«The development process is in large part about addressing market failures. Whether it is the inability of productive borrowers to obtain financing or ...»
Jan. 9, 2009
Research Statement - Summary
Asim Ijaz Khwaja
The development process is in large part about addressing market failures. Whether it is the inability of
productive borrowers to obtain financing or children to receive quality education, private and public responses
can be significantly constrained by market imperfections. A central theme in my research has been to identify and
examine such limitations, particularly in the context of failures in physical and human capital accumulation – two prominent ingredients of modern growth theory. The broader effort is to build towards a more complete understanding of the market micro-structure, both in finance and education, and ultimately link the micro- findings to macro implications.
The effort to identify specific market features, mechanisms, and interactions at the level of market players (individuals, firms, and organizations) while facilitating tighter empirical/causal identification has lead me to devote considerable time and effort in the collection and compilation of the extensive micro-level data needed for such an examination. Two such examples are a financial database and an education panel data set. The financial data is a compilation of all loans, identified at the borrower and lender level, given out by 173 formal lending institutions during a seven year period to over 140,000 borrowers in Pakistan. The education data includes conducting a census of all 85,000 households in over a hundred villages in Pakistan, four (panel) surveys in these villages of all 800 primary schools, 5,000 teachers, 2,000 households and repeat tests of over 12,000 children matched to the schools, teachers, and households.
Most of this empirical work so far has focused on Pakistan. This has been a conscious choice because of my familiarity with the environment and overall market context, and the large fixed costs of gathering such detailed data sets. Nevertheless, the insights gained from my research are intended to both highlight issues faced by other emerging economies and also provide lessons that even developed markets may benefit from, a point that is increasingly salient as we begin to deal with the current global financial and economic crisis.
A. Finance and Firms Financial markets intermediate the matching of money to ideas. To the extent that there are imperfections in such markets, the effectiveness of this role will be diminished, with potentially large and real consequences for the economy. A significant part of my research in this area, developed jointly with Atif Mian (Chicago, GSB), examines this question from a series of related perspectives, from directly identifying market manipulation and imperfections to examining market micro-structure and alternate financial instruments.
I identify the presence of market manipulation in equity and debt markets in two papers. “Unchecked Intermediaries: Price Manipulation in an Emerging Financial Market” (Journal of Financial Economics, 05, with A.
Mian) highlights the poor governance of market intermediaries in equity markets. Using a unique data set of all daily trades for each broker trading during a two and a half year period on the main stock exchange in Pakistan, we identify price manipulation through “pump and dump” schemes. We estimate significant rents accruing to manipulating brokers – 50 to 90 percentage points higher returns - with conservative estimates suggesting a $100 million annual transfer of wealth from outside investors to manipulating brokers, 10% of market capitalization.
In “Do Lenders Favor Politically Connected Firms? Rent Provision in an Emerging Financial Market” (Quarterly Journal of Economics, 05, with A. Mian) we identify the extent of politically-based rent provision in debt markets.
We find that politically connected firms, those with a politician on their board, get differential treatment from public banks – they borrow one and a half times as much and have 50% higher default rates. Conservative estimates for the economy-wide loss due to such inefficient lending are upto 1.9% of GDP a year. Both these papers not only highlight the presence of significant manipulation in financial markets but highlight that even though the overall economic costs are substantial, the large individual rents accrued – especially by those with influence - may explain why governance reforms, even if straightforward, may be hard to implement.
Aside from explicit market manipulation, economies may face significant losses due to systematic constraints and agency problems endemic in financial markets tasked with intermediating capital between relatively uninformed investors and privately informed borrowers. Of particular interest is understanding how economies respond to liquidity shocks – both negative and positive. In “Tracing the Impact of Bank Liquidity Shocks: Evidence from an Emerging Market” (American Economic Review, 08, with A. Mian), we document the presence of a large bank lending channel in which firm lending drops by 0.37% for every percent drop in its bank’s liquidity. We isolate this “supply-side” channel by taking advantage of differential (across banks) liquidity shocks arising from unanticipated nuclear tests in Pakistan in 1998, and examining the borrowing of a given firm across multiple and differentially affected lenders. While large firms hedge this shock by borrowing elsewhere, small firms are unable to do so and therefore are more likely to enter financial distress. In a related paper, “Dollars Dollars Everywhere, Nor Any Dime to Lend” (Review of Financial Studies, R&R, with A. Mian and B. Zia), we find that positive liquidity shocks do not necessarily have a symmetric (positive) impact. While the events of 9/11 resulted in large capital inflows in Pakistan due to pulling back foreign investments and reverse capital flight, and the cost of capital dropped dramatically, there was no commensurate increase in lending. We show that this inability to extend credit is driven by “back-looking” credit constraints that lenders employ. Since firm collateral didn’t show commensurate increases, banks were unwilling to extend further credit to those borrowing up to their collateral limits. We estimate this cost the economy as much as 2.3% in forgone GDP.
In a series of related papers I have also begun to examine the organizational structure in which firms operate and how this structure may both be affected by and influence market performance. In “Identifying Business Networks in Emerging Economies,” (with A. Mian & A. Qamar) we document the presence of a robust “supernetwork” of firms linked through common directors. It is a hundred times larger than the next largest network in the economy and, while constituting around 7% of all corporate firms, absorbs two-thirds of all formal financing.
