Tropical Lending: International Prices, Strategic Default and Credit Constraints among Coffee Washing Stations

Transcription

1 Tropical Lending: International Prices, Strategic Default and Credit Constraints among Coffee Washing Stations Arthur Blouin Toronto Rocco Macchiavello Warwick Ascea Summer School June th 2014

2 Motivation High and heterogeneous F K among microenterprises consistent with imperfect credit markets: - De Mel et al. (2008, 2009), Sri Lanka - Fafchamps et al. (2011), Ghana - Kremer et al. (2011), retailers in Kenya - Karlan et al. (2012)), taylors in Accra - Field and Pandhe (2013), MFI clients in Kolkatta - Keniston (2011),... To bridge with macro evidence and improve policies, evidence must be complemented in two ways: 1. Sources of credit market imperfections (in specific settings) 2. Credit constraints among larger firms

3 This Project: Coffee Washing Stations (CWS) as a case study Attempt progress on both dimensions.

4 This Project: Coffee Washing Stations (CWS) as a case study Attempt progress on both dimensions. Internal records from a lender: working capital loans to 200 CWS in 18 countries

5 This Project: Coffee Washing Stations (CWS) as a case study Attempt progress on both dimensions. Internal records from a lender: working capital loans to 200 CWS in 18 countries Why CWS? Methodological Intrinsic Interest

6 Strategy Identifying sources of credit market imperfections: requires exogenous variation in incentives (Karlan and Zimman (2009)) 1. Technology: stark time separation of (-) and (+) cash flows

7 Strategy Identifying sources of credit market imperfections: requires exogenous variation in incentives (Karlan and Zimman (2009)) 1. Technology: stark time separation of (-) and (+) cash flows 2. Market: fluctuations in current and futures international coffee prices

8 Strategy Identifying sources of credit market imperfections: requires exogenous variation in incentives (Karlan and Zimman (2009)) 1. Technology: stark time separation of (-) and (+) cash flows 2. Market: fluctuations in current and futures international coffee prices identify ex-post MH (strategic default)

9 Strategy Identifying sources of credit market imperfections: requires exogenous variation in incentives (Karlan and Zimman (2009)) 1. Technology: stark time separation of (-) and (+) cash flows 2. Market: fluctuations in current and futures international coffee prices identify ex-post MH (strategic default) Testing for credit constraints: requires exogenous variation in K (and knowledge of production function) 3. Lender: internal scoring system RDD design on working capital loans size

10 Strategy Identifying sources of credit market imperfections: requires exogenous variation in incentives (Karlan and Zimman (2009)) 1. Technology: stark time separation of (-) and (+) cash flows 2. Market: fluctuations in current and futures international coffee prices identify ex-post MH (strategic default) Testing for credit constraints: requires exogenous variation in K (and knowledge of production function) 3. Lender: internal scoring system RDD design on working capital loans size 4. Technology: simple and known (interpretation)

11 Summary of Results 1. Detect strategic default: Unanticipated international coffee price defaults (OLS, event study, )

12 Summary of Results 1. Detect strategic default: Unanticipated international coffee price defaults (OLS, event study, ) 2. Identify credit constraints: Additional funds used to purchase more cherries, not to substitute other loans Returns to capital about 30%, well above interest rate Identify externality along the chain: prices paid to farmers increase

13 Summary of Results 1. Detect strategic default: Unanticipated international coffee price defaults (OLS, event study, ) 2. Identify credit constraints: Additional funds used to purchase more cherries, not to substitute other loans Returns to capital about 30%, well above interest rate Identify externality along the chain: prices paid to farmers increase 3. Quantify Value of Informal Enforcement: Ex-post IC constraint bounds value of informal enforcement V for borrowers V > 50% of value of sale contract Credit and market access equally important sources

14 The Plan 1. Context 2. Framework 3. Evidence 4. Discussion 5. Ongoing work on coffee

15 Context

16 What do Coffee Washing Stations do? Add pictures?

17 What do Coffee Washing Stations do? Add pictures?

18 What do Coffee Washing Stations do? Add pictures?

19 What do Coffee Washing Stations do?

20 What do Coffee Washing Stations do? Add pictures?

21 What do Coffee Washing Stations do? Add pictures?

22 Descriptive: Washing Stations and Loans Variable Observations Median Mean Panel A: Washing Station Descriptive Total Sales ('000 $) Number of Employees (Permanent) Number of Employees (Seasonal) Total Assets ('000 $) Costs of Goods Sold / Total Sales Number of Loans with Lender Panel B: Loan Descriptive Loan Amount ('000 $) Interest Rate Previous Loan with Lender Numerical Score Length of Loan Write - off Write - off or Restructured Write - off, Restructured, or Delay World Price at Closing World Price at Maturity

