can be appreciated by keeping η fixed and observing that the expected return from originating an asset under transparency (7) is lower than the expected return from investment under opacity (10). The reason for this property is that opacity gives the originators a positive probability to replace a low quality asset with a high quality one on date 1, which as we see below alleviates the under-investment inefficiency. The expected life-time utility from moving forward to date 1 with one unit of cash is expressed as





q R 2−q 1 EC = 1 + op , 2 p

which together with the market clearing implies

EC =





1 1−η 1 1 + qR . 2 η 2−q

(11)

Setting (10) and (11) equal, I obtain that the date-0 equilibrium allocation of cash that gets carried over to date 1 fulfils

1 2

ηop < , and also that the equilibrium price level in the market satisfies

pop <

is higher than the returns to investment with transparency, and hence more investment takes place in an equilibrium with opacity compared to an equilibrium with transparency. The rise in asset origination raises the macroeconomic output in equilibrium, since qR > 1. I note that asset swap works only because of the presence of adverse selection in the date-1 asset market.10

1 2−q

for Property 4 (EA = EC ) to hold since qR > 1. Therefore, when assets are of opaque nature, the economy carries over less cash from date 0 to date 1 compared to the case of transparent markets. Since qR > 1, this implies that macroeconomic development is higher with opaque assets relative to the case of transparent assets. Of course, since positive quantities of cash get carried over to date 1 (guaranteed by Property 2), the aggregate welfare with asset opacity is still lower than the aggregate welfare implied by the planer’s allocation. op

tr Proposition 4. Vsoc > Vmarket > Vmarket .

The intuition behind this important result, that establishes the qualitative model feature of efficient adverse selection, can be explained as follows. When assets are opaque, patient investors can opportunistically unload low quality assets in the date-1 asset market and obtain high quality assets with positive probability - supplied by the q-fraction of impatient sellers. This implies an asset swap possibility for originators of assets, which decreases the likelihood of ending up with a low quality asset on date-2 for those agents who are patient. However, there is no corresponding (offsetting) reduction in the utility of impatient suppliers of high quality, because due to the liquidity-constrained asset market prices are determined by the available cash in the market and not by the quality of assets supplied. Thus, in an economy where assets are priced below fundamentals, with asset opacity the returns from investing in assets

Studying equilibrium transparency choice allows to address two important questions. First, it allows me to understand whether asset opacity could be an equilibrium choice for forward-looking agents. Second, and as importantly, I would like to study whether there could be any inefficiency associated with decentralized opacity decisions, that is, whether a social planer would choose a different degree of overall market opaqueness compared to the decentralized asset opacity choices of individual agents. Addressing both of these questions will shed light on the generalizability of counterproductive development implications of financial transparency that I derived in the previous section. I extend the setup from Section 2 as follows, while maintaining the general features of the model that give rise to properties 1–5. On date 0 agents choose whether to hold cash or to originate an asset, as before. Different from the analysis of Section 2, however, asset originators decide whether to make the originated asset opaque - and receive private news on date-1 regarding date-2 payoffs - or leave it as transparent. Creating an opaque asset requires a (non-monetary) cost equivalent of cO units of consumption. I incorporate this cost figure in order to have a non-trivial equilibrium choice between opacity and transparency and also to capture a cost for foregoing transparency that is orthogonal to the mechanism of the paper - but relevant for development consequences of opacity. Especially in the context of developing countries, for financial investors originating assets in novel & innovative industries is likely to be a costly process compared to investing in more traditional industries. Such costs are expected to be borne by investors to a certain extent because of the time and effort it takes for searching for novel investment options. Originating assets in novel segments of the market is also subject to financial opacity due to the lack of prior information associated with these industries. In this respect, by incorporating costly opacity creation into the benchmark model, I also capture a relevant property associated with opacity investment in developing countries. Opacity creation is also subject to a stochastic process. Specifically, when the opacity cost is incurred on date 0, the originated asset turns out to be opaque on date 1 with probability α (α ∈ (0, 1)). With probability 1 − α , as of date 1 it will be identical to a transparent asset. Therefore, not all of the “ex-ante” opaque assets turn out to become opaque “ex-post”. This randomness in opacity creation allows for the coexistence of opaque and transparent assets on date 1, despite the linear structure of the model.11 The realization of a particular asset’s opaqueness on date 1 is public information, i.e. all agents in the economy know on date-1 whether a particular asset is opaque or transparent. The adjusted timing of events is presented in Timeline 2.

10 In order to appreciate the role of asset swapping for aggregate outcomes, one can assume a trading structure where sellers on date-1 are not allowed to re-enter the market as buyers. Under this set-up we can observe that ηop equals 12 , which is the quantity cash carried forward on date-0 with transparent assets. Thus, without asset swapping the implied aggregate welfare of transparent and opaque asset markets would coincide with each other. 11 Co-existence of opaque and transparent assets can also be achieved by incorporating an optimal opacity choice at the level of a financier (asset originator). However, this requires a deviation from the current linear structure of the model.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

7

agent cannot affect the aggregate outcomes. Hence, unilateral deviations of a single agent from the rest of the originators’ behavior will not alter the market clearing condition stated at (13). I can therefore express the return from sourcing transparent assets as before:

EAtr =

 1  tr qp + qR . 2

(15)

An asset originator is indifferent between staying transparent and op originating an opaque asset if EAtr = EA , or more explicitly if

α (1 − q ) qR

2−q

2

I will solve for the equilibrium first and then contrast the private returns to originate opaque assets against the social returns to opacity in order to discuss the efficiency properties of equilibrium opacity. I note that since transparent and opaque assets can now coexist, there are separate markets for high quality transparent assets and for opaque assets. The recursive solution for the competitive equilibrium is presented as follows. On date 1, a proportion 1 − α of the ex-ante opaque assets turns out to be transparent. When such assets are of low quality, they are not traded in any market. For an (ex-post) opaque asset the probability of being high quality in the opaque asset market – as q before – is 2−q . Absence of arbitrage opportunities between the two markets requires that the relative price of assets fulfills

ptr 2−q = . pop q

(12)

ptr pop

ptr pop

<

2−q q )

or transparent as-

2−q q )

sets (the case of > are not traded in date 1 markets, neither of which can constitute an equilibrium. As in the benchmark analysis of Section 2, patient opaque asset originators that end up with low quality assets sell their original assets and buy assets sold by other asset originators. There is )α mass of (1 − η ) (1−q of such originators. Utilizing (12), we can 2 op then solve for p as

pop =

η . (1 − η )[(2 − q )(1 − α ) + α ]

(13)

The expected return from an opaque asset can be expressed as the sum of a term that captures the expected return from originating a transparent asset and another term that captures the net benefit from opacity:

EAop

 1  tr = qp + qR + 2



α (1 − q ) 2



qR ( − pop ) − cO . 2−q

(14)

The benefit from opacity, represented by the term α (12−q ) ( 2qR −q − pop ) in (14), arises because opacity may enable the asset originator to opportunistically swap a low quality asset in the date 1 opaque assets market with a high quality one, but this has to be weighed against the date-0 cost of opacity cO . The return to moving forward with cash on date-0 is expressed as in Section 2:







(16)

Combining (16) with the equilibrium no-arbitrage condition stemop ming from Property 4 (EAtr = EA = EC ) I obtain

Timeline 2. Opacity choice.

Otherwise, opaque assets (the case of



− pop = cO .



