Lecture 3 section 1 l.jpg
This presentation is the property of its rightful owner.
Sponsored Links
1 / 31

LECTURE 3: Section 1 PowerPoint PPT Presentation


  • 55 Views
  • Uploaded on
  • Presentation posted in: General

LECTURE 3: Section 1. Forecasting Currency Returns: Exchange Rate Puzzles. Puzzle 1: Why don’t the “fundamentals” explain exchange rates?. If not fundamentals, then ??? Speculative bubble? Nonrational behavior (systematic forecasting errors)? Omitted variables? Recall the hybrid model:

Download Presentation

LECTURE 3: Section 1

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Lecture 3 section 1 l.jpg

LECTURE 3: Section 1

Forecasting Currency Returns:

Exchange Rate Puzzles


Puzzle 1 why don t the fundamentals explain exchange rates l.jpg

Puzzle 1: Why don’t the “fundamentals” explain exchange rates?

If not fundamentals, then ???

  • Speculative bubble?

  • Nonrational behavior (systematic forecasting errors)?

  • Omitted variables?

    Recall the hybrid model:

  • Order flow reflects expectations revisions omitted from macro models


Puzzle 1 why don t the fundamentals explain exchange rates3 l.jpg

Puzzle 1: Why don’t the “fundamentals” explain exchange rates?

Daily model of foreign exchange market: round 1

i) Observe daily returns as realized increments of macro info:

ii) Dealers quote to customers,

  • equilibrium has all quoting same

    iii) Receive customer order


Puzzle 1 why don t the fundamentals explain exchange rates4 l.jpg

Puzzle 1: Why don’t the “fundamentals” explain exchange rates?

Round 2: Interdealer trades

  • Each dealer quotes scalar 2-way price:

  • All trade simultaneously at same price:

  • Net interdealer trade initiated by dealer i:

  • At close of period all dealers observe net interdealer order flow:

    • Flow conveys size & sign of public order flow in round 1


Puzzle 1 why don t the fundamentals explain exchange rates5 l.jpg

Puzzle 1: Why don’t the “fundamentals” explain exchange rates?

Round 3: Dealers share risk with public

  • each dealer simultaneously & independently quotes to public to square positions

  • Public demand is function of expected return conditioned on public info:

  • Public willingness to absorb larger FX position for given reflected in

  • Portfolio balance effect required to induce public to absorb order flow:

  • At end of day:


Puzzle 1 why don t the fundamentals explain exchange rates6 l.jpg

Puzzle 1: Why don’t the “fundamentals” explain exchange rates?

Evans & Lyons estimated model of:

  • Interest differential reflects macro effects (overnight rates)

  • Order flow significant determinant of exchange rates (daily flow)


Puzzle 2 excess volatility l.jpg

Puzzle 2: Excess volatility

Managed exchange rates less volatile than floating exchange rates, but macro determinants shouldn’t vary with regime so can macro factors really be important?

  • Order flow may be more informative with floating rates

    • With credible fixed rate FX is riskless asset & order flow has no market impact

  • Killeen, Lyons, & Moore model shift from flexible to fixed regime

    • Fixed regime is absorbing state

    • Model similar to previous except in fixed regime:

    • Now elasticity of demand for currency is function of volatility


Puzzle 2 excess volatility8 l.jpg

Puzzle 2: Excess volatility

KLM estimate model of flexible rates prior to euro fixings of May 4, 1998

  • Prior to May 4, 1998 expect exchange rate & its determinants to all be nonstationary & cointegrated

  • After, exchange rate is stationary but determinants still nonstationary

    • So need coefficient on order flow to be zero for balanced equation

  • Excess volatility beyond macro variables coming through order flow impact


Puzzle 3 forward bias puzzle l.jpg

Puzzle 3: Forward bias puzzle

Forward rate typically predicts change in exchange rate with wrong sign

  • Null of unbiasedness:

    Economists perspective:

  • Inefficiency: systematic errors & ignore profit opportunities

  • Risk premium: but models unable to capture magnitude of bias without unreasonably large risk aversion


Puzzle 3 forward bias puzzle10 l.jpg

Puzzle 3: Forward bias puzzle

Practitioner’s perspective: is bias large enough to generate a Sharpe ratio comparable to buy & hold equity strategy

  • SR = 0.4 for equities

  • Need b=-1 or 3 to get similar SR for currencies, so band of rational inattention

    • True that backtest IR is determinant of investment strategy

      Why is b estimate closer to -1 lower bound?

  • Leveraged investors won’t pursue strategy

  • Unleveraged (real money) & nonfinancial corporates have “limited participation” in currency strategy

    • Will adjust portfolios slowly to change in interest rates

      • So F adjusts to new interest rates, but S adjusts slowly as funds are slow to move & may appreciate for a while


Lecture 3 section 2 l.jpg

Lecture 3: section 2

Galati & Heath

Why are nonfinancial customers the fastest growing segment of the FX market?