23 Lending Model 1. scoring system 2. monitoring of use of funds 3. account receivable with a foreign buyer used as collateral

24 Lending Model 1. scoring system 2. monitoring of use of funds 3. account receivable with a foreign buyer used as collateral Lower bound to the cost of credit market imperfections

25 Account Receivables as Collateral for WK Loans Not atypical: 1. Primary source of SME financing in US (Klepper (2004)) 2. Demirguc-Kunt and Maksimovic (2001) 3. Larson and Varangis (2006)

26 Sources of Finance (Rwanda) Financial Institutions Internal Funds Service Providers Cherries Purchased on Credit

27 Lending Model OUR LENDER FOREIGN BUYER STATION FARMERS

28 Lending Model OUR LENDER FOREIGN BUYER STATION FARMERS 1. Buyer and station negotiate a contract

29 Lending Model OUR LENDER FOREIGN BUYER STATION FARMERS 1. Buyer and station negotiate a contract 2. Lender extend loan to station (formula + value of the contract)

30 Lending Model OUR LENDER FOREIGN BUYER STATION FARMERS 1. Buyer and station negotiate a contract 2. Lender extend loan to station (formula + value of the contract) 3. Station purchases cherries during harvest time.

31 Lending Model OUR LENDER FOREIGN BUYER STATION FARMERS 1. Buyer and station negotiate a contract 2. Lender extend loan to station (formula + value of the contract) 3. Station purchases cherries during harvest time. 4. After harvest ends station delivers coffee to buyer and lender is paid.

32 Lending Model OUR LENDER FOREIGN BUYER Price (?) STATION FARMERS Three types of contracts: 1. Fixed price, 2. Price to be fixed (or at differential), 3. Mixed contracts (minimum price)

33 Default OUR LENDER FOREIGN BUYER STATION FARMERS OTHER BUYER

34 Theory

35 Theoretical Framework: Set-Up A washing station (station), a lender and a foreign buyer. Station has i) fixed capacity, ii) CRS technology below capacity, iii) only cost is purchasing cherries. Price of cherries p₀ is fixed and known to all parties. Station is subject to limited liability Delivery failure is very costly for the buyer.

36 Timing of Events t = 0 Contract with buyer

37 Timing of Events t = 0 Contract with buyer t = 1 Contract with lender

38 Timing of Events Negative Cash Flows: Ex-ante MH constraint t = 0 Contract with buyer t = 1 Contract with lender t = 2 Loan disbursed, Stations purchases Inputs or diverts

39 Timing of Events Negative Cash Flows: Ex-ante MH constraint t = 0 t = 1 t = 2 t = 3 Contract with buyer Contract with lender Loan disbursed, Stations purchases Inputs or diverts International price is realized

40 Timing of Events Negative Cash Flows: Ex-ante MH constraint Positive Cash Flows Ex-post MH constraint t = 0 t = 1 t = 2 t = 3 t = 4 Contract with buyer Contract with lender Loan disbursed, Stations purchases Inputs or diverts International price is realized Station repays or side-sells (default)

41 Ex Post IC Constraint: Set - Up Assuming a fixed price contract, the station side-sells if q c p c D p Q p q c V pq p U D 1 q c p c D U L Contract Extra sales Continuation Value if repay Value of defaulting Contract Continuation Value if late

42 Ex Post IC Constraint: Set - Up Assuming a fixed price contract, the station side-sells if q c p c D p Q p q c V pq p U D 1 q c p c D U L Contract Assume: Extra sales Continuation Value if repay Value of defaulting Contract Continuation Value if late U D U and U L V 1 U 1. Punishment only through probability σ of relationship s renewal 2. Continuation value conditional on relationship s death does not depend on default

43 Ex Post IC Constraint: Set - Up Assuming a fixed price contract, the station side-sells if q c p c D p Q p q c V pq p U D 1 q c p c D U L Contract Assume: Extra sales Continuation Value if repay Value of defaulting Contract Continuation Value if late U D U and U L V 1 U 1. Punishment only through probability σ of relationship s renewal 2. Continuation value conditional on relationship s death does not depend on default Rewrite as: V U Value of Informal Enforcement V 1 1 D p p c q c Temptation to Deviate