1 R q 1 R EC = 1 + op = 1 + tr . 2 p 2−q 2 p I derive next the condition at which - as of date 0 - asset originators are individually indifferent between opacity and transparency. Given that we have a continuum of agents, an individual

ptr =

1 , q

(17)

pop =

1 . 2−q

(18)

Plugging (18) in (16), I obtain the following results. Proposition 5. Agents invest in opacity on date 0 if and only if α (1−q ) (qR − 1 ) > cO . 2(2−q ) Corollary 1. All assets originated on date-0 are of opaque nature if cO = 0. Proposition 5 implies if agents are allowed to choose the degree of their asset transparency freely in an environment with relatively low cost opacity creation, a high degree of market opaqueness could prevail. Corollary 1 shows that in the special corner of cO = 0 (costless opacity origination) all date-0 asset investment would be in opaque assets. Furthermore, when cO = 0 the higher the degree of asset opacity - α - the larger would be the macroeconomic output. This result confirms that if the degree of financial transparency is a choice variable for originators of assets, the liquidity-constrained asset market would exhibit limited financial transparency. The result also shows that reducing the cost of opacity - by lowering cO or increasing α - would improve the macroeconomic development. On social efficiency of equilibrium opacity. Next I investigate whether the equilibrium-opaqueness produced by markets is constrained efficient also for the general case of cO > 0. Specifically, I compare the (net) expected utility of a laissez-faire economy to the (net) expected utility achieved when a social planner (a transparency regulator in the aggregate) dictates investment in asset opacity on date-0, but leaves agents free to choose the date0 allocations in terms of holding cash and investing in assets. Specifically, a transparency regulator in this current context can be thought as a policy maker, who lets markets clear on date-1 (at equilibrium prices {pop , ptr }) and decides on whether on date-0 individual agents should be induced to invest in opacity or not. We know that when there is no investment in opacity, which is equivalent to the case of α = 0 in the current context, the equilibrium quantity of cash carried forward from date 0 would be η (α = 0 ) = 12 . I also note that with opacity investment on date 0, equilibrium cash holdings satisfy

η (α > 0 ) =

(2 − q ) − α (1 − q ) , 2 (2 − q ) − α (1 − q )

which is less than η (α = 0 ) = 12 for all parameter values. This 1 means incurring the cost of opacity when pop = 2−q and generating opaque assets with probability α raises the aggregate quantity of assets originated on date-0. I derive the aggregate benefit of

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF 8

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

opacity relative to full-transparency as

This is an important finding, because it shows that for any level of cO opacity-vs-transparency choice of individual agents is socially efficient. The intuition for this property is related to the liquidity constrained asset market: when the asset market is liquidity constrained, cash-in-the-market pricing prevails and the asset prices are pinned down by the no-arbitrage condition. This implies that the quantity of cash carried to the date-1 market at such prices gets passively determined. Specifically, the no-arbitrage between holding liquidity and originating assets ensures that any motive of asset originators to source opacity in order to scrutinize liquidity holders would result in a 1-to-1 reduction in asset prices. Since the implication of this mechanism is a general equilibrium contraction in liquidity carried over to date 1, with endogenous asset opacity adverse selection continues to exhibit a welfare-improving property.12

These two extensions are also important to highlight the channels which strengthen the welfare-improving consequences of asset opacity through its impact on long-term investment on date0. My analysis reveals that the aggregate welfare implications of long-term investment induced by opaqueness are stronger whenever the liquidity shortages on date-1 are severe (representing a developing country financial market). I also show that the utilityloss associated with not compensating early liquidity needs of impatient agents dampens the aggregate implications of opacity. With respect to the latter the linear utility specification analyzed in the benchmark provides a framework, where not-compensating early liquidity-needs does not generate any utility-loss, and hence the aggregate welfare benefits from asset opacity are maximized. This result implies that the benefits of financial opacity through boosting long-term investment relate to the transferability of (linear) utility in-between patient and impatient agents. Furthermore, in Section 3 I have assumed that sourcing opaque assets bears a cost for the asset originator. Alternatively, there could also be a productivity differential between opaque assets of private-information nature and transparent assets. This case is considered in Appendix D, where I assume that opaque assets yield lower returns than transparent assets. In this case opacity can also imply a cost for the eventual buyers of opaque assets (as they obtain lower return assets). However, as Appendix D shows, the private and the social returns to opacity are still equalized when the cost of opacity is introduced through a productivity loss, and hence the same qualitative results as in Section 3 are obtained. Fourth, in Appendix E I introduce the possibility of information acquisition for cash holders. The results show that information acquisition undermines the benefits from opacity, namely the benefits of private information and the resulting adverse selection. However, when access to information-acquisition platform is limited, welfare-improving opacity continues to prevail as a key property of the extended model set-up. Fifth, in the benchmark analysis I have assumed that assets cannot be discontinued on date 1, i.e. the asset scrap value is zero, as a result of which impatient agents are willing to sell assets in the market at any price. Introducing scrapping creates a benefit to transparency, as transparency allows assets to be discontinued on date 1 when expected asset returns are low. This means when asset scrapping is allowed, the standard Akerlof-style adverse selection mechanism becomes active, in which case opaque secondary asset markets may hurt the efficient origination of assets in the primary market. As I show in Appendix F, opacity raises aggregate welfare as long as the scrap value of assets are sufficiently low, in other words, when the standard adverse selection mechanism is not strong enough.

4. Extensions

5. Empirical evidence: transparency, liquidity and development

In the Appendix of the paper I present a spectrum of additional extensions and conduct robustness checks to show that the key result of the paper is not driven by special case structural assumptions. In this section I refer to these further extensions and leave the details to Appendix sections. The benchmark analysis in Section 2 was based on two simplifying assumptions to ensure tractability and a clean illustration of the key insight from the theoretical model: (i) transferable risk-neutral (linear) preferences and (ii) a cash-in-the-market pricing structure without aggregate uncertainty. In Appendix B and Appendix C I relax these assumptions and show the validity of the main mechanism of my model under these general formulations.

The novel conclusion from the model is that adverse selection could have a positive influence on economic welfare if asset markets are sufficiently underdeveloped, such that market liquidity is scarce and assets are priced below fundamentals. This positive effect of opacity survives with endogenous asset transparency choice and endogenous information acquisition, and also under alternative structural assumptions. In order to understand the empirical validity of the highlighted theoretical argument of the paper in this section I refer to World Bank data and conduct a cross-country empirical analysis. The empirical question that I am interested in addressing is whether the interaction between low market liquidity and financial transparency has a negative association with aggregate economic development. The key variable of interest that I need to proxy for this purpose is financial transparency of asset originators. World Bank Doing Business database provides a suitable indi-

(η (α = 0 ) − η (α > 0 ) )(qR − 1 )   1 (2 − q ) − α (1 − q ) = − (qR − 1 ) 2 2 (2 − q ) − α (1 − q ) =

α (1 − q ) (qR − 1 ), 2[2(2 − q ) − α (1 − q )]

(19)

where the first multiplicative term on the right-hand-side of (19) is the expansion in the base of the asset origination on date-0 provided by opacity (relative to the case of full transparency) and the second multiplicative term is the net consumption gain from originating an asset. I also derive the aggregate cost of opacity as

(1 − η (α > 0 ) )cO =

2−q cO , 2 (2 − q ) − α (1 − q )

(20)

where the first multiplicative term on the right-hand-side of (20) is the total quantity of assets originated and the second multiplicative term is the unit cost of date-0 investment in opacity. Finally, comparing (19) against (20), I can show that in the aggregate the economy’s social planner would be indifferent between letting private agents invest in opacity on date 0 and having the private agents keep their assets fully transparent if and only if

α (1 − q ) (qR − 1 ) = cO . 2 (2 − q )

(21)

I can note that the condition in Proposition 5 and (21) are identical to each other. Hence, there is no wedge between private returns and social returns to opacity for any cO > 0. Proposition 6. The private and social returns to opacity coincide with each other.

12 I would like to note an important remark. This result is analytically derived thanks to the linear structure of the model. With concave preferences - although maybe still be valid - supporting this property is not analytically tractable.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx Table 1 Descriptive statistics.

Log Real GDP Per Capita Business Extent of Disclosure Index Market Liquidity (% of GDP) Property Rights Index Legal Rights Index Primary-School Enrollment (%) Secondary-School Enrollment (%) Tertiary-School Enrollment (%)

Mean

St Dev Min

Max

# of Obs

8.46 5.07 48.18 2.96 4.55 103.72 79.63 35.80

1.44 2.25 77.46 0.31 1.45 4.44 14.86 14.73

11.68 10 952.66 4 7 112.71 105.41 76.51

2878 2878 1200 102 102 102 102 102

5.01 0 0.00 2.25 1.15 89.50 39.71 9.73

Notes. All data items are from World Bank’s World Development Indicators.

cator to proxy financial transparency: Business Extent of Disclosure. The business extent of disclosure index is a country-level indicator for the protection of external investors - by law - through enforced public disclosure of business financial information. Among other measures of transparency, importantly, this indicator also captures the enforcement of external audit requirements imposed by a country’s law to publicly disclose business financials to potential investors. World Bank started to construct this index in 2005 and therefore the cross-country analysis that I conduct will concentrate on the period of 2005–2017.13 In order to understand the interactions between market liquidity, financial transparency and economic development I estimate various forms of the following cross-country regression specification using ordinary-least-squares:

ln(GDP PCapita )it = α + β Discit + γ MLit + θ (Il l iquidit × Discit ) + φ  Xit + μt + it .

