Jose Gonzalo Rangel

1
Macroeconomic Announcements, Price Discovery, and
Order Flow Effects in the Stock Market Evidence
from Incomplete Data and Multiple Financial
Markets

• Jose Gonzalo Rangel
• UCSD
• Capri Workshop
• May 25, 2006

2
Motivation

• Asset prices are affected by revisions in
  expectations driven by news about changing
  economic conditions (e.g. output, employment and
  inflation shocks).
• The ultimate objectives of monetary policy are
  expressed in terms of same macroeconomic
  variables (Bernanke and Kuttner, 2005).
• The stock market response to macroeconomic news
  is linked to market assessments (investors
  beliefs) of future states of the economy and/or
  Fed actions.
• However, the mechanism through which these
  beliefs enter equity prices remains an intriguing
  empirical question.

3
Main Approach

• Announcements are pure symmetric information
  events. Beliefs are homogeneous. The transmission
  mechanism involves a nearly instantaneous price
  adjustment (jumps, little trading activity
  involved)
• Andersen et al.(2003, 2005), Boyd, Hu, and
  Jagannathan(2005), Bernanke and Kuttner(2005)
• Problem under asymmetric information the
  market needs to aggregate heterogeneous beliefs.
  Transmission mechanism involves a learning
  process. Learning occurs through trades.
  Fundamental price is affected by the order flow
  (sum of signed trades).
• Important effects on price dynamic behavior
  (price discovery), liquidity, and volatility.
• Evans and Lyons (2004), Brandt and Kavajecz
  (2004)

4
Price Effects of Trading Activity on Announcement
Days

• Data Evidence Amihud (2002) illiquidity ratio

5
Order Flows Role Graphically
Symmetric Information Approach

Microstructure Approach
Private info
Price
Order flow
Hybrid
Information
Price
Order flow
6
Main Empirical Results

• Significant instantaneous news impacts of news
  related to real activity, investment, inflation,
  and monetary policy.
• Significant order flow and/or asymmetric
  information effects on employment days due to
• Uncertainty on the implications of employment
  news for stock prices
• Increases in the volatility of fundamental prices
• Asymmetric Information effects come from the
  interest rate component of equity prices.
• Evidence of excess sensitivity of long term
  interest rates to employment shocks
• Private agents revise expectations about future
  Fed policies and/or long run states of the
  economy (Gurkaynak, Sack, and Swanson, 2005)
• Revisions are not homogeneous

7
The Microstructure View

• Observed transaction price
• Random walk representation of the unobserved
  fundamental (log) price mt
• qttrade direction (1 if buy, -1 if sell)
• Under symmetric information
• where ut accounts for arrival of public
  information over (t-1,t

8
The Microstructure View

• Under asymmetric information (Hasbrouck, 2004,
  2005)
• where
• 1(or -1) if transaction k was initiated by
  the buyer (or seller). Nt number of trades over
  (t-1,t.
• Similar results if Qt is proxied by signed
  volume, Vtqt, where Vtf(volume), and qt
  represents sign(Qt).

9
A Simple Asymmetric Information Model (SAIM)

• Theoretical basis Kyle (1985), and Glosten and
  Milgrom(1985)
• Suppose Kyles framework in a one period model

10
Results and Comparative Statics of (SAIM)

• Proposition A1 There is a unique linear
  equilibrium in which
• Comparative statics
• ?? when fundamental price volatility ? (s2 ?)
• ? ? when volatility of liquidity demands ? (su2 ?
  )
• ?? when precision of the signal ?
• (se2 ?, for M sufficiently large)
• ?? when number of informed traders ?
• (M ?, for M sufficiently large)

11
Empirical Specification (SAIM)

• Estimation of a structural microstructure model
  based on
• Hasbrouck (2004) extension of Glosten and Harris
  (1988), including price impacts of trades on
  fundamental prices.
• Incomplete data, daily frequency
• Additional extensions
• News effects on the efficient price
• Average incremental effects of order flow on
  announcement regimes vs the non-announcement
  regime.