A follow-up paper, “The Value of Business Networks”, identifies the impact of a firm entering this supernetwork and shows that this leads to a 16.5% increase in a firm’s borrowing and a 9.7% lower probability of it defaulting.
Networks also play an important role in contracting and retail settings. In “Subcontractors for Tractors: Theory and Evidence on Flexible Specialization, Supplier Selection, and Contracting,” (Journal of Development Economics 05, with T. Andrabi & M. Ghattak) we utilize contract-level data between a large assembler and its network of suppliers to show that the network setting not only allows the assembler to manage aggregate demand shocks but allows both low and high quality suppliers to co-exist. Low quality suppliers exist because they are more likely to undertake dedicated investments and are willing to receive lower prices and less stable orders and ultimately act as “capacity” buffers for the assembler in the absence of requisite insurance markets. Networks can also be an important organizational form in retail settings. In an ongoing project, “Divided We Fall: Understanding Retail Clusters” (with J. Das and A. Mian), we examine an organization structure prevalent in developing countries – retail clusters – defined as a large network of shops selling the same product at a common location. We analyze a unique primary survey of 700 retail shops in 15 retail categories in a large city (Lahore, Pakistan) and show that the apart from inventory sharing considerations, clusters also arise due to an inability of retailers to smooth input price shocks produces which in turn produces strong “price comparison” incentives on the part of buyers and hence, a preference to shop at clusters.
More recently, as a result of my own work and recent events, I have begun to examine alternate financial mechanisms. In “Can Individual Lenders Infer Borrower Creditworthiness? Evidence from a Peer-to-Peer Credit Market” (with R. Iyer, E. Luttmer & K. Shue) we show that, despite not being financial experts, individual lenders can collectively infer up to a third of the relevant information content regarding a borrower’s creditworthiness, as proxied by their credit score. They do so by utilizing both soft and hard information available to them through a borrower’s postings on the p2p lending website. This degree of efficiency suggests that such alternate lending markets may even provide advantages over traditional lenders by extracting information from non-traditional “softer” sources. I have recently also obtained a grant from Google.org to set up a multi-year “Entrepreneurial Finance Lab” (www.cid.harvard.edu/efl) at the Center for International Development at Harvard. The focus of the lab is exploring alternate ways to extend capital to smaller entrepreneurs. Initially, we plan to pursue this using screening methods that draw on psychometric testing tools, and then embed these tools in micro-equity like contractual arrangements.
2B. The Educational Market
There is an increasing realization that the educational marketplace is no longer a simple choice environment where a parent decides whether to enroll a child in a single (public) school or not. Instead, it is one where parents choose among multiple schools, of differing price and quality, across multiple children. While recent empirical work has made substantial contributions through experimental interventions that estimate returns to educational inputs, we have just started examining the overall educational market place and its evolution in developing countries. My interest in education is to further such an understanding of the educational marketplace by drawing both on observational and experimental evidence.
A challenge in doing so is the substantial data requirements imposed – in order to understand the overall market, one needs matched information on all significant actors in the educational market and be able to follow this over time. For the past several years I have jointly lead, with J. Das (World Bank) and T. Andrabi (Pomona College) an exercise to gather a comprehensive four-year panel dataset that captures the educational universe in 112 villages in rural Pakistan (www.leapsproject.org). This has been an extremely intensive process and, while we are still in the preliminary stages of producing work, we anticipate the project will generate significant payoffs.
The “LEAPS” data consists of four (annual) rounds of matched surveys of all 800 primary schools, 5,000 teachers, and 2,000 households, with achievement tests we designed and implemented of the same 12,000 children followed over the four years. A baseline educational census of all 85,000 households confirms that these villages constitute closed educational markets at the primary level. By tracking schools, teachers, households, and children over this period we can also examine how these markets evolve over time and respond to various interventions. To our knowledge there are few such datasets in developing or developed countries that provide such a comprehensive picture of the educational landscape.
The initial findings are detailed in “Learning and Educational Achievements in Punjab Schools (LEAPS):
Insights to inform the education policy debate” (Oxford University Press, forthcoming, with T. Andrabi, J.Das, T. Vishwanath, & T. Zajonc,). We provide basic but potentially revealing insights regarding enrollment and learning outcomes, schools’ resources and competition, teachers’ competency and incentives, and parental/child knowledge and behavior. We find that while enrollment and learning levels are quite low, there are large differences – whereas enrollment is affected by both household and village level factors, learning differs much more across schools within villages. The schooling market is characterized by lots of school choice and competition, with most schools located close to one another and fees responsive to quality. Private school teachers, while less qualified and paid substantially less than their public counterparts, add significantly more educational value, highlighting the value of incentives. Finally, we document that while parents spend considerable financial resources on their children’s education, they invest little time/personal attention and are not as well informed about school choices. An examination within the household shows that parents discriminate across children with the biggest margin being across perceived child ability rather than gender or other child attributes. Building on this work, we are now examining various specific features of the market.