(22)

The dependent variable of the analysis is the log of real percapita GDP - proxying the level of macro development. The key right hand-side variable is Disclosure, captured by business extent of disclosure index, which varies between 0 and 10.14 Another important explanatory variable for the empirical analysis is Market Liquidity (ML) of the country, which is captured by the total value of stocks traded in the market as a percentage of GDP - measuring the level of asset market liquidity in the economy. Utilizing the data on market liquidity I also create a dummy variable to proxy an “illiquid financial market status” for a country. The average market liquidity ratio in the cross-country sample - for the years between 2005 and 2017 - roughly equals to 50% of GDP. I assign Il l iquid = 1 if ML is less than 50% for a country in a particular year. In a subset of the regressions I also condition the macro-development with additional right-hand-side variables (X), namely property rights, legal rights and enrollment rates at primary, secondary and tertiary schools, which are likely to be relevant factors to explain macroeconomic development. Finally, μt denotes the time fixed-effects. All data variables used in the analysis are from WDI database of World Bank. Descriptive statistics for the relevant range of variables are presented in Table 1. The empirical results are presented in Table 2. The first two columns show that there is a highly significant positive correlation between transparency and the level of economic development in specifications both with and without time fixed effects. This is an expected empirical pattern, which was also pointed out by Porta et al. (1998) using cross-country data on accounting standards. Economically developed countries are more lenient towards information sharing and following the findings of the past liter13 An alternative to the business extent of disclosure index is Porta et al. (1998)’s accounting standards. However, accounting standards data has a narrow coverage of countries and also it is rather an indirect measure of disclosure as also stated by the authors of the paper. 14 Higher values of business disclosure index proxy stricter enforcement of transparency.

9

ature it is natural to expect that there could be positive consequences of transparency (and the lack of adverse selection) that are not directly captured in the benchmark model for the sake of tractability and ease of exposition of the novel insights. For instance, as covered in Section 3 and Appendix D of the paper opaque assets could be costly to create or of lesser productivity for reasons that are orthogonal to the mechanism uncovered here - which could generate welfare gains from transparency. The claim of my theoretical foundation is that the gains from transparency would be substantially lower in countries with illiquid asset markets to which I turn next. Columns III and IV of Table 2 present the key empirical results. In those regressions I maintain the disclosure index as a right hand-side variable and additionally include stock market liquidity and the interaction between the dummy for being an illiquid market and disclosure index. The results show that while the disclosure index continues to have a positive coefficient estimate to explain per-capita income, the interaction between an illiquid (underdeveloped) stock market and disclosure is negative and significant at 1% level. This interaction effect - which supports the theoretical argument presented in the previous sections - is economically significant as well, because as the coefficient estimates reveal in countries with relatively illiquid stock markets the positive association between transparency and development is roughly 40–50% less strong compared to the economies with relatively more liquid stock markets. Also fulfilling standard conjectures, regression results show that market liquidity itself has a positive and statistically significant coefficient estimate in explaining economic development. The empirical results presented in Columns III and IV are robust to the inclusion of time fixed effects in the regression specification. Finally in columns V and VI of Table 2, I analyze the effects of inclusion of additional variables - that could condition aggregate welfare - on the empirical interaction between transparency, liquidity and economic development. The variables that are included are legal rights index - capturing the enforcement of collateral and bankruptcy laws, property rights, and measures of school enrollment ratios. To the right hand side of the regressions I also incorporate interactions of legal rights and property rights with the illiquidity status of a country. I control for these additional interaction terms to understand whether the interaction of market liquidity with an institutional variable is especially negative for the case of disclosure index or whether it exhibits negative coefficient estimates with other institutional variables as well. Results from columns V and VI confirm the earlier findings (both with and without time fixed effects): disclosure has a significantly positive level effect on economic development, but this positive effect gets completely reduced (and even becomes negative) for economies with illiquid asset markets. Therefore, the key empirical pattern that is relevant for our theoretical analysis is robust to the inclusion of variables that condition the level of economic development. Other variables of interest in this regression have largely expected signs: property rights and economic development are positively related for the case of financially underdeveloped economies (unlike for the interaction of liquidity with disclosure) and secondary and tertiary school enrollment rates are positive significant explanatory variables for the economic development. Legal rights and market liquidity are not significant determinants of economic development. It is not too surprising that some variables do not come out significant in these two regressions. I attribute this outcome to the fact that when all variables are included in the analysis, the sample size shrinks to a total of 102 observations with a time span of 4 years. Nevertheless, I continue to observe a negative association between transparency and economic development in countries with illiquid asset markets.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF 10

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx Table 2 Economic development, market liquidity and transparency. Dependent variable: log real GDP per capita

Disclosure

I

II

III

IV

V

VI

0.147∗ ∗ ∗ (0.011)

0.139∗ ∗ ∗ (0.011)

0.147∗ ∗ ∗ (0.019) 0.001∗ ∗ ∗ (0.000) −0.059∗ ∗ ∗ (0.013)

0.142∗ ∗ ∗ (0.020) 0.002∗ ∗ ∗ (0.000) −0.061∗ ∗ ∗ (0.013)

0.053 No

0.066 Yes

0.118 No

0.140 Yes

0.305∗ ∗ (0.144) -0.000 (0.000) −0.494∗ ∗ ∗ (0.123) −0.685∗ ∗ ∗ (0.244) 0.001 (0.068) 1.000∗ ∗ ∗ (0.215) −0.145 (0.110) −0.040∗ ∗ ∗ (0.011) 0.023∗ (0.014) 0.024∗ (0.013) 0.738 No

0.292∗ ∗ (0.137) 0.000 (0.000) −0.486∗ ∗ ∗ (0.119) −0.725∗ ∗ ∗ (0.232) −0.031 (0.070) 0.950∗ ∗ ∗ (0.209) −0.105 (0.112) −0.038∗ ∗ ∗ (0.013) 0.019 (0.013) 0.028∗ ∗ (0.013) 0.768 Yes

Market Liquidity (% of GDP) Illiquid Mkt x Disclosure Property Rights Legal Rights Illiquid Mkt x Prop R Illiquid Mkt x Legal R Prim-School Enroll (%) Sec-School Enroll (%) Tert-School Enroll (%) R-sq Year FE Observations Min # of Countries Time-Span

2,878 208 2005–2017

1,200 78 2005–2017

102 24 2013–2016

Notes. ∗ ,∗ ∗ ,∗ ∗ ∗ denote 10%, 5% and 1% significance. Robust standard errors are reported in parentheses.

The empirical analysis presented in this section has a descriptive nature and therefore it should not be taken as a causal inference due to shortcomings of an OLS specification with crosscountry data and relatively small sample size. However, the robust negative association between transparency and illiquidity in explaining economic development that is documented throughout regression specifications suggests an indicative evidence that the theoretical argument highlighted in the paper might be carrying important insights for policy makers in developing countries. 6. Conclusion I have developed a framework to study the general equilibrium implications of lack of transparency (opacity) in financial markets and the implied adverse selection when assets are priced below fundamentals. The key result from the analysis shows that asset opacity improves macroeconomic development by spurring investment and diminishing the incentives to hold liquidity. The mechanism rests on the fact that when the asset market is illiquid, asset prices become less sensitive to fundamentals. Opacity then allows holders of low quality assets to sell in the market and replenish their asset holdings. This in turn has general equilibrium implications: asset origination becomes more attractive - improving the aggregate welfare of the economy. The results are robust with respect to endogenous opacity choice of asset originators, endogenous information acquisition of liquidity providers and a spectrum of deviations from benchmark structural assumptions and thus are likely to be relevant for financial development policies conducted in merging markets and developing countries. The findings of the paper are complementary to the past research - both from theoretical and empirical perspectives. As highlighted throughout, it is only natural to expect also positive implications of financial transparency on macroeconomic development that works through different channels, that did not constitute the core focus in this paper. The important message from my analysis is that transparency is likely to have an adverse consequence on macro outcomes, “if asset markets are liquidity-constrained”. This prediction is in line with cross-country empirical evidence.

