finance discussion writing

boleroblack_qi

Lecture6.pdf

Home >Business & Finance homework help >Financial markets homework help >finance discussion writing

Alok Kumar Department of Finance Miami Business School

[email protected]; 305‐284‐1882

FIN 686 Psychology of Financial Markets and Financial Decision Making

LECTURE NOTES 6

Asset Prices Under The Behavioral Framework (Summary from Week 1)

Behavioral View: Asset Prices Recall: Under the noise trader framework, in equilibrium, the asset

price at time t is given by:

Yt = total value of stocks demanded by ordinary investors.

When  ≈ 0: we get the efficient market hypothesis.

When  is large: noise traders determine asset prices.

Behavioral View: Asset Prices II  Our goal this week is to find different ways of estimating and

predicting the noise trader demand Y. We refer to Y as “investor sentiment”.

We will also try to estimate , i.e., the proportion of noise traders in the market.

 Specifically, first, we will examine the pricing effects of local investor sentiment.

 Next, we will examine the pricing effects of political sentiment.  Last, we will study how geographical dispersion of information

affects asset prices because investors are slow in aggregating this information.

Geography and Finance

In the first scenario, we will see that local economic conditions affect the risk aversion of local investors. As a result, local asset prices are predictable, especially when local investors exhibit stronger local bias and are less sophisticated.

Main Question Can state‐level (i.e., local) macroeconomic variables

predict state portfolio returns?

Example: When the economy of Texas is doing poorly, do firms headquartered in Texas (i.e., the Texas portfolio) systematically earn higher or lower returns in the future?

Approach:  Construct geographical portfolios and identify “local”

recession indicators.  Predict returns using local macro variables and achieve out‐

of‐sample predictability.

Predictability Mechanism

When local economic conditions are poor (relative to the U.S.):

 Local risk aversion rises, risk sharing opportunities become scarce.

 Local investors reduce exposure to risky assets (including local stocks).

 They sell local stocks; Local stock prices are depressed.

 In the future, local stocks should yield higher returns.

Local bias could amplify these effects for local stocks.

State‐Level Predictors Three local economic indicators  BEA State Personal Income Growth

 Proxy for returns to human capital

 Relative State Unemployment (from BLS)  u – moving average of u past 16Q (results similar if we use 8‐10Q)  Simple measure of unemployment news to proxy for risk aversion

 State‐level hy  hy = cointegrating residual between h and y  h = state housing wealth, y = state personal income  Measures the severity of borrowing constraints and the degree of risk

sharing across investors

Descriptive Statistics of State Portfolios

State‐Level Portfolios: Performance

State‐level portfolios = Value‐weighted portfolios of all firms headquartered in the state

AR 0.54 CT 0.04 UT -0.06 NH -0.13

MN 0.28 WI 0.03 OH -0.06 SC -0.15

CA 0.16 NV 0.02 MD -0.07 FL -0.16

WA 0.12 IA 0.01 MI -0.08 AZ -0.17

NC 0.10 VA 0.00 AL -0.08 LA -0.18

GA 0.09 KY -0.01 MA -0.10 OK -0.27

NY 0.08 IL -0.01 OR -0.11 KS -0.30

MO 0.04 IN -0.02 PA -0.12 CO -0.40

NJ 0.04 TN -0.03 TX -0.12

Average Monthly Characteristic Adjusted Returns (DGTW)

State‐Level Portfolios: Size

NY 1119.47 PA 196.58 IN 79.01 LA 24.38

CA 953.12 MA 189.66 CO 78.44 NV 24.26

TX 652.93 MN 185.90 TN 73.79 KY 22.55

NJ 452.92 VA 172.30 MD 61.15 IA 17.98

IL 440.44 NC 164.14 WI 58.74 SC 13.61

CT 290.25 MI 144.38 AZ 28.88 UT 12.26

GA 248.34 AR 114.70 AL 28.47 KS 8.77

OH 245.65 MO 103.70 OK 25.98 NH 6.37

WA 203.58 FL 85.16 OR 24.45

Average Monthly Firm Size ($m)

State‐Level Portfolios: B/M

NC 0.96 LA 0.82 MD 0.71 CO 0.61

IA 0.94 AL 0.78 NH 0.70 TX 0.61

IN 0.90 PA 0.76 IL 0.70 AZ 0.61

KS 0.87 VA 0.76 WA 0.68 CA 0.60

MI 0.87 OH 0.75 FL 0.68 NJ 0.58

GA 0.86 WI 0.73 MA 0.68 MO 0.56

KY 0.86 TN 0.73 UT 0.65 OK 0.56

AR 0.83 OR 0.72 NY 0.64 CT 0.35

SC 0.83 NV 0.71 MN 0.63

Average Monthly Firm Book-to-Market Ratio

Panel Predictive Regression

 Fixed effects regression model  Simple specification that can be used to form trading strategies even

when the look‐back period to estimate the model is short  We fix slopes across states because slopes from state‐by‐state

regressions are not statistically different from the common slopes estimated by the FE model

