[Link].

net

Quantitative
Methods

2017CFA二级培训项目
讲师：周琪

1-90
[Link]

周琪
工作职称：金程教育金融研究院CFA/FRM高级培训师
教育背景：中央财经大学国际经济学学士、澳大利亚维多利亚大学金融风
险管理学学士
工作背景：学术功底深厚、培训经验丰富，曾任课AFP、CFP多年，参与教
学研究及授课，现为金程教育CFA/FRM双证培训老师，担任CFA项目学术
研发负责人，对CFA教学产品的研发工作负责，曾亲自参与中国工商银行总
行、中国银行总行、杭州联合银行等CFA、FRM培训项目。累计课时达400
0小时，课程清晰易懂，深受学员欢迎。
服务客户：中国工商银行、中国银行、中国建设银行、杭州联合银行、杭
州银行、国泰君安证券、深圳综合开发研究院、中国CFP标准委员会、太平
洋保险等
主编出版：参与金程CFA项目参考书目的编写工作，包括金程CFA一级中文
Notes等

2-90
[Link]

Topic Weightings in CFA Level II

Session NO. Content Weightings

Study Session 1-2 Ethics & Professional Standards 10-15

Study Session 3 Quantitative Methods 5-10

Study Session 4 Economic Analysis 5-10

Study Session 5-6 Financial Statement Analysis 15-20

Study Session 7-8 Corporate Finance 5-15

Study Session 9-11 Equity Analysis 15-25

Study Session 12-13 Fixed Income Analysis 10-20

Study Session 14 Derivative Investments 5-15

Study Session 15 Alternative Investments 5-10

Study Session 16-17 Portfolio Management 5-10

3-90
[Link]

 SS3 Quantitative Methods for

Framework Valuation
• R9 Correlation and
Quantitative Methods regression
• R10 Multiple regression and
issues in regression analysis
• R11 Time-series analysis
• R12 Excerpt
from ’’Probabilistic
Approaches: Scenario
Analysis, Decision Trees, and
Simulation’’

4-90
[Link]

Reading
9

Correlation and regression

5-90
[Link]

1. Scatter Plots
2. Covariance and Correlation
Framework 3.
4.
Interpretations of Correlation Coefficients
Significance Test of the Correlation
5. Limitations to Correlation Analysis
6. The Basics of Simple Linear Regression
7. Interpretation of regression coefficients
8. Standard Error of Estimate & Coefficient of
Determination (R2)
9. Analysis of Variance (ANOVA)
10. Regression coefficient confidence interval
11. Hypothesis Testing about the Regression
Coefficient
12. Predicted Value of the Dependent Variable
13. Limitations of Regression Analysis

6-90
[Link]

Scatter Plots
 A scatter plots is a graph that shows the relationship between the
observations for two data series in two dimensions.

7-90
[Link]

Sample Covariance and Correlation

 Covariance:
 Covariance measures how one random variable moves with
another random variable. ----It captures the linear relationship.

n
 Cov( X , Y )  (X
i 1
i  X )(Yi  Y ) /( n  1)

 Covariance ranges from negative infinity to positive infinity

Cov( X , Y )
 Correlation: r 
sx s y
 Correlation measures the linear relationship between two
random variables
 Correlation has no units, ranges from –1 to +1

8-90
[Link]

Interpretations of Correlation Coefficients

 The correlation coefficient is a measure of linear association.
 It is a simple number with no unit of measurement attached, so the
correlation coefficient is much easier to explain than the covariance.

Correlation coefficient Interpretation

r = +1 perfect positive correlation
0 < r < +1 positive linear correlation
r=0 no linear correlation
−1 < r < 0 negative linear correlation
r = −1 perfect negative correlation

9-90
[Link]

Interpretations of Correlation Coefficients

10-90
[Link]

Significance Test of the Correlation

 Test whether the correlation between the population of two variables is
equal to zero.
 H0: ρ=0; HA: ρ≠0 (Two-tailed test)
 Test statistic:
r-0 r n-2
t=  , df = n-2
2 2
1-r 1-r
n-2
 Decision rule: reject H0 if t>+t critical, or t<- t critical
 Conclusion: the correlation between the population of two variables is
significantly different from zero.