Appendix A. Proofs A1. Proof of Proof of Proposition 1 Property 1. In order to see why this should be the case, suppose that no assets are originated on date-0. Since agents are forward-looking, this implies that on date-1 there is no asset demand, which is possible only if cash is un-dominated in the sense that ptr > R. That would mean ptr > qR > 1 holds in equilibrium, yielding that holding cash (whose unit return equals 1) is dominated by holding an asset between date-0 and date-1. This contradicts that no assets are originated on date-0. Property 2. If there is no cash carried forward from date-0, then there would not be any demand for the assets supplied in the spot market on date-1. Then, ptr = 0 must hold, which means asset origination is dominated by holding cash between date-0 and date-1 - another contradiction. Property 3. Suppose not, such that ptr = R. Expected life-time return to originate an asset then equals to qR, while expected return to hold onto cash equals 1. Since qR > 1, then no cash would be carried forward to date-1 (since its return is dominated), violating property 2. Properties 1, 2 and 3 imply that expected returns to originate an asset (EA ) and to hold cash (EC ) as of date-0 must satisfy the following additional property. Property 4. If not, either asset or cash is not held in equilibrium, violating property 1 or property 2. Property 5. From Property 3 (cash-in-the-market pricing) it follows that liquidity is inelastically supplied in the date-1 spot market by patient cash-holders. Also because of Property 3 only those high-quality assets are supplied in the spot market that are held by impatient agents. The market clearing condition on date-1 is thus given by

1 tr 1 p q (1 − η ) = η , 2 2 where the left-hand-side is the demand for liquidity on date-1 and the right-hand-side is the supply of liquidity on date-1. 

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

A2. Proof of Proposition 3 Property 1. If not, then this will mean that for each unit of cash held on date-1, it is more profitable to hold onto the cash - and carry it forward to date-2 - rather than to supply it in the spot market. This would then in turn imply that pop = 0, such that holding an asset between date-0 and date-1 is dominated by holding cash. Therefore, no asset will be originated on date-0, which would violate Property 1. Property 2. On date-1 - because of cash-in-the-market pricing - all impatient investors and patient investors with low quality assets sell regardless of the market price, whereas patient investors of high quality assets hold onto their assets. Also because of the cash-in-the-market pricing, patient cash holders supply all of their liquidity to purchase assets. By Property 5 patient sellers of low quality assets use the liquidity obtained from selling assets to acquire new assets in the market. The (gross) supply of liquidity in the date-1 asset market thus comes both from liquidity that has been carried over, and from opportunistic buying of patient agents who sold low quality assets. The total proceeds from selling by the latter group is given by 1 op 2 p (1 − η )(1 − q ), thus I obtain for the total liquidity supply:

1 η 2 

+

which yield EC > EA . It follows from Property 4 then that in equilibrium, the allocation of cash (η∗tr ) should satisfy η∗tr > 12 , which is larger than the welfare-optimal amount. The equilibrium with transparent assets thus remains inefficient. The resulting expression for the welfare (expected utility) becomes:



Liquidity Demand





  z η∗tr − (η∗tr + q(1 − η∗tr )) . 2



i.

ηop

1−ηop

> z,

ηop ii. 1− ηop < z.

1 ηop 3 − 2q ( 1 + β )z + − z + qR 2 1 − ηop 2−q





1 1 − ηop EC = (1 + β )z + 1 − z + qR 2 ηop

(23)

op Vmarket =

,

  1 2−q

,

qR(1 − ηop )







+

z (1 + β ) 2 

 z ηop −  2

+

from cash for c1 < z

from cash for c1 ≥ z

ηop Case ii. With 1− ηop < z, the life-time return from investing in

an asset and holding onto cash until date 1 are expressed as



u L = c2 . This alteration has two implications. First, the utility of early consumers exhibits risk aversion. Second, consumption cannot be perfectly transferred to the late consumers without incurring a welfare loss.15 I assume β > qR − 1, such that a social planner would keep η = 2z < 12 units of liquidity until date 1 to compensate early consumers’ consumption needs and invest 1 − η = 1 − 2z units of cash in asset origination. Doing so generates a welfare of



implying that EA = EC and hence the equilibrium allocation of cash with opaque assets continues to equal to ηop under the utility specification (23), which satisfies ηop < 12 . Under case i, the aggregate welfare of the society is then given by

c2 from asset returns

For a late consumer (L), I continue to assume

For β = 0 the framework collapses to the benchmark specification.

2

from cash for c1 < z



In this section I investigate the equilibrium consequences of an alternative utility function, which relaxes the perfectly transferable linear utility specification I concentrated on in previous sections. Following Dang et al. (2017), let us index an early consumer with E and suppose that an early consumer’s utility is given by

15



1 − η∗tr 2

Next I investigate the efficiency of the equilibrium under opacity. Recalling that the equilibrium cash allocation implied by markets in Section 2.3 was denoted with ηop , two cases arise:

EA =

.

1 1 (2 − z )qR + z(1 + β ), 2 2



+q

op

Liquidity Supply

uE = c1 + β min{z, c1 }, z ∈ (0, 1 ).



c2 from asset returns



η∗tr

η Case i. With 1− ηop > z, the respective life-time returns from investing in an asset and holding cash are expressed as

Appendix B. Non-transferrable preferences and risk-aversion

Vsoc =



+ z (1 + β )

from cash for c1 ≥ z

which solving for the unit price of an asset yields

1−η

qR(1 − η∗tr )

tr Vmarket =

Re-injected Funds

1 op 1 1 p (1 − η )(2 − q ) = η + pop (1 − η )(1 − q ), 2 2 2





pop =

1 [q(1 + β )z + 1 − qz + Rq], 2 1 EC = [(1 + β )z + 1 − z + Rq], 2

EA =

+

The aggregate supply of assets comes from impatient investors (supplying 21 (1 − η ) in total) and patient investors with low quality assets (supplying 12 (1 − η )(1 − q )). Combined this gives an asset supply of 12 (1 − η )(2 − q ). Hence, the market clearing is stated as

η

where Vsoc is the expected utility implied by the social planner’s allocation with the preference structure at (23). I start by showing that the equilibrium allocation remains inefficient under transparency. With transparent assets and standard risk neutral preferences, in equilibrium the aggregate units of cash carried over to date 1 equaled to ηtr = 12 . I first examine whether ηtr = 12 still constitutes an equilibrium. Plugging ηtr = 12 into the expected return functions from originating assets and holding cash, I get

1 op p (1 − η )(1 − q ) . 2



From Date-1

11

EA =



1 ηop (1 + β ) 2 1 − ηop





+ qR



3 − 2 q 

,

2−q

1 1 − ηop EC = (1 + β )z + 1 − z + qR 2 ηop

  1 2−q

,

yielding that EA < EC . Therefore, if z is sufficiently large, then under the alternative utility specification the cash allocation induced op by markets with opaque assets is pinned down by an η∗ , where op 1 op η < η∗ < 2 . Under case ii, the aggregate welfare is either given by op Vmarket =

qR(1 − η∗op ) +







c2 from asset returns

z (1 + β ) 2 

from cash for c1 < z

+

 z η∗op − , (24)  2

from cash for c1 ≥ z

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF 12

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

or by op Vmarket

qR(1 − η ) + η (1 + β ),

=





op ∗





c2 from asset returns

op ∗



(25)



c1 from cash

depending on whether in equilibrium

η∗op

op

1−η∗

> z (the case of

op Eq. (24)) or η∗ op < z (the case of Eq. (25)). 1−η∗

Proposition 7. The relative aggregate welfare between opacity and transparency have the following properties: i. If

ηop

op

tr > z, Vmarket > Vmarket .

1−ηop

ii. If

ηop 1−ηop

iii. If

ηop 1−ηop

< z and < z and

where zˆ ∈ (0, 1 ).

η∗op

op

1−η∗

op

tr > z, Vmarket > Vmarket .