 Dependent variable Yj: Idiosyncratic component of portfolio return of state j  Our conjecture is that a component of the local return is affected by

local conditions  Residual returns from factor models (e.g., the CAPM, 4‐factor model)  Market‐, characteristic‐ , or industry‐adjusted returns

Main Predictors  Four state‐level predictors Xj

 “Recession” indicators: State income growth rate, relative unemployment rate, housing collateral ratio

 State‐level dividend/price ratio

 Seven U.S. predictors XUSA  U.S. income growth, U.S. hy, U.S. relative unemployment, U.S. cay, and

three spreads (paper‐bill, term: 10y bond – 1y bond, default: Baa – 10y bond)

 Account for known predictability at the aggregate stock market level

 Driscoll and Kraay standard errors  Account for cross‐sectional dependence, heteroscedasticity, and

autocorrelation

One‐Quarter Ahead Predictability Regressions

The regression results show that when REL UNEMP rises, investors become more risk averse, and future returns rise. When hy drops, harder to share income risks, variance of consumption growth rises, and future returns also rise.

(1) (2) (3) (4) (5) (6)

CAPM 4 Factor 7 Factor

State-Level Predictors

Inc Gr -0.099 -0.060 0.038 -0.105 0.029 0.017 -0.97 -0.91 0.47 -0.85 0.35 0.23

Rel Un 0.014 0.014 0.013 0.012 0.012 0.009 3.94 2.95 3.09 3.49 3.99 2.99

hy -0.032 -0.036 -0.034 -0.028 -0.031 -0.023 -2.66 -5.26 -3.66 -2.08 -2.98 -2.11

Residual R loc

Market Adj. Char. Adj. Ind. Adj.

Economic Significance of Predictability Regressions: Trading Strategies

Evaluation period: 1984 – 2011 Five steps 1. At the end of quarter q of year y, predict characteristic‐

adjusted return in quarter q+1 using data from 1980Q1 to year y quarter q (no look ahead bias)

2. Rank the states according to their predicted returns next quarter

3. Two portfolios: Long = top 3 states, Short = bottom 3 states 4. Hold long and short portfolios for 3 months 5. Rebalance at the end of quarter q+1 ….

Performance of Trading Strategies: Baseline Estimates

N s = 3

CAPM 4 FF Factors 4 FF Factors + 3 IND 4 FF Factors + 3 IND + 2

Reversal

Long 0.294 0.421 0.429 0.442 3.04 2.44 2.54 2.59

Short -0.360 -0.363 -0.381 -0.374 -2.13 -2.17 -2.46 -2.39

Long - Short 0.654 0.784 0.809 0.816 2.16 3.18 3.51 3.50

Alpha Estimates

Both long and short portfolios have significant excess returns. Excess performance of the long portfolio = 0.442 x 12 = 5.30% annually Short portfolio under‐performance = 0.374 x 12 = 4.49%. Performance differentials are economically significant.

Sub‐Sample Estimates

Trading strategy performance is robust across sub‐periods.

Sensitivity to Ns

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Number of States (Ns)

M on

th ly

9 -F

ac to

Long‐Short performance is robust to alternative definitions of long and short portfolios.

About 5% annualized alpha for Ns = 6

Retail Investor Local Bias Predictability must be stronger among stocks with a more local clientele.

Data: US discount brokerage data.

Period: Brokerage data are available for 1991 – 1996; Evaluation period for trading strategies is set to 1991 – 2004.

Two firm‐level local bias proxies  Average distance of the firm from all investors in the sample MINUS

average distance of the firm from its own shareholders.

 Average portfolio concentration of firm’s investors = portfolio variance / average variance of stocks in portfolio.

Evidence of “State” Bias

CA 44 PA 17 AZ 15 KS 11

WA 43 MI 17 OH 15 KY 10

MN 28 RI 17 UT 14 CO 10

TN 22 LA 17 NJ 14 NH 9

OK 21 NY 16 MO 13 WV 9

IL 21 FL 16 GA 13 NC 9

OR 21 MD 16 MA 13 NM 8

AR 19 VA 16 NV 12 IA 8

IN 19 SC 15 MS 12 NE 8

TX 18 MT 15 VT 12 ID 8

WI 18 AL 15 CT 12 ME 8

% of State-level Ownership by Local Investors of Local Firms

Conditioning on Retail Investor Local Bias

Trading strategy performance is statistically insignificant among stocks with low local bias and economically large among stocks with high local bias.