12
Empirical Specification (SAIM)

• Observed and efficient prices
• where
• Ik,t type k announcement Indicator
• Vt f(trading volume) (1, Vt)
• qt trade direction
• Yk,tRealization of type k macro variable
  (Yhatforecast)

13
Empirical Specification (SAIM)

• Observed Price
• Returns
• Conditional variance of returns

14
Estimation Issues and Econometric Approach (SAIM)

• Remarks
• Model keeps the same form under time aggregation
• Parameters of interest ?, , and ßs
• Observed variables prices, volume, news
  variables
• Unobserved (latent) variables qt and mt
• 2T possible paths of qt (Qt2T, TSample size,
  e.g. 3,120 days)
• Likelihood

15
Estimation Issues and Econometric Approach (SAIM)

• How to estimate?
• Simulated Maximum Likelihood (SMLE)
• Bayesian Markov Chain Monte Carlo (MCMC)
• Hasbrouck (2004, 2005)
• MCMC advantages
• Computationally convenient
• Parameter uncertainty
• Uncertainty about news effects
• Uncertainty about asymmetric information

16
The MCMC Algorithm

• Desired posterior
• Given the set of parameters
• and q(0)
• Step1 Draw T(1) from
• Step2 Draw q(1) from
• Step3 Continue in this fashion until
• generate a sequence
• whose limit is the desired posterior F

17
Data

• SP500 (daily) data on closing prices and volume
  from CRSP. Sample period 1992-2003.
• Futures (daily) data on closing prices and volume
  for SP500, bonds (5Y, 10Y) and exchange rates
  (US/YEN) from Datastream
• 19 Macroeconomic announcements and forecasts from
  MMS regarding
• Real Activity IP, RS, NFP, UMP, CU, PINC, and
  CCR
• Consumption NHS, PCE
• Investment DGO, CS, and BI
• Trade GSTB
• Price Level CPI, PPI
• Forward Looking LI, NAPM, and HS
• Monetary Policy FOMC/FFR

18
Hist
19
(No Transcript)
20

• Slide lambda

21

• Comparative statics
• ?? when fundamental price volatility ?
• ? ? when volatility of liquidity demands ? (su2 ?
  )
• ?? precision of the signal ?
• (se2 ?, for M sufficiently large)
• ?? when number of informed traders ?
• (M ?, for M sufficiently large)
• Back

22
Robustness Evidence from Bond Markets
23
Evidence from Exchange Rates
24
Summary of Empirical Results

• For the stock market
• Instantaneous fundamental news impacts
  (asymmetric)
• Order flow effects on employment days
• For long term bond markets
• Fundamental news effects as predicted by the
  asset pricing view
• Strong order flow effects on employment days
• For exchange rates
• Just fundamental news effects on employment days

25

• Results consistent with recent literature
• Pasquariello and Vega (2005) Day-to-day bond
  yield changes and order flow are most sensitive
  to Nonfarm Payroll Employment announcements
  (based on intradaily data) .
• Morris and Chin (2002) Overreaction to
  employment news. Bond yields are most reactive to
  the types of news emphasized by the press.
• Does employment convey more information about
  future growth? No evidence
• Does employment convey more information about
  future inflation? More likely

26
Contribution Why is this distinction interesting?

• Relevant for practitioners and policy makers.
• Provides new methods for measuring impacts of
  output and inflation shocks in financial markets.
• Provides new measures on how homogeneous is the
  market evaluation of future Fed reactions to
  these shocks.
• Provides an explanation for observed patterns in
  different price characteristics, such as
  volatility and liquidity.
• Contributes to a better understanding of link
  between macroeconomic information and the price
  discovery process (one of the main functions of
  financial markets).
• Assets trade in markets, markets provide
  liquidity and price discovery, and asset prices
  are influenced by the transaction costs of
  liquidity and the risk of price discovery
  (OHara, 2003)

27
Concluding Remarks

• Evidence of incremental asymmetric information
  costs on employment days.
• Changes in the asymmetric information coefficient
  on employment days due to
• Uncertainty on the implications of employment
  news for asset prices
• Increases in volatility of fundamental prices.
• Bond markets point to asymmetric information on
  the interest rate component of stock prices.
• Consistent with the excess sensitivity of
  long-term interest rates
• Not only investors change their long run
  expectations of the state of the economy and
  long-run Fed policies, but also they have
  heterogeneous beliefs.

28
Future Research

• Analysis with complete information
• More flexible specification for conditional
  volatility
• Time varying news effects
• Time varying order flow effects
• Explore correlations in the trade direction (or
  order flow)
• Analysis of individual stocks
• Include earnings announcements

29
(No Transcript)
30

• Conditional Posterior for the latent trade
  direction
• Where,
• Back

31

• Observed Price
• Returns

32
Posterior Distributions for News Effects and
Asymmetric Information Parameters
33
Asymmetric Information Effect SP500
Cumulative Impact (Basis Points)
34
Asymmetric Information Effect (Futures SP500)

• Cumulative Impact (Basis Points)

35
Assumptions

• Post announcement true value,
• M informed traders get noisy signals about the
  true price impact of a particular news event,
• Informed agent i demands xi units of the asset
• Noise traders demand
• A market maker sets prices after observing
  aggregated order flow. Fundamental post
  announcement price satisfies market efficiency