η∗op op 1−η∗

op

tr < z, Vmarket > Vmarket as long as z ≤ zˆ,

Proof. (i and ii). The result follows from the argument that the quantity of cash carried over with transparent assets is inefficiently high and that only a fraction q of asset originators who turned out to be impatient could consume. Formally, η∗tr > ηop > z and q < 1 op (at i) and η∗tr > η∗ > z (at ii). (iii). First note that the quantity of cash carried over with transparent assets (η∗tr ) is larger than the cash carried over with opaque op op tr assets (η∗ ). Vmarket > Vmarket if



qR(η − η ) + η (1 + β ) − η > zβ tr ∗

op ∗

op ∗

tr ∗



1 − η∗tr +q 2 2

η∗tr



,

gate uncertainty on date 1 allows for the cases where cash-in-themarket pricing do not apply and where assets are priced according to fundamentals. At first I assume that γ is sufficiently close to zero, such that in some aggregate states of the economy cash-inthe-market pricing prevails in equilibrium. The adjusted timing of events is presented in Timeline 3. The modification in the setup does not alter the social planner’s solution, i.e. it is still optimal to not carry any liquidity over to date 1 from an aggregate welfare point of view. In what follows, I compare the equilibrium efficiency with transparency and opacity. First, I consider transparency. Due to the aggregate uncertainty associated with γ , there is now a distribution of prices in the date1 spot market, p(γ ). Note that for γ¯ large enough, there could be realizations of γ such that p(γ ) = R. This is because for sufficiently large amounts of date-1 cash, there will no longer be cash-in-the market pricing and prices will equal fundamentals. The life-time expected returns from originating assets and carrying over cash can then be expressed as follows

  γ  γ 1 γ f (γ ) tr = q p (γ ) f (γ )dγ + qR + R dγ , tr 2 γ γ p (γ )    γ 1 (1 + γ ) f (γ ) ECtr = 1+R dγ . 2 ptr (γ ) γ EAtr

This proposition shows that the qualitative result I obtained in Section 2 regarding the macro effects of adverse selection continue to hold as long as the welfare consequences of foregoing “liquidity demand” of the early consumers are not very high. This is very intuitive, because when the preference for early consumption becomes very high, there is no longer an under-investment problem in the economy. Appendix C. Relaxing the aggregate liquidity constraint In order to relax the special-case liquidity-constrained spot market that I captured in Section 2, I consider the following alteration. On date 1 patient agents receive γ˜ units of cash, which they can utilize to make asset purchases. I assume that γ˜ is a random variable drawn from a well-behaved distribution function with support [γ , γ¯ ] on density f(γ ).16 The introduction of aggre-

EAop =

1 2 +



γ γ

 γ γ

pop (γ ) f (γ )dγ + qR

(1 − q ) pop (γ )

q R f ( γ )d γ 2 − q pop (γ )

 γ f (γ ) +R dγ , 2 − q γ pop (γ )  γ

3 − 2q  1 = pop (γ ) f (γ )dγ + qR

q 

2

γ

γ

q 



γ

For γ = γ¯ = 0 the framework collapses to the benchmark specification.

2−q

γ f (γ ) dγ , pop (γ ) 

q   γ (1 + γ ) f (γ )  1 ECop = 1+R dγ . 2 2−q γ pop (γ ) +R

2−q

γ

(28) (29)

Proposition 8. Under a generalized cash-in-the-market pricing framework, a market-equilibrium with asset opacity has higher aggregate welfare compared to a market-equilibrium with asset transop tr parency, i.e. Vmarket > Vmarket . op

16

(27)

Next I turn to the case of opaque assets, where asset quality news arrive privately and adverse selection prevails in equilibrium. The respective expected returns can now be expressed as follows

op

tr where - since for z = 0 Vmarket > Vmarket - it follows that there exists op tr a zˆ ∈ (0, 1 ) such that for all z ≤ zˆ we have Vmarket > Vmarket . For z ∈ op op tr tr [zˆ, 1 ), whether Vmarket > Vmarket or Vmarket < Vmarket depends on the parameter constellations. 

(26)

op

Proof. In equilibrium EA = EC and EAtr = ECtr must hold as in the benchmark analysis of Section 2, otherwise cash or assets are not sourced, neither of which can constitute an equilibrium. In order op to prove the claim in the proposition I need to show that EC > ECtr op op tr tr (or EA > EA ). Suppose that Vmarket = Vmarket , which implies that

ECop = ECtr .

(30)

Using (27) and (29), I can note that the equality (30) would hold if and only if

 γ

1

γ

pop (γ ) =

Timeline 3. Generalized Cash-in-the-Market Pricing.

2−q q

f ( γ )d γ +



γ γ

 γ

1 ptr (γ )

γ

γ

pop (γ )

f ( γ )d γ +

f ( γ )d γ  γ γ

 γ f ( γ ) d γ . ptr (γ )

(31)

Next I need to check, when the asset price condition (31) holds, op whether the condition EA = EAtr could be satisfied. Using (26) and

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx op

(28), at asset prices given by (31), the condition EA = EAtr could hold only if

qR

3 − 2q  2−q

 γ

− qR = q

γ

−

 γ q ptr (γ ) f (γ )dγ , 2−q γ

⇒ qR

1 − q 2−q

γ



Appendix D. Productivity-reducing opacity



= γ pop (γ ) f (γ )dγ

=q

1 − q 2−q

γ γ

ptr (γ ) f (γ )dγ .

(32)

As long as γ is not too high (such that cash-in-the-market pricγ ing binds in some states), E[ ptr ] = γ ptr (γ ) f (γ )dγ < R, implying op

that (32) can never be satisfied. This means that EA > EAtr given op op the price condition (31) and also that EA > EC . Therefore, in an equilibrium with asset opacity

 γ

1 pop

γ

f ( γ )d γ +

(γ ) 

2−q > q

γ γ

 γ

1 ptr (γ )

γ

γ

pop (γ )

f ( γ )d γ +

f ( γ )d γ  γ γ

γ f ( γ )d γ ptr (γ )

 (33)

should hold in order to ensure the no-arbitrage between origop op inating assets and holding cash (EA = EC ). Then, the inequalop op op tr ity (33) yields that EC > EC (and also EA > EAtr ) and Vmarket > tr Vmarket .  The finding at Proposition 8 shows that relaxing the degree of cash-in-the-market pricing in the economy does not alter the key qualitative result that the economy with opaque assets is more efficient in its date-0 portfolio allocation compared to an economy with transparent assets. Building upon the insight from Proposition 7, I also derive the following Corollary. Corollary 2. If γ is sufficiently large, such that the spot market on date-1 can fully absorb (and ex-post insure) liquidity risk, market allocations attain the socially optimal aggregate output derived at (6) and the nature of arriving information on date-1 has no implication for macroeconomic performance. Proof. Suppose that γ is large enough such that assets are always priced based on fundamentals. Let us first consider the case of transparent assets in this environment, where only high quality assets are sold. The price of an asset satisfies ptr = R because of the large liquidity supply on date-1. This implies that only impatient high quality asset holders would sell with the following expected return to asset origination:

EA =

and EA > EC and therefore all cash is invested on date-0, resulting op tr in Vmarket = qR = Vmarket .  Corollary 2 formalizes that adverse selection is likely to be welfare-improving when assets are exchanged in liquidity constrained markets.

ptr (γ ) f (γ )dγ



 1  tr qp + qR = qR. 2

It could be the case that opaque financial assets are inherently of lesser quality. Such quality differentiation would imply a cost of holding opaque assets, not only for originators but also for purchasers and potentially undo the macro-development effects of adverse selection.17 In order to dig deeper into the implications of this alternative cost structure, let us assume that opacity investment on date 0 results in a “productivity loss” for the originated asset. Specifically, as in Section 3, on date 0 an asset originator decides whether as of date 1 he wants to have a fully transparent asset with certainty or an opaque asset with probability α . Different from Section 3, there is no cost of originating an opaque asset on date 0 (i.e. cO = 0); however, if an asset becomes opaque on date 1, it loses fractions of its date 2 (high state) return: I assume that the date 2 high-state return of a transparent asset is RH whereas H that of an opaque asset is RL with RRL ≡ μ > 1 and qRL > 1. Different from the case of costly origination with an extensive margin lump-sum cost, this alternative structure taxes high-paying assets at the intensive margin. Hence, the current cost structure has the potential of being distortionary. The rest of the model structure remains. The recursive solution is as follows. Date-1 Market Clearing. In this case, the equilibrium noarbitrage between pop and ptr on date 1 is given by the following:

pop q RL q 1 = = . tr p 2 − q RH 2−qμ



1 R 1 + tr 2 p



= 1,

EA > EC and therefore all cash is invested on date-0, yielding tr Vmarket = qR. In the case of opaque assets, both high quality and low quality q assets are sold; and, therefore we have pop = 2−q R. This implies that all impatient asset holders and patient holders of low quality assets would sell in the market, yielding

