The microstructure of exchange rate management: FX intervention and capital controls in Brazil

Abstract

This article uses a microstructure approach to analyse the effectiveness of capital controls introduced in Brazil to counter an appreciation of the Real. Based on a rich data set from the Brazilian foreign exchange market, we estimate a reduced-form VAR to characterize the interaction of the central bank, financial and commercial customers in times of regulatory policy measures. We find that capital controls change market participants’ behaviour, and that central bank interventions elicit a significant response in financial order flow. Referring to the source of order flow, we find no direct price impact by financial flows and thus no evidence that the appreciation of the Real is driven by financial customers’ activity. Instead, commercial customers seem to be a primary driver of the Real within our model.

Keywords:

• foreign exchange
• sterilized intervention
• capital controls
• market microstructure
• VAR

JEL Classification:

• F31
• E58
• G14
• G15

Acknowledgements

The authors are in deep debt with Harmen Lehment for the organization of the Advanced Studies Program of the Kiel Institute for World Economy. We thank seminar participants at the Kiel Institute for World Economy and the 3rd ISCEF conference in Paris. We are also grateful for comments from Jörg Breitung, Emanuel Kohlscheen, Lucas Menkhoff and Lucio Sarno.

Notes

¹ See ‘What’s the currency war about?’ from BBC, 22 October 2010.

² See the article ‘Brazil in “currency war” alert’ from 27th September 2010, in the Financial Times. The perception of a currency war is at the time of writing not over. On 8th February 2013, Guido Mantega warns again, see ‘Global currency war could get nastier’ from Reuters.

³ See ‘The Liberalization and Management of Capitals Flow: An Institutional View’, IMF 2012.

⁴ More generally, Canales-Kriljenko (2003) summarizes the use of capital controls and regulatory requirements across a larger set of emerging economies.

⁵ That is, 570 out of 913 trading days.

⁶ While Wu is primarily concerned with a description of the anatomy and dynamics of the market, our purpose here is to address the context for exchange rate management. In addition, our use of recent macroprudential policies controls for parameter instability derived from changes in the policy environment, a concern that Neely (2005) explains faces similar structural models of intervention.

⁷ Kohlscheen (2014) shows that there is no statistical difference between the use of the CRB commodity index and an index weighted by the Brazilian trade balance. VIX is the implied volatility of the S&P 500 and is used as a proxy for international risk perception.

⁸ The same perception is found in Kohlscheen (2012) and Menkhoff (2013).

⁹ Here, we follow Lutkepohl (2005) terminology, where the A matrix represents the restrictions on the endogenous variables and B the restrictions on the covariance matrix of the error terms.

¹⁰ Another way to show that the error term is orthogonal is to change the ordering of the variables and show that the Cholesky impulse responses produce the same results, but as this is not a formal testing procedure, we opt for the method described above. Nevertheless, as a check for robustness we carried out the test in both formats, confirming that the results are consistent across these methods.

¹¹ The rank of the largest exporters in 2012 from the Ministry of Development, Industry and Foreign Trade reported 5 foreign companies in the top 10.

¹² As a check for robustness we also considered a 10 day moving average for cumulative order flow and found it significant to the 1% confidence interval.

¹³ We suggest that this is the outcome of the autoregressive interaction of the variables. Since order flows from private and official sources exhibit negative reaction coefficients with respect to the second-order lag of the exchange rate return, a positive correlation among order flow variables might be observed, but cannot be interpreted as a direct influence.

¹⁴ It is notable that the variation in the Fed Funds rate is negligible, as it did not move above 0.25% during our observation horizon.

¹⁵ The literature has been accumulating empirical evidence for the failure of uncovered interest parity since the seminal work of Fama (1984).

¹⁶ This argument is independent of the role that fundamentals play in driving the FX activity of financial participants, which is the common focus of the debate. We recall that unlike most advanced countries, commercial customers in Brazil are a relatively significant share of the market.