11-90
[Link]

Example
 The covariance between X and Y is 16. The standard deviation of X is 4
and the standard deviation of Y is 8. The sample size is 20. Test the
significance of the correlation coefficient at the 5% significance level.

 Solution :
 The sample correlation coefficient r = 16/(4×8) = 0.5. The t-
statistic can be computed as: 20  2
t  0.5   2.45
1  0.25
The critical t-value for α=5%, two-tailed test with df=18 is 2.101.
Since the test statistic of 2.45 is larger than the critical value of
2.101, we have sufficient evidence to reject the null hypothesis. So
we can say that the correlation coefficient between X and Y is
significantly different from zero.

12-90
[Link]

Limitations to Correlation Analysis

 Outliers
 Outliers represent a few extreme values for sample observations.
Relative to the rest of the sample data, the value of an outlier may be
extraordinarily large or small.
 Outlier can result in apparent statistical evidence that a significant
relationship exists when, in fact, there is none, or that there is no
relationship when, in fact, there is a relationship.

13-90
[Link]

Limitations to Correlation Analysis

 Spurious correlation
 Spurious correlation refers to the appearance of a causal linear relationship
when, in fact, there is no relation. Certain data items may be highly
correlated purely by chance. That is to say, there is no economic explanation
for the relationship, which would be considered a spurious correlation.
 correlation between two variables that reflects chance relationships in a
particular data set,
 correlation induced by a calculation that mixes each of two variables
with a third (two variables that are uncorrelated may be correlated if
divided by a third variable,
 correlation between two variables arising not from a direct relation
between them but from their relation to a third variable. (height may be
positively correlated with the extent of a person's vocabulary)

14-90
[Link]

Limitations to Correlation Analysis

 Nonlinear relationships
 Correlation only measures the linear relationship between two variables,
so it dose not capture strong nonlinear relationships between variables.
 For example, two variables could have a nonlinear relationship such as
Y= (1-X) 3 and the correlation coefficient would be close to zero, which is
a limitation of correlation analysis.

15-90
[Link]

The Basics of Simple Linear Regression

 Linear regression allows you to use one variable to make predictions
about another, test hypotheses about the relation between two
variables, and quantify the strength of the relationship between the
two variables.
 Linear regression assumes a linear relation between the dependent
and the independent variables.
 The dependent variable is the variable whose variation is
explained by the independent variable. The dependent variable
is also refer to as the explained variable, the endogenous
variable, or the predicted variable.
 The independent variable is the variable whose variation is used
to explain the variation of the dependent variable. The
independent variable is also refer to as the explanatory variable,
the exogenous variable, or the predicting variable.

16-90
[Link]

The Basics of Simple Linear Regression

 The simple linear regression model

Yi  b0  b1 X i   i , i  1,..., n
 Where
Yi = ith observation of the dependent variable, Y
Xi = ith observation of the independent variable, X
b0 = regression intercept term
b1 = regression slope coefficient
εi= the residual for the ith observation (also referred to as the disturbance
term or error term)

17-90
[Link]

Interpretation of regression coefficients

 Interpretation of regression coefficients
 The estimated intercept coefficient ( b̂0 ) is interpreted as the
value of Y when X is equal to zero.
 The estimated slope coefficient ( b̂1 ) defines the sensitivity of
Y to a change in X .The estimated slope coefficient ( b̂1 ) equals
covariance divided by variance of X.
 Example
 An estimated slope coefficient of 2 would indicate that the
dependent variable will change two units for every 1 unit change
in the independent variable.
 The intercept term of 2% can be interpreted to mean that the
independent variable is zero, the dependent variable is 2%.