44 Ex Post IC Constraint: Implications Rewrite as: V U Value of Informal Enforcement V 1 1 D p p c q c Temptation to Deviate

45 Ex Post IC Constraint: Implications Rewrite as: V U Value of Informal Enforcement V 1 1 D p p c q c Temptation to Deviate Test 1 (Strategic Default): Given pc, unanticipated in p => likelihood of default Test 2 (Heterogeneity): Higher relationship value ( age) likelihood of default Fixed price contracts likelihood of default Note: All variables on RHS are observed in the data

46 Ex Ante MH and AS: Remarks Ex-Ante MH: - messier - Test: unexpected prices just after signing contract => default (Empirical analysis: in progress, still coding data). Adverse Selection: - we only observe loans granted, not applications.

47 Strategic Default

48 Empirical Strategy: Timing Consider the case of Rwanda Negative Cash Flows: Ex-ante MH constraint t=0 Contract with buyer t=1 Contract with lender t=2 Loan disbursed, Stations purchases Inputs or diverts Positive Cash Flows Ex-post MH constraint t=3 International price is realized t=4 Station repays or side-sells (default)

49 Empirical Strategy: Timing Consider the case of Rwanda Negative Cash Flows: Ex-ante MH constraint Contract Signed Harvest Season February March March to Late June / Early July t=0 Contract with buyer t=1 Contract with lender t=2 Loan disbursed, Stations purchases Inputs or diverts Positive Cash Flows Ex-post MH constraint Contract Executed August - December t=3 International price is realized t=4 Station repays or side-sells (default)

50 Prices and Default Default Contract Density Price ratio (maturity/future) Default Density kernel = triangle, degree = 0, bandwidth =.23

51 Prices and Default Density of Default Density of Contracts Price ratio (normalized by ratio from previous year) Default Contracts kernel = triangle, degree = 0, bandwidth =.21

52 Prices and Default: Robustness Regression Analysis: - Different Specifications (OLS / Probit) - Different Controls - Different Samples - Different Definitions of Default Event Study : - Look at 25 largest price increases on a month-on-month - RDD Heterogeneity: effect stronger on - New relationships - Fixed Price Contracts

53 Timing of Contracts: Closing and Maturity Dates

54 Prices and Default: Regression Analysis

55 Prices and Default: An Event Study Dependent Variable: Default T-Test [1] [2] [3] ** ** ** [0.0259] [0.0214] [0.0199] Window 2 Months 3 Months 4 Months Number of observations

56 Prices and Default: RDD

57 Prices and Default: Heterogeneity Dependent Variable: Default [1] [2] Mature After P Jump X First Loan Mature After P Jump X Fixed Price Contract Optimal Bandwith ** *** [0.0371] [0.0179] 75% Optimal Bandwith ** ** [0.0377] [0.0182] 125% Optimal Bandiwth * * [0.0311] [0.0371] Number of observations

58 What about Selection into Contracts? Analogy with the Risk management and borrowing constraints literature: 1. Negative selection (Froot et al. (1993)) 2. Positive Selection (Rampini and Viswanathan (2011)) [Still getting/coding the data]

59 Credit Constraints

60 Credit Constraints: Definition, Strategy and Test A firm is credit constrained if: MPK i 1 r i Challenge: MPK is not observed (and difficult to estimate)

61 Credit Constraints: Definition, Strategy and Test A firm is credit constrained if: MPK i 1 r i Challenge: MPK is not observed (and difficult to estimate) Strategy: loan size depends on a score RDD on loan size (holding constant r) Test (Banerjee and Duflo (2012)): firm is credit constrained if additional loan is used to expand inputs and sales, without (completely) substituting for existing, more expensive, loans.

62 Credit Constraints: Definition, Strategy and Test A firm is credit constrained if: MPK i 1 r i Challenge: MPK is not observed (and difficult to estimate) Strategy: loan size depends on a score RDD on loan size (holding constant r) Test (Banerjee and Duflo (2012)): firm is credit constrained if additional loan is used to expand inputs and sales, without (completely) substituting for existing, more expensive, loans. Remarks: - Decision at the margin - Other loans can be substituted.