1 op [ p + qR + (1 − q ) pop ] = qR, 2   q R 1 2−q EC = 1 + op = 1, 2 p

EA =

(34)

I utilize the date 1 no-arbitrage condition (34) and express the market clearing on date 1 as

pop (1 − η )

( 2 − q )μ ( 1 − α ) 2

+ pop (1 − η )

( 2 − q )α

1 ( 1 − q )α = η + pop (1 − η ) , 2 2

2 (35)

which solves for pop

pop =

η . ( 1 − η )[ ( 2 − q ) μ ( 1 − α ) + α ]

(36)

Date-0 Portfolio Choice. Moving on to the individual agents’ decision to invest in opaqueness by the time the portfolio choices are made, I can express the value function of originating an asset (with an embedded opacity choice) as

EA =

Since expected return to hold onto cash is given by

EC =

13

α 2

qRL

 3 − 2q 2−q





− μ − pop (μ(2 − q ) − 1 ) +

 1  tr qp + qRH , 2 (37)

where the value of carrying cash forward is



EC =

1 RL q 1 + op 2 p 2−q





=



1 RH 1 + tr . 2 p

(38)

I again look for the parameter configurations of the model that would keep the individual asset originator indifferent between investing in opacity and keeping his asset fully transparent and then 17 A highlighted previously for financial investors originating assets in novel & innovative industries is likely to be costly - due frictions in decentralized market exchange. Such costs are expected to be borne by investors to some extent. Productivity-reducing opacity allows to capture the implications of such costs when they are partly also borne by purchasers of assets.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF 14

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

evaluate the macro benefits of adverse selection in the spot market against its costs for such configurations of the parameter space. An individual agent is indifferent towards opacity as of date 0 if

qRL

 3 − 2q 2−q



− μ = pop (μ(2 − q ) − 1 ).

(39)

Since the right hand side is always positive, the condition q (39) requires that μ ≤ 32−2 −q - an assumption that I will keep for the rest of the analysis. I again assume that when an asset originator is indifferent, on date 0 he prefers to invest in opacity. Then, using (39) in date 0 no-arbitrage condition (EA = EC ) I derive the equilibrium price of opaque assets

pop =

1 1 , 2−qμ

(40)

and also the equilibrium price of transparent assets

1 . q

ptr =

(41)

Using (40) in (39) I get

qRL =

μ (2 − q ) − 1 , μ[ ( 3 − 2 q ) − μ ( 2 − q ) ]

(42)

which is the full parameter characterization of an individual agent’s opacity indifference condition that I sought to express. This yields the following proposition: Proposition 9. Agents invest in opacity on date 0 if and only if qRL > μ(2−q )−1 μ[(3−2q )−μ(2−q )] . Plugging (40) in (36) gives

η (α > 0 ) =

(2 − q )μ − α [(2 − q )μ − 1] . 2μ(2 − q ) − α [(2 − q )μ − 1]

(43)

Similar to the equilibrium opacity analysis from Section 3, we know that when there is no investment in opacity (implying α = 0), then the equilibrium quantity of cash carried forward from date-0 would be η (α = 0 ) = 12 . I can note that

η (α > 0 ) =

(2 − q )μ − α [(2 − q )μ − 1] 1 < η (α = 0 ) = 2μ(2 − q ) − α [(2 − q )μ − 1] 2

for all parameter values, i.e. incurring the expected productivity cost associated with investing in opacity on date 0 and generating opaque assets with probability α raises the aggregate quantity of assets originated on date 0. Given condition (42), the welfare in the economy when all agents invest in opacity is derived as: op Vmarket = EAop ( pop , ptr ) =

1 [1 + qR]. 2

(44)

We can note that given equilibrium prices derived at (40) and (41), the unilateral deviation of an asset originator to full transparency gives him

EAtr =

1 1 [qptr + qR] = [1 + qR], 2 2

(45)

confirming the state of indifference towards opacity. In order to evaluate an aggregate transparency regulator’s willingness to allow for opacity (while letting the date 1 asset markets clear), I first observe that the aggregate benefit from opacity results from the expansion of the asset origination on date-0:

Aggregate Bene f it Opacity = (η (α = 0 ) − η (α > 0 ) )[α (qRL − 1 ) + (1 − α )(qRH − 1 )]



=



1 (2 − q )μ − α [(2 − q )μ − 1] − [α (qRL − 1 ) 2 2μ(2 − q ) − α [(2 − q )μ − 1]

+ (1 − α )(qRH − 1 )]

α [(2 − q )μ − 1] [qRL [α + (1 − α )μ] − 1]. 2[2μ(2 − q ) − α [(2 − q )μ − 1]] (46)

=

The first multiplicative term on the right-hand-side measures the expansion in the asset base whereas the second multiplicative term is the net consumption benefit for the society for having originated one unit of asset - which takes into account that α fraction of the additional assets will be of “low” quality and 1 − α fraction will be of “high” quality. The aggregate cost on the other hand is associated with the contraction in asset productivity, which I can express as

Aggregate Cost Opacity = (1 − η (α = 0 ) )α qRL (μ − 1 ) =

α 2

qRL (μ − 1 ),

(47)

where the first multiplicative term (α /2) on the right-hand-side is the total quantity of assets that suffer a productivity loss with opaque asset investment - compared to the case of full transparency - whereas the remaining multiplicative term is productivity loss at the asset level. Finally, comparing (46) against (47), an aggregate transparency regulator would be indifferent towards opacity if the following condition holds:

( 2 − q )μ − 1 [qRL [α + (1 − α )μ] − 1] 2μ(2 − q ) − α [(2 − q )μ − 1] ?



= qRL (μ − 1 ).

(48)

To understand the net social returns to opacity relative to its net private returns, I will investigate whether condition (48) would hold with equality given the level of qRL that keeps an individual agent indifferent (derived at (42)). Therefore, plugging the expression for qRL from (42) to both sides of (48), I can evaluate the net aggregate benefits of opacity given parameter configurations of the model that would keep an individual asset originator indifferent towards opacity. Specifically, using (42) in (48), (48) reduces to the following:

[α + (1 − α )μ][(2 − q )μ − 1] − μ[(3 − 2q ) − μ(2 − q )] 2μ(2 − q ) − α [(2 − q )μ − 1] ?



=

μ − 1.

(49)

Simplifying the left-hand-side (LHS) of (49) further shows that the LHS expression in fact equals to μ − 1. Hence, aggregate benefits of opacity equal to the aggregate costs of opacity given the parameter conditions (42) that would keep an individual agent indifferent towards opacity. Proposition 10. The private and social returns to opacity coincide with each other. This result implies that incorporating a productivity loss associated with originating (and holding onto) an opaque asset does not alter the key equilibrium efficiency implication of decentralized opacity decisions of financial market participants and the prevailing adverse selection that I derived in Section 3. Appendix E. Information acquisition in equilibrium In this section I introduce the possibility of information acquisition for cash holders. This analysis allows to explore when the funding-side could uncover asset-quality news through information acquisition, whether such equilibrium behavior would generate room for transparency regulation or for regulation of information acquisition. I thus extend the benchmark model of Section 2 in