18-90
[Link]

The assumptions of the linear regression

 The assumptions
 A linear relationship exists between X and Y
 X is not random, and the condition that X is uncorrelated with the error
term can substitute the condition that X is not random.
 The expected value of the error term is zero (i.e., E(εi)=0 )
 The variance of the error term is constant (i.e., the error terms are
homoskedastic)
 The error term is uncorrelated across observations (i.e., E(εiεj)=0 for all
i≠j)
 The error term is normally distributed.

19-90
[Link]

Calculation of regression coefficients

 Ordinary least squares (OLS)
 OLS estimation is a process that estimates the population parameters Bi
with corresponding values for bi that minimize the squared residuals (i.e.,
error terms).
 the OLS sample coefficients are those that:
n

Cov( X , Y )
(X i  X )(Yi  Y )
b1   i 1
n b0  Y  b1 X
(X
Var ( X )
i  X) 2

i 1

 The estimated intercept coefficient ( b̂0) : the point ( X , Y ) is on the

regression line.

20-90
[Link]

Example: calculate a regression coefficient

 Bouvier Co. is a Canadian company that sells forestry products to
several Pacific Rim customers. Bouvier’s sales are very sensitive to
exchange rates. The following table shows recent annual sales (in
millions of Canadian dollars) and the average exchange rate for the
year (expressed as the units of foreign currency needed to buy one
Canadian dollar).
Year i Xi = Exchange Rate Yi = Sales
1 0.40 20
2 0.36 25
3 0.42 16
4 0.31 30
5 0.33 35
6 0.34 30
 Calculate the intercept and coefficient for an estimated linear
regression with the exchange rate as the independent variable and
sales as the dependent variable.

21-90
[Link]

Example: calculate a regression coefficient

 The following table provides several useful calculations:
Year i Xi = Exchange Rate Yi = Sales (Xi -X)2 (Yi -Y)2 (Xi -X)(Yi -Y)
1 0.4 20 0.0016 36 -0.24
2 0.36 25 0 1 0
3 0.42 16 0.0036 100 -0.6
4 0.31 30 0.0025 16 -0.2
5 0.33 35 0.0009 81 -0.27
6 0.34 30 0.0004 16 -0.08
Sum 2.16 156 0.009 250 -1.39

22-90
[Link]

Example: calculate a regression coefficient

 The sample mean of the exchange rate is:
n
X   X i / n  2.16 / 6  0.36
i 1

 The sample mean of sales is:

n
Y   Yi / n  156 / 6  26
i 1
 We want to estimate a regression equation of the form Yi = b0 + b1Xi
+εi. Then
estimates of the slope coefficient and the intercept are
  Yi -Y  Xi -X  -1.39
bˆ 1 = i=1 n = = -154.44, and
  Xi -X 
2 0.009
i=1

bˆ 0  Y  bˆ 1 X  26   154.444  0.36   26  55.6  81.6

 So the regression equation is Yi = 81.6 – 154.444Xi

23-90
[Link]

Analysis of Variance(ANOVA) Table

 Components

Y
 
Yi  b0  b1 X i (Yi  Yi )  SSE
__

 _
(Yi  Y )  SST
(Yi  Y )  RSS
__
Y

b0
X

24-90
[Link]

ANOVA Table
 ANOVA Table

df SS MSS
Regression k=1 RSS MSR=RSS/k
Error n-2 SSE MSE=SSE/(n-2)
Total n-1 SST -
SSE
 Standard error of estimate: SEE  n2
 MSE

 Coefficient of determination (R²)

 RSS SSE
R2   1
SST SST
explained variation unexplained variation
 =1-
total variation total variation

25-90
[Link]

Standard Error of Estimate

 Standard Error of Estimate (SEE) measures the degree of variability of the
actual Y-values relative to the estimated Y-values from a regression equation.
 SEE will be low (relative to total variability) if the relationship is very strong
and high if the relationship is weak.
 The SEE gauges the “fit” of the regression line. The smaller the standard
error, the better the fit.
 The SEE is the standard deviation of the error terms in the regression.

