Capital Controls and Misallocation in the Market for Risk

Lorena Keller, Northwestern University Email: [email protected] Website: https://sites.google.com/site/kellerlorena/

Research Question: Do capital controls on carry trade inflows affect firms’ foreign currency liabilities? Are there consequences on employment after a depreciation shock? Contribution 1

I show a new side effect of capital controls

Details Capital controls induce local banks to substitute 10-20% of their lending in local currency for lending in dollars Banks in EM have a fundamental risk problem: households save partially in dollars. Absent capital controls banks hedge the associated FX risk with foreign investors.

2

New mechanism

3

Natural experiment to show channel at work

I exploit heterogeneity in the strictness of capital controls across Peruvian banks and obtain unique, very detailed data to provide causal evidence of this mechanism

5

Importance: Large real effects

I show capital controls led to 6-11% decrease in employment after a sudden stop

With capital controls, banks can no longer hedge with foreigners. Hence, banks shit FX risk to firms by lending dollars.

Consensus post financial crisis: Countries should set capital controls on carry trade inflows to prevent sudden stops • Importance of preventing sudden stops today: • Greater risk of sudden stops: • Low dollar rates led to demand of emerging markets (EM) assets to earn the int. rate differential between EM and dollars (carry trade)

2 Without capital controls: Lend PEN (A.2.2). Hedge FX w/ foreigners (A.3)

• Banks have $ 100 deposits and hedge buying $ 100 forward to foreigners. Foreigners want this because during inflows foreigners use forwards to buy PEN (asset), sell USD (liabilities)

• Forwards liquidate at t + 1 → Banks have $ 100 deposits at t to lend in

Use DiD to disintangle credit supply from credit demand • I compare how treated banks change USD & PEN loans after capital controls compared to non-treated to the same pool of firms.

q=−1 q=1 X X Loans in USD =β0 + β1CC + βiCC ∗ P ostt=2011m1+q mo + βiCC ∗ P ostt=2011m1+ Total Loans b,f,t q=−11 q=12

+ ΓX + F irm ∗ DateF E + υbf t

PEN to firms (FX hedged)

• Greater sensitivity to sudden stops:

• Data: Confidential loan level data (bank-firm-time) and all outstanding

• Non-US banks increasing dollar liabilities (hold $ 10tr. liabilities)

forward contracts for the universe of Peruvian banks. Source: SBS

Implementation of capital controls • Focus on capital controls on carry trade inflows

Results at the bank level Treated banks increase USD lending by 10-15%

• Foreigners use forwards to obtain assets in EM currency and liabilities in dollars

.2 .2

3 With capital controls: Restrictied hedge w/ foreigners. Hedge the rest

0 .1

lending USD. β

• Then, countries (eg. Colombia, Peru, Korea) set capital controls as lim-

Treated banks decrease PEN lending by 20-40%

its to forward positions of banks (counterparties of foreign investors).

β

Context

• Assume capital controls limit forwards to $25. To hedge remaining $75: -.4

-.1

ing in EM currency

• Hedging regulation: EM bank regulators force banks to hedge FX risk (Canta(2006)) post 1998 Asian crisis

• How do banks hedge if capital controls from hedging with forwards?

I find a new side effect of capital controls This is the first paper to study and show that capital controls increase dollarization of EM firms’ debt because capital controls induce local banks to lend more in dollars.

Fi r ms

As s e t s Li a bi l i t i e s

D 50PEN 25US

As s e t s Li a bi l i t i e s

75US D 50PEN

100USD

25US D

50PEN

Theoretical Predictions

Att Househol ds

Att +1

• When capital controls are present, banks (1) lend more in USD and (2) less in PEN (scenario (3) versus (2))

Testing Predictions: Natural Experiment (Peru) • Peru: 2011 capital controls limited banks’ forward positions. Heterogeneous treatment as limits varied across banks (function of equity)

Mechanism

• Treatment (CCb,22Jan11): Use % used of limit at capital controls announce-

Toy Example

ment to split banks into: • Treated: Above threshold (CCb,22Jan11 = 1) •

• Assumptions: EM country (Peru). (1) Households save 100 dollars (2)

Control: Below threshold (CCb,22Jan11 = 0)

2010m7

2011m1

2011m7

2012m1

2010m1

2010m7

B. Log(USD Credit + 1) (FX:2005m2)

2011m1 C. Log(PEN Credit + 1)

Results at the firm level • Firms can substitute across banks. Hence, aggregate credit at firmmonth level to study the currency composition of loans at the firm level

• I find that firms borrowing from treated banks increased USD loans by 29.6% compared to those borrowing from non-treated banks.

Employment in firms that borrowed from treated banks after a sudden stop • I split firms into (i) firms borrowing from treated banks (treated firms) and (ii) firms borrowing from non-treated banks (non-treated firms) at capital controls announcement

• Study employment of treated vs non-treated firms after a sudden stop (FED taper tantrum, May 2013. After this, PEN depreciated 30% (shaded 1.05

Perm. workers if firm exp22Jan2011 < 100% (lhs, Normalized MA) Perm. workers if firm exp22Jan2011 = 100% (lhs, Normalized MA) FX (PEN/USD) (rhs)

3.2

1 3

.95

1.05

1

CC in Effect

Normalized Ratio USD (Date:2011m1)

1 Lend in dollars (A.2.1). Then FX hedged (A.3)

2010m1

1.1

firms (A.2.3) (3) Banks hedge FX risk (A.3). These hold broadly in EM

• Scenarios:

-.6

area, Fig. 2))

• Outcome Variables: USD and PEN loan of firm f from bank b at time t

CC Announcement

Banks decide either: (1) lend dollars (A.2.1) or (2) lend EM curr (PEN) to

-.2

.9

Taper Tantrum

• Therefore, banks have risk management problem: (FX risk) when lend-

banks lend $75 to firms. Then banks can only lend $25 in PEN Loc a l Ba nks

CC Announcement

EM Setting: Banks have Risk Management Problem • Dollarization: Banks in EM have dollar deposits from households.

-.2

0

2.8

2.6

.85 2010m1

.95 2010m1

2010m7

2011m1 Below 100% Limit

2011m7

2012m1

2012m1

2013m1

2014m1

2015m1

2016m1

Figure 2: Currency depreciation and employment of treated and non-affected firms

Above 100% Limit

A. Ratio USD (bp, FX:2005m2)

Figure 1: Percentage of local bank’s lending in dollars for Treated and Non-Treated Banks

Figure 1 shows treated banks increase the share of lending in USD after capital controls

2011m1

Balanced sample

Figure 2 shows treated firms reduce employment after sudden stop

• Using DiD: Treated firms decreased permanently employed workers by 11%∗∗ compared to non treated post taper tantrum.

2011m7