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

JID: JBF

ARTICLE IN PRESS

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

such a way that a liquidity-provider could receive exclusive information on future asset returns on date-1 through a technology platform, entry to which is stochastic. I maintain the rest of the model structure that yield properties 1–5. The information acquisition works the following way. Each cash-holder from date-0 ends up on an “informed platform” on c date-1 with probability λ, provided he has incurred 2I units of non-monetary cost (of consumption equivalent) on date-0 to have stochastic accessibility to the informed platform. Every cash-holder on the platform gets to observe the future returns to every single asset that is brought to the platform by originators. The information remains in the platform and cannot be communicated to anyone outside. The would-be-asset-sellers do not pay any fee when entering the platform to supply assets. I have three remarks to highlight before I continue with the analysis. 1. I assume λ < 1, such that if all cash-holders invest in information acquisition on date-0 we do not have a trivial replication of the economy with transparent assets. 2. I concentrate on the case of cI small enough and as a result all cash-holders do invest in information acquisition on date-0, such that we do not have a trivial replication of the economy with private information and opaque assets. As a matter of fact most of my analysis will concentrate on the special case of negligibly small cI . 3. I also note that if the asset-side of the financial market fully consists of transparent assets then cash-holders will have no incentives to invest in information acquisition (and the econtr omy’s equilibrium welfare will equal to Vmarket that I derived in Section 2). Therefore, I assume that all assets originated on date-0 are of opaque nature with private quality news - as described in Section 2. Since the core structure of the framework remains from the benchmark, patient high quality asset holders keep their assets till date-2, while impatient high quality asset holders and all low quality asset holders desire to sell on date-1. Furthermore, it is clear based on the revised structure that only the impatient high quality asset holders will enter the informed platform on date 1 to sell, while both impatient high quality and low quality asset holders might transact with uninformed cash holders on date 1, who do not get the chance of informed trading. Timeline 4 presents the sequence of events in this extended framework. Let us adopt θ to denote the equilibrium fraction of impatient high-quality asset holders who enter the informed platform on date 1. Characterization of the equilibrium with endogenous information acquisition requires solving for θ . There are two relevant cases: (1) 0 < θ < 1 and (2) θ = 1. I present them next under the

15

assumption that cI is small enough such that all cash-holders on date-0 invest in information acquisition. Case 1: 0 < θ < 1 In this particular equilibrium fractions of the high-quality assetsales take place in the uninformed market. This equilibrium can prevail if and only if high quality sellers are indifferent between selling in the informed platform and selling in the uninformed financial market, which implies pin f ormed = puni f ormed = p. If pinformed > puniformed , then high-quality sellers transition to the platform until prices are equalized. Similarly, if pinformed < puniformed , then high-quality sellers would transition to the uninformed market until prices are equalized. Date-1 Market Clearing. We have two market clearing conditions on date-1, one for the informed platform and another one for the uninformed financial market - with the same asset price clearing both markets. Clearing in the informed platform gives

1 1 λη = pq(1 − η )θ , 2 2 where as before η is the aggregate quantity of cash carried-forward on date-0, yielding:

η

p=

λ1 . θ

1−η q

(50)

In the uninformed financial market on the other hand

Supply =

1 1 η (1 − λ ) + p(1 − q )(1 − η ), 2 2

and

Demand =

1 1 pq(1 − η )(1 − θ ) + p(1 − q )(1 − η ) 2 2 1 + p(1 − q )(1 − η ), 2

which yield the following market clearing

p(1 − η )[1 − q + q(1 − θ )] = η (1 − λ ), with

p=



η  1−λ . 1−η (1 − q ) + q (1 − θ )

(51)

Then, (50) and (51) together imply

λ

1−λ

=

qθ , (1 − q ) + q (1 − θ )

solving for the equilibrium fraction of high-quality asset sellers who select into the informed trading platform, θ , as

θ=

λ q

.

(52)

Timeline 4. Information acquisition.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF 16

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

Based on (52) we can infer that θ < 1 if λ < q. Otherwise, there are enough informed asset purchasers on date-1, who fully absorb all of the high quality asset sellers in the informed platform. For the corner case of θ = 1, which is sustained if λ ≥ q, I will provide a separate analysis below. I note that the other corner at θ = 0 is ruled out by λ > 0. Then, (52) in (50) yields

p=

η

1−η

,

(53)

which is identical to the cash-in-the-market equilibrium asset price function that I derived for the case of opaque financial markets in Section 2. Equilibrium η in this current case will be modified though, since it will be a function of λ and cI - as to be delineated below. Date-0 Portfolio Choice. The proportion of high quality assets relative to the aggregate pool of assets in uninformed financial market of date-1 gives 1 q 2

(1 − θ )

1 q 2 + 12

(1 − θ ) q (1 − θ ) = , 1 q ( 1 − θ ) + 2 (1 − q ) (1 − q ) + 2 (1 − q )

using which and (53) I can express the expected return to originate assets as of date-0:



1 R q (1 − θ ) η EA = + qR + (1 − q ) p 2 1−η p q (1 − θ ) + 2 (1 − q )





1 1 − θ + 2 (1 − q ) η = + qR 2 1−η q (1 − θ ) + 2 (1 − q )





.

(54)

I first note that for λ = 0 (and hence for θ = 0) (54) reduces to the expected return to asset origination expression obtained in Section 2 for the case of opaque financial markets, namely

EA =



1 η 3 − 2q + qR 2 1−η 2−q





1 1−θ 1 1 1 + (1 − λ )qR + λR − cI 2 q (1 − θ ) + 2 (1 − q ) p p



=

1 1 + qR 2

1 − η  η







1−η

= 1 + qR

1 − θ + 2 (1 − q ) q (1 − θ ) + 2 (1 − q )

1 − η  η



 (1 − λ )(1 − θ ) +θ , q (1 − θ ) + 2 (1 − q )

differentiating which with respect to λ shows that ∂ η/∂ λ > 0. op Since ∂ Vmarket /∂ η < 0, I conclude that more informed trading under the case of θ ∈ (0, 1), i.e. higher λ, is destructionary for the aggregate welfare. Therefore, I continue to obtain the relevance of limited financial transparency on macroeconomic development in liquidity constrained asset markets. Interestingly though I also find out that when the funding-side of the market could acquire information, restricting accessibility to the information platform could improve the allocation efficiency and macroeconomic outcomes. Case 2: θ = 1 In this case λ ≥ q and all high-quality asset sellers on date-1 are absorbed by the informed platform and anticipating this, uninformed cash holders do not engage with any exchange on date-1, because they expect that only low quality asset holders will visit the uninformed financial market. Date-1 Market Clearing. Date-1 market clearing on the informed platform is given by

1 1 λη = pq(1 − η ), 2 2 where η again denotes the equilibrium quantity of aggregate cash carried to date-1, solving for the equilibrium price of assets in the informed platform as

η

λ

1−η q





(1 − λ )(1 − θ ) + θ − cI . q (1 − θ ) + 2 (1 − q )



+ qR

.

(56)

Date-0 Portfolio Choice. Taking the future asset price figure as given the date-0 expected returns to originate assets and to hold cash are expressed as follows:



(55)

Note also that for λ = 0 (and hence for θ = 0), (55) reduces to the expected return to cash origination expression that I obtained in Section 2 for the case of opaque markets, namely

EC =

η

p=

.

Turning to the expected return to hold cash as of date-0,

EC =

Finally, since equilibrium η is implicitly solved from,



1 1−η 1 1 + qR . 2 η 2−q

I move on with the characterization of the equilibrium cash holdings in the economy, which - as before - is given by η that solves EA = EC . In this respect, suppose that η = 12 , which is the equilibrium cash holding in an economy with transparent financial markets. Furthermore without loss of generality assume that cI = 0, and then for η = 12 , EA > EC if

1 − θ + 2(1 − q ) > (1 − λ )(1 − θ ) + θ q(1 − θ ) + 2(1 − q )θ , or using (52), if

q > λ, which always holds by construction that gives rise to θ < 1. Therefore, EA > EC given η = 12 , implying that as long as λ < q (and cI negligibly small) the equilibrium cash holding with opacity remains to satisfy η < 12 in order to maintain Property 4 (EA = EC ) and as a result aggregate welfare with opaque financial markets continues to exceed the aggregate welfare under full transparency op tr (Vmarket > Vmarket ).



EA =

1 λη + qR , 2 1−η

EC =

1 1−η 2 − λ + qR 2 η



(57)





− cI .