63 Are Firms Credit Constrained? Definition MPK, r MPK Loans from other lenders Credit constrained firm K

64 Case 1: Our Lender is Mg. MPK, r MPK Loans from our lender Loans from other lenders K

65 Case 1: Our Lender is Mg. MPK, r MPK Loans from our lender Loans from other lenders Extra loans from our lender K

66 Case 2: Our Lender is Infra-Mg. MPK, r MPK Loans from our lender Loans from other lenders K

67 Case 2: Our Lender is Infra-Mg. MPK, r MPK Loans from our lender Loans from other lenders Extra loans from our lender K

68 Case 2: Our Lender is Infra-Mg. MPK, r MPK Loans from our lender Loans from other lenders Extra loans from our lender K

69 A Credit Constrained Firm MPK, r MPK Loans from our lender Loans from other lenders K

70 A Credit Constrained Firm MPK, r MPK Loans from our lender Loans from other lenders Extra loans from our lender K

71 RD on Loan Size (logs) Assignment variable (distance to score cutoff) NOTE: mean loan 300K sales contract 600K experiment approx. is: Loan 0.6 X 600K 0.4 X 600K = 120K

72 RD on interest rate r Assignment variable (distance to score cutoff)

73 Decision Process Committee in US who takes final decision OCS Card + Score Analysts in the US aggregates info for financial scores Commercial director deals with evaluation of buyer Local Offices collect information about client + information on local mkt. Station Buyer

74 RD Design on Score: no sorting Density Assignment Variable (distance to cutoff)

75 RD Design on Score: no manipulation

76 RD on Loan Size, Outcomes (logs)

77 RD on loans ( 000 $) No substitution

78 Value of Informal Enforcement V

79 Ex Post IC Constraint: Taking it to the Data Recall: V U 1 1 D p p c q c Ingredients: - debt level D - contract value - price at maturity - μ : probability of default, conditional on late payment and default - σ : probability of new loan conditional on late payment: We normalize it by the value of the contract amount

80 Defaults, Delays and Future Loans

81 2 Value of Informal Enforcement V Mean=0.540 St. Dev= Density Mean=0.211 St. Dev= Value of relationship with lender relative to sales Defaulters Non-defaulters 1.5

82 Value of Informal Enforcement V : Remarks What is a benchmark? 1. Fixed costs of exporting (Shapiro (1982)) 2. Roses from Kenya (Macchiavello and Morjaria (2012))

83 Value of Informal Enforcement V : Remarks What is a benchmark? 1. Fixed costs of exporting (Shapiro (1982)) 2. Roses from Kenya (Macchiavello and Morjaria (2012)) What does a large number mean? 1. Many valuable relationships out there do not emerge due to frictions 2. What is the source of relationship value?

84 What is the Value of Finance? V F 1

85 What is the Value of Finance? V F 1. 1 Probability of loan renewal ρ ¾. Discount factor: δ 0.9

86 What is the Value of Finance? V F 1 1. Probability of loan renewal ρ ¾. Discount factor: δ loan of approx. 120K $ => in profits of approx. 30K $. => 25% return

87 What is the Value of Finance? V F 1 1. Probability of loan renewal ρ ¾. Discount factor: δ loan of approx. 120K $ => in profits of approx. 40K $. => 30% return 3. Linear technology: extrapolate to average loan size (50% of sale contract) = 0.30% X 0.5 = 0.15% value of the sale contract => VF => approx. 60% of the value of informal enforcement.

88 Conclusions

89 Discussion & Policy Implications Strategic default & institutional features a) b) Imperfect enforcement in international trade Fixed price contracts Why fixed price contracts? hedging vs. counterparty risky challenges in providing access to hedging to LDCs exporters/producers Interlinked transactions Credit constraints and externalities along the chain Vertical integration Trade credit from farmers and trust Costa Rica Fair Trade as a bundle of contractual provisions: FT premium Minimum floor price + differential Pre-Financing

90 The Industrial Organization of the Coffee Chain Many (export oriented) agricultural chains in LDCs share the following structure (e.g.: coffee, tea, cocoa, ) FOREIGN BUYER EXPORTER bulking & sales 1st STAGE PROCESSING FARMERS (Small holders/ estates) These chains tend to be characterized by imperfect markets, e.g.: economies of scale in processing, perishability and hold-up, harvest seasonality, price and weather uncertainty, high post-harvest financing requirements, pre-harvest finance to farmers, commercialization and quality standards Throughout time and place, government interventions of various forms (boards, monopsony, subsidies, zoning regulations, ) Within crops, enormous variation across countries in industrial structure