Growth of nonfinancial customers attractive returns l.jpg

Growth of nonfinancial customers: attractive returns

Leveraged investors have earned attractive SR from currency strategies

  • Graph 1 shows SR from carry trade (short JPY, long AUD, NZD) compared to alternatives

    • Since 2001, SR in range of 1.5-2 for AUD, NZD

    • AUD & NZD were had fast growth in turnover


Growth of nonfinancial customers growth of hedge fund aum l.jpg

Growth of nonfinancial customers: growth of hedge fund AUM

AUM grew with currencies’ development as an asset class

Growth of PB services

  • Fund trades with multiple counterparties but give up trades to PB

    • 1 credit relationship, settle with 1 bank


Growth of nonfinancial customers what else l.jpg

Growth of nonfinancial customers: what else?

Algorithmic trading (don’t know how much)

Growth of emerging markets

  • Faster than developed markets

  • HKD particularly fast growth due to China equity markets & IPOs


Lecture 3 section 3 l.jpg

Lecture 3: section 3

Cheung & Chinn

Survey of US FX Dealers


Key survey results l.jpg

Key survey results

65% of trades interbank, 35% customer

Conventional spreads

  • GBPUSD, 5bps; USDCHF, 5 bps; USDDEM, 3 bps; USDJPY, 3bps

    • Quote market norm to maintain “equitable & reciprocal trading relationship”

    • Most frequent reason for deviating: thin & volatile market

    • Of course, the spread rises with quantity

      Large player advantage from large customer base

      Exchange rates unpredictable intraday but more predictable over period of months


Key survey results20 l.jpg

Key survey results

Trading methods evenly distributed over

  • Technical trading, customer orders, fundmentals, “jobbing”

    • Jobbing: continuously buying & selling to make small profits from many trades (a precursor to algo??)

      News effects reflected in exchange rates largely within 1 minute

  • Importance of different news shifts over time

    • Maybe think of varying parameter models

      Exchange rates only reflect fundamental value in long-run

  • Over 6 months, not at all intraday

  • PPP not seen as important concept & not relevant to most traders even at >6 months


Lecture 3 section 4 l.jpg

Lecture 3: Section 4

Brunnermeir et al.

Carry Trade & Crash Risk


Crash risk l.jpg

Crash Risk

Reflected in skewness of currency returns

  • JPY has negative interest differential & positive skewness

  • AUD, NZD have positive interest differential & negative skewness

    Price of crash risk

  • Risk reversal: long out of money call; short out of money put

    • = 0, If risk neutral distribution of exchange rates is symmetric

    • Not 0, if skewed distribution

    • RR measures expected skewness & skewness risk premium


Empirical approach l.jpg

Empirical approach

8 currencies versus USD

Abnormal return measured as:

  • 3 month interbank interest rates

  • Exchange rate is foreign currency price of USD

  • USD is funding currency

  • With UIP, abnormal return = 0

    • Use daily data to compute skewness by quarter

  • Use CFTC positioning data to measure “Carry trade”

    • Positions of non-commercial traders (not hedgers)


Empirical results l.jpg

Empirical Results

Estimate models with excess returns, positioning, & skewness on LHS of regression

  • Interest differential >0 & strong predictor of excess return & marginal predictor of positioning

    • Lags die off quickly

  • Interest differential <0 & highly significant predictor of skewness

    • Lags persist & only die out slowly

    • Chance of “coming down by the elevator” are greater, the higher interest differential

      Also estimate density of FX returns conditional on interest differential

  • Higher interest differential has higher mean but bigger left-tail risk


Forecasting crashes price of crash risk l.jpg

Forecasting Crashes & Price of Crash Risk

Regress skewness at t+1 & risk reversals at t on

  • Interest differentials:

    • <0 & significant for skewness

    • Insignificant for RR

  • Excess return z:

    • <0 & significant for skewness

    • >0 & significant for RR (part of unwind comes from buying insurance against downside risk?)

  • Futures positioning:

    • Not significant for skewness if z in regression (z may be better indicator of positioning than positioning variable)

    • >0 and significant for RR

  • Lagged Skewness: >0 & significant for Skewness & RR

  • Lagged RR: <0 & significant for Skewness (higher insurance price, lower downside risk)

    • Post crash, speculators pay more for insurance yet positioning is lower, lowering crash risk


Carry unwind state l.jpg

Carry Unwind State

Would expect redemptions, margin calls, lower risk tolerance to motivate unwinds

Estimate model of change in carry positioning as f(VIX, lagged positioning)

  • Interact VIX with interest differential to get sign right & find <0 & significant

    Estimate model of change in RR as f(VIX, lagged RR)

  • VIX effect <0 & significant (price of crash risk rises)

    Estimate model of z on same

  • VIX effect <0 & significant (return to carry falls)

    Replace VIX with TED spread & find useful for predicting future change in RR & z


Predicting co movement of exchange rates l.jpg

Predicting co-movement of exchange rates

If carry trades affected by shifts in risk tolerance should expect funding currencies to all move together and same should be true for investment currencies

  • Implication is that currencies with similar interest rates should move together, while currencies with different interest rates should move apart

    Estimate model of pairwise correlation of exchange rate changes (13 week non-overlapping) as f(abs value of interest differential, correlation of interest rate levels (monetary policy control), and average correlation across all pairs of exchange rates (common factor in all correlations)

  • Correlations are negative function of interest differential (reduction in int diff raises exchange rate correlations)


  • Login