N s = 3

All Stocks

Low (bottom third) High (top third) Low (bottom third) High (top third)

Long 0.421 0.057 0.547 0.028 0.655 2.44 0.53 2.85 0.23 3.03

Short -0.363 -0.002 -0.385 0.036 -0.443 -2.17 -0.05 -2.69 0.11 -2.85

L - S 0.784 0.059 0.932 -0.008 1.098 3.18 0.32 2.96 -0.17 3.16

Distance-based LB Measure Portfolio Concentration

4 Factor Alpha Estimates

Summary 1. Main idea

 State recession indicators (unemployment and housing collateral) can predict the returns of state portfolios.

2. Main finding  During the 1980‐2004 period: Annual performance differential is over

7% on a risk‐adjusted basis. Stronger in high local bias regions. 3. Main “story”: time‐varying local discount rates.

 During local recessions, consumption smoothing becomes more difficult and risk sharing levels decline.

 This generates regional variation in risk aversion and predictable patterns in returns.

 Local bias amplifies these effects. 4. Main recommendation

 Buy stocks in depressed states, short stocks in booming states.

Investor Sophistication and Return Predictability

Basic Conjecture Local stock returns would be more predictable in regions

where investors are less sophisticated. Why?

Investors in more sophisticated states would be able to blunt the effects of local recessions more effectively.

In contrast, relatively less sophisticated investors may exhibit stronger behavioral biases and a lower ability to exploit financial assets to reduce their income risk.

Graphical Evidence

-0.05

-0.03

-0.01

0.01

0.03

0.05

Ju n-

79 A

ug -8

0 O

ct -8

1 D

ec -8

2 Fe

b- 84

A pr

-8 5

Ju n-

86 A

ug -8

7 O

ct -8

8 D

ec -8

9 Fe

b- 91

A pr

-9 2

Ju n-

93 A

ug -9

4 O

ct -9

5 D

ec -9

6 Fe

b- 98

A pr

-9 9

Ju n-

00 A

ug -0

1 O

ct -0

2 D

ec -0

3 Fe

b- 05

A pr

-0 6

Ju n-

07 A

ug -0

8 O

ct -0

Low Sophistication minus High Sophistication

Characteristic‐Adjusted Return Differential of Long‐Short Portfolios between Low and High Sophistication States.

Local stock returns are indeed more predictable in regions where investors are less sophisticated.

Political Sentiment and Returns

Politics and Financial Markets

In the second scenario, we will examine how the changing political environment affects financial markets. In particular, the political climate affects the optimism level of certain types of investors and through this channel it affects asset prices.

Key Takeaways 1. Political sensitivity of firms and industries varies and this can

be quantified as political betas. 2. A Long‐Short trading strategy based on political sensitivity

earns about 6‐8% per year (stronger in recent years); characteristic‐adjusted returns are about 6% per year.

3. Return predictability is not generated through the cash flow channel.More likely to be induced by investors’ demand shifts (i.e., political sentiment).

4. A significant part of momentum profits (about one‐third) can be attributed to under‐reaction to political information.

Politics and Markets How do financial markets respond to the changing political climate? There is heightened interest in the finance media.  Headlines: Market Reaction

 How will market react on day after election? (Marketwatch, 10/23/12).  Does the election matter to the markets, investors? (Fox Business,

10/25/12).

 Headlines: Portfolio Allocation  The winner for investors is... (WSJ, 10/22/12).  Should the election change your investment mix? (Marketwatch, 10/23/12).  How to position your post‐election portfolio (USA Today, 10/26/12).  Top performing timers react to election (Marketwatch, 11/07/12).

Our Story… Demand for certain types of risky assets could vary with

the political climate.  At any given point in time, Republicans and Democrats do

not have similar views about the markets and the overall economy.

 Their opinions vary with the political climate: people more optimistic about the economy when their preferred party is in power.

 Consequently, they may increase their holdings of risky assets (which they prefer).

 Some sophisticated investors could engage in hedging too.

Main Question and Key Findings

Does heterogeneity in political sentiment induce predictable patterns in certain segments of the market?  Yes! Strong evidence of predictability.  1939 – 2011 period: Long‐Short strategy earns 6‐8% per year;

characteristic‐adjusted returns are about 6% per year.  Return predictability is not generated through the cash flow

channel.  Predictability is stronger around elections and party

transitions.  Systematic demand shifts induced by changes in the political

climate generate a mispricing and correction pattern.

Evidence of Aggregate Predictability

Estimating Political Sensitivity

Political Sensitivity Based Portfolios

Politically Sensitive Industries

Annual Performance Estimates

Positive performance differential during most years.

Cumulative Value of $1

Green: Long Black: Market Red: Short Grey: Riskfree

Performance of Industry Portfolios

Monotonically increasing returns. Both raw and characteristic‐ adjusted returns exhibit similar patterns.