91 The Industrial Organization of the Coffee Chain Paper is part of a broader agenda on the industrial organization of the coffee chain focusing on coffee washing stations FOREIGN BUYER EXPORTER bulking & sales COFFEE WASHING STATIONS FARMERS (Small holders/ estates) Why coffee (washing stations)? - Intrinsic: Millions farmers, grown in 40+ countries, many extremely poor After oil, commodity with largest trading value in the world. Methodological: - Simple production processes: ideal to measure dispersion in productivity and MPKs, - Volatile harvest and market conditions to study market failures.

92 The Industrial Organization of the Coffee Chain Project I: Tropical Lending (with Arthur Blouin) COFFEE WASHING STATIONS FOREIGN BUYER FARMERS (Small holders/ estates) LENDER Question: Is the sector characterized by widespread credit constraints? If yes, what is the underlying source? What are the consequences for the stations and the farmers? Data: Loan level database (+ contract & stations financial records) of an international lender supplying credit to 300+ stations in around 20 countries.

93 The Industrial Organization of the Coffee Chain Project II: Competition and Efficiency in Rwanda (with Ameet Morjaria) FOREIGN BUYER EXPORTER bulking & sales COFFEE WASHING STATION A COFFEE WASHING STATION B FARMERS (Small holders/ estates) Question I: What is the impact of competition on station efficiency and farmers? Are there pecuniary externalities non mediated by prices? Question II: What role does trust play in the functioning of stations? Data: Comprehensive survey of (almost) all stations in Rwanda + detailed admin. & GIS data. Engineering model used to instrument for competition

94 The Industrial Organization of the Coffee Chain Project III: Adaptation and the Costs and Benefits of Vertical Integration (with Pepita Miquel-Florensa) EXPORTER bulking & sales FOREIGN BUYERS EXPORTER bulking & sales COFFEE WASHING STATION A FARMERS (Small holders/ estates) COFFEE WASHING STATION A Question: Is supply assurance a key driver of vertical integration decisions? Are firms boundary decisions efficient or is there too much / too little integration? Data: Comprehensive administrative data on sales, inputs, costs, payments from Costa Rica

95 Related Literatures I Credit Constraints (in larger firms), Credit and Exports, Trade Credit: - Banerjee and Duflo (2012): key reference on reduced form tests for credit constraints - McMillan and Woodruff (2000): trade credit as a measure of trust - Other work in CF: e.g., Klapper (2006), Klapper et al. (2011) - Other work in Trade: e.g., Paravisini et al. (2011), Manova (2013a, 2013b) Reputation & (Relational) Contracts: - Banerjee and Duflo (2000): what matters is what people think you ll do, not what you actually do - Macchiavello and Morjaria (2014): - More on dynamics: Antras and Foley (2014), Macchiavello (2010)

96 Related Literatures II Agriculture & Coffee: - Fafchamps (et al., various): agricultural value chains - Banerjee et al. (2001): why coops often do not work well? - Mullhainathan and Sukhatankar (2014): private vs coop: does it matter? - Dragusano and Nuun (2014): does fair trade work? - De Janvry et al. (2014): why fair trade is unlikely to work? - Older literature on interlinked transactions: Bardhan et al. - Casaburi and Reed (2013): pass-through with interlinked transactions - Contract Farming: e.g., Deb and Suri (2013), Casaburi et al. (in progress)

97 Related Literatures III Empirics of Contracts - Chiappori and Salanié (2002): a reduced form test to detect AI - Karlan and Zimman (2010): a clever two-stage experiment to distinguish MH and AS - Adams et al. (2009): if you can credibly estimate demand for loans then you can structurally back out MH and AS - Townsend (et al., various): too much! Vertical Integration - Atalay et al. (2014), Forbes and Laderman (2010): what is integration about? Productivity Dispersion & Management - Syverson (2012): growing interest in understanding PPDs across firms - Hsieh and Klenow (2009): the dispersion appears to be larger in LDCs - Bloom et al. (various): management appears to have something to do with it - Madrigan and Xu (2013): nice paper on misallocation and credit.

98 Thanks!