(58)

Then, I proceed to characterize the equilibrium quantity of cash carried to date-1. For this suppose that η = 12 , which is the quantity of cash carried forward in an economy with transparent financial markets as noted before. Plugging this in (57) and (58), EA < EC if and only if

λ + qR < 2 − λ + qR − cI , which holds as long as cI < 2(1 − λ ). Suppose again that cI = 0. This will imply then that η > 12 is needed in order to maintain op tr EA = EC , such that Vmarket < Vmarket . Finally, equilibrium η is solved by

1 − η λη + qR = 2 − λ + qR , 1−η η

with the qualitative feature of ∂ η/∂ λ < 0. Since ∂ Vmarket /∂ η < 0, I conclude that more informed trading under the case of θ = 1 benefits the aggregate welfare. I summarize the key results of this section in the following proposition. op

Proposition 11. Assuming that all assets originated on date-0 are of opaque nature and also that the cost of accessing the informed platform for the funding-side is zero (or negligibly small),

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

ARTICLE IN PRESS

JID: JBF

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

17

op ∂ Vmarket < 0, ∂λ op ∂ Vmarket op tr ≥ q, Vmarket < Vmarket and > 0. ∂λ

tr i. for λ < q, Vmarket > Vmarket and op

ii. while for λ

Transparency Regulation with Information Acquisition. Information acquisition undermines the benefits from opacity, namely the benefits of private information and the resulting adverse selection. Formally, the fraction of cash-holders eligible to enter the information platform (λ) that would maximize the aggregate equilibrium welfare equals zero, which implies that the optimal regulation would let the asset-side of the market remain opaque and restrict access to information at the funding-side. However, if access to information cannot be constrained by regulation, but transparency of assets can, then the optimal design of regulatory framework might ask for financial transparency. In order to observe this, first we can note that for λ < q as λ increases, the benefits of opacity - relative to the case of transparency - start to decline, although opaque markets continue to dominate transparent markets in terms aggregate welfare as long as λ is small enough. The intuition for this property is related to the reduction in the extent of asset swap opportunities generated by opacity. For λ large enough the benefits of an economy with transparent asset markets exceed that of with opaque markets. In particular, when λ ≥ q, then all high quality assets are absorbed in the informed platform, such that there is no room for the assetswap to improve welfare. In this case 1 − λ fraction of the liquidity providers on date-1 do not get to finance asset purchases. This means that the equilibrium price of assets are too low - without the beneficial asset swap opportunity - such that the returns to originate an asset turns out to be lower with opacity compared to the case of transparency. In this respect, interestingly, the need for transparency intervention arises when private liquidity providers’ information technology efficiency (captured by λ) takes interim values. This latter result relates to the theoretical findings of Andolfatto et al. (2014): Andolfatto et al. (2014) utilize a search-theoretic framework of money to show that aggregate uncertainty (nondisclosure) about future payoffs of a commonly held asset could help with risk sharing when information acquisition does not come at too low of a cost. I also uncover a detrimental effect of information, where as an important difference from Andolfatto et al. (2014) I work with a model of asset-specific information and adverse selection. I show that a novel benefit of adverse selection prevails through an asset swap possibility when asset sales are priced based on the amount of liquidity available in the market. Similar to the finding of Andolfatto et al. (2014), in my framework endogenous information acquisition could dampen the benefits of opacity, though in the current model information acquisition does not necessarily make opacity undesirable even if the cost of learning future asset payoffs is low. Appendix F. Asset reversibility I now allow for assets to be discontinued on date 1. In particular, I assume that the scrap value of an asset on date 1 is x with x ∈ (0, 1) in case it is of high quality, and zero otherwise. I consider first the case of transparency. The social planner can now decide - next to the amount of investment on date 0 - also whether assets should be scrapped on date 1. On date 1, the continuation value of a high quality asset is R, while its scrapping value is only x. The social planner will hence never scrap, regardless of the nature of arriving information on date 1. The social planner thus allocates exactly the same way as in the benchmark model: it will invest all cash in assets and welfare continues to be expressed as

Vsoc = qR.

(59)

Timeline 5. Asset scrapping.

I now turn to the analysis of the equilibrium. The timing of the various decisions for agents is presented in Timeline 5. I show that the market equilibrium in the benchmark model with transparent assets remains an equilibrium with scrapping. The asset price in the baseline model was ptr = 1q , which is strictly larger than x. Hence it is never optimal to scrap assets, since agents will always prefer to sell instead of scrapping. Thus, I obtain the same outcome as in the baseline model: ηtr = 12 and welfare is given by

Vtr =

1 [qR + 1]. 2

Therefore, Vtr < Vsoc and hence the economy with transparent assets continues to produce a socially inefficient aggregate allocation. Next I turn to the market equilibrium with opaque assets and analyze the case with adverse selection in the spot market. There are now two cases to consider, depending on the date-1 price pop in an equilibrium without scrapping: (i) pop < x and (ii) pop > x. Consider first pop < x. In this case, impatient agents who hold high quality assets would prefer scrapping assets instead of selling them. This implies that only low quality assets would remain in a date 1 market. Trade in the asset market hence breaks down due to the standard Akerlof-style adverse selection. Since asset transparency generates gains from trade compared to the autarkic state, for x large enough asset transparency turns out to be socially desirable. Consider next pop > x. The asset price is now higher than the scrapping value, hence scrapping will not be undertaken by high quality asset holders. The equilibrium in this case will thus remain as in the benchmark analysis without scrapping. I can therefore conclude with the following result. Proposition 12. Asset opacity raises welfare as long as the scrap value of high-quality assets, x, is sufficiently low. References Allen, F., Gale, D., 1994. Liquidity preference, market participation and asset price volatility. Am. Econ. Rev. 84 (4). 933–55 Andolfatto, D., Berentsen, A., Waller, C., 2014. Optimal disclosure policy and undue diligence. J. Econ. Theory 149, 128–152. Bencivenga, V., Smith, B., Starr, R.M., 1995. Transactions costs, technological change and endogenous growth. J. Econ. Theory 67 (1), 153–177. Bencivenga, V., Smith, B., Starr, R.M., 1996. Liquidity of secondary markets: allocative efficiency and the maturity composition of the capital stock. Econ. Theory 7 (1), 19–50. Bencivenga, V., Smith, B., Starr, R.M., 20 0 0. Secondary capital markets, long-run growth, and the term structure of asset yields. Int. Econ. Rev. (Philadelphia) 41 (3), 769–800. Castro, R., Clementi, G.L., MacDonald, G., 2004. Investor protection, optimal incentives, and economic growth. Q. J. Econ. 119 (3), 1131–1175. Chinn, M.D., Ito, H., 2006. What matters for financial development? capital controls, institutions, and interactions. J. Dev. Econ. 81 (1), 163–192. Dang, T.V., Gorton, G., Holmström, B., Ordonez, G., 2017. Banks as secret keepers. Am. Econ. Rev. 107 (4), 1005–1029. Diamond, D.W., Dybvig, P.H., 1983. Bank runs, deposit insurance, and liquidity. Journal of Political Economy 91 (3), 401–419.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721

JID: JBF 18

ARTICLE IN PRESS

[m5G;December 20, 2019;12:40]

B.R. Uras / Journal of Banking and Finance xxx (xxxx) xxx

Feldman, R.A., Kumar, M.S., 1995. Emerging equity markets: growth, benefits, and policy concerns. World Bank Res. Obs. 10 (2), 181–200. Hirshleifer, J., 1971. The private and social value of information and the reward to inventive activity. Am. Econ. Rev. 61 (4), 561–574. Kaminsky, G.L., Schmukler, S.L., 2008. Short-run pain, long-run gain: financial liberalization and stock market cycles. Rev. Financ 12 (2), 253–292. Monnet, C., Quintin, E., 2013. Rational opacity. Working Paper. Pagano, M., Jappelli, T., 1994. Saving, growth and liquidity constraints. Q. J. Econ. 109 (1), 83–109. Philippon, T., 2010. Financiers versus engineers: should the financial sector be taxed or subsidized? Am. Econ. J. 2 (3). 158–82 Philippon, T., 2015. Has the US finance industry become less efficient? on the theory and measurement of financial intermediation. Am. Econ. Rev. 105 (4), 1408–1438.

Plantin, G., 2009. Learning by holding and liquidity. Rev. Econ. Stud. 76 (1), 395– 412. Porta, R.L., Lopez-de Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law and finance. J. Polit. Econ. 106 (6), 1113–1155. Rajan, R.G., Zingales, L., 2001. Financial systems, industrial structure, and growth. Oxford Rev. Econ. Policy 17 (4), 467–482. Rajan, R.G., Zingales, L., 2003. The great reversals: the politics of financial development in the twentieth century. J. Financ Econ. 69 (1), 5–50. Singh, A., 1997. Financial liberalisation, stockmarkets and economic development. Econ. J. 107 (442), 771–782. Stiglitz, J.E., Weiss, A., 1981. Credit rationing in markets with imperfect information. Am. Econ. Rev. 71 (3), 393–410. World Bank Group, 2005–2017. Doing business.

Please cite this article as: B.R. Uras, Finance and development: Rethinking the role of financial transparency, Journal of Banking and Finance, https://doi.org/10.1016/j.jbankfin.2019.105721