Factor Model Estimates

Results are similar when we use various factor models to adjust for risk.

Incumbent versus Challenger

Results are stronger when the challenger party wins.

Predictability Around Elections

Results are stronger during high attention periods (i.e., around Presidential elections).

Republicans or Democrats

Results are significant for both Republican and Democratic Presidential periods.

Economically Large Segment?

The extreme portfolios represent about 17‐27% of the aggregate market.

Demand‐Induced Predictability? Consistent with demand‐ induced mispricing and correction pattern. Cash flow channel predicts an opposite pattern.

Politics and Momentum

Main Question and Key Findings

Preliminary Evidence

Our measure of political sensitivity is strongly correlated with the recent performance of stocks.

Conditioning on Political Sensitivity

Politically inconsistent momentum strategy is not profitable.

Raw Returns

Characteristic‐Adjusted Returns

Performance of Momentum Strategy

Cumulative Value of $1

Summary and Conclusion Political sentiment generates predictable patterns in

stock returns.  Present a novel way of identifying firms that are more likely

to be influenced by the changing political climate (political betas).

 1939 – 2011 period: Long‐Short strategy earns 6‐8% per year; characteristic‐adjusted returns are about 6% per year.

 Return predictability is not generated through the cash flow channel.

 Systematic demand shifts generate a mispricing and correction pattern.

 A significant part of momentum profits can be attributed to under‐reaction to political information.

Geographic Dispersion and Return Predictability

Main Questions

 Do market participants aggregate geographically dispersed information efficiently?

 In particular, are equity analysts able to aggregate firms’ geographically dispersed information? How about sophisticated investors?

 Do potential frictions in the information aggregation process generate predictable patterns in stock returns?

Motivation

 Previous research shows that value‐relevant firm‐specific information is distributed geographically. E.g., stock returns comove more strongly with firms in economically connected states.

Further, managers are able to gather information from plants and operations more efficiently when those locations become more accessible following the introduction of airline routes (exogenous events).

Key Findings

 Future earnings and cash flows of firms can be predicted using current earnings and CF of other firms in both headquarters (HQ) and economically‐connected (EC) states.

 Equity analysts focus on the information in HQ‐states but not EC states. Even analysts located in EC states do not aggregate EC‐ state information.

 Return patterns indicate that geographically dispersed information is aggregated slowly. Can form profitable Long‐Short trading strategies.

How to Identify Firms’ Operations?

 Use firms’ 10‐K filings to come up with a proxy of geographical dispersion: Conduct textual analysis of information in Business, Properties, Consolidated Financial Data, and MD&A sections. Compute citation share (CS) of each state.

Count number of times references made to each U.S. state from 1994 to 2010.

Citation‐Share and Economic Connections

An example: Compute CS to capture economic connections:

State # Mentioned Citation‐Share

Texas 5 0.5

California 4 0.4

Illinois 3 0.3

New York 2 0.2

Florida 1 0.1

Total = 10 1.0

Trading Strategies

 To construct trading portfolios based on Expected Earnings Surprise for stocks with scheduled announcements in month t+1: (Fitted EPS – Analyst Consensus EPS)/Lag Price Fitted EPS is from baseline regressions.

 Value‐weight and hold for 3 months.

Performance Results

Alphas using Other Factor Models

Impact of Arbitrage Costs Institutional Ownership Idiosyncratic Volatility Low High Low High

Portfolio (1) (2) (3) (4) Long – Short (5 – 1) 0.735 0.486 0.439 0.512 (2.45) (2.33) (1.30) (2.51)

Long (5) 1.158 1.068 0.995 1.089 (2.59) (2.86) (1.54) (3.13) 4 0.739 0.765 0.594 0.856 (2.08) (1.68) (0.87) (2.52) 3 0.444 0.593 0.580 0.598 (0.88) (1.29) (0.87) (1.50) 2 0.477 0.644 0.654 0.624 (0.97) (1.36) (1.00) (1.59) Short (1) 0.422 0.582 0.556 0.577 (0.99) (1.27) (0.88) (1.52) Number of months 184 184 184 184

Mispricing and Correction

Overall…

Both HQ and EC states contain value‐relevant information about firms’ future performance.

Equity analysts are unable to aggregate this information; they focus on the information in HQ‐ states but not EC states.

There is gradual diffusion of geographic information, which generates predictable patterns in stock returns.

Conclusions  In three different settings, we saw that investor psychology

affects asset prices.  First, we saw the pricing effects of local investor sentiment.  Next, we examined the pricing effects of political sentiment.  Last, we studied how geographical dispersion of information

affects asset prices because investors are slow in aggregating this information.

 You can look at the reading list (and references there) for other examples.

Next week, we will study how corporate managers take advantage of potential inefficiencies in the market.