SSE
SEE   MSE
n2

26-90
[Link]

Coefficient Determination (R2)

 A measure of the “goodness of fit” of the regression. It is interpreted as a
percentage of variation in the dependent variable explained by the
independent variable. Its limits are 0≤R2≤1.
 Example: R2 of 0.63 indicates that the variation of the independent
variable explains 63% of the variation in the dependent variable.
 For simple linear regression, R² is equal to the squared correlation
coefficient (i.e., R² = r² )
 The Different between the R2 and Correlation Coefficient
 The correlation coefficient indicates the sign of the relationship between
two variables, whereas the coefficient of determination does not.
 The coefficient of determination can apply to an equation with several
independent variables, and it implies a explanatory power, while the
correlation coefficient only applies to two variables and does not imply
explanatory between the variables.

27-90
[Link]

Example
 An analyst ran a regression and got the following result:
Coefficient t-statistic p-value
Intercept -0.5 -0.91 0.18
Slope 2 12.00 <0.001

ANOVA Table df SS MSS

Regression 1 8000 ?
Error ? 2000 ?
Total 51 ? -
 Fill in the blanks of the ANOVA Table.
 What is the standard error of estimate?
 What is the result of the slope coefficient significance test?
 What is the result of the sample correlation?
 What is the 95% confidence interval of the slope coefficient?

28-90
[Link]

Regression coefficient confidence interval

 Regression coefficient confidence interval
bˆ1  t c sbˆ
1
 If the confidence interval at the desired level of significance dose not
include zero, the null is rejected, and the coefficient is said to be statistically
different from zero.
 sb̂ is the standard error of the regression coefficient.
1

 As SEE rises, sb̂ also increases, and the confidence interval widens
1
because SEE measures the variability of the data about the regression
line, and the more variable the data, the less confidence there is in the
regression model to estimate a coefficient.

29-90
[Link]

Hypothesis Testing about Regression Coefficient

 Significance test for a regression coefficient
 H0: b1=The hypothesized value(usually 0)
 Test Statistic:
bˆ1  b1
t , df=n-2
sbˆ
1

 Decision rule: reject H0 if +t critical <t, or t<- t critical

 Rejection of the null means that the slope coefficient is different from
the hypothesized value of b1.

30-90
[Link]

Predicted Value of the Dependent Variable

 Predicted values are values of the dependent variable based on the
estimated regression coefficients and a prediction about the value of the
independent variable.
 Point estimate
Yˆ  bˆ0  bˆ1 X '
 Confidence interval estimate
Yˆ  t c  s f 
t c = the critical t-value with df=n−2
sf = the standard error of the forecast
1 ( X '  X )2 1 ( X '  X )2
s f  SEE  1    SEE  1  
n (n  1) s X
2
n  ( X i  X )2

31-90
[Link]

Limitations of Regression Analysis

 Regression relations change over time
 This means that the estimation equation based on data from a specific
time period may not be relevant for forecasts or predictions in another
time period. This is referred to as parameter instability.
 The usefulness will be limited if others are also aware of and act on the
relationship.
 Regression assumptions are violated
 For example, the regression assumptions are violated if the data is
heteroskedastic (non-constant variance of the error terms) or exhibits
autocorrelation (error terms are not independent).

32-90
[Link]

Reading
10

Multiple regression and issues in regression analysis

33-90
[Link]

1. The Basics of Multiple Regression

Framework
2. Interpreting the Multiple Regression
Results
3. Hypothesis Testing about the Regression
Coefficient
4. Regression Coefficient F-test
5. Coefficient of Determination (R2)
6. Analysis of Variance (ANOVA)
7. Dummy variables
8. Multiple Regression Assumptions
9. Multiple Regression Assumption
Violations
10. Model Misspecification
11. Qualitative Dependent Variables

34-90
[Link]

The Basics of Multiple Regression

 Multiple regression is regression analysis with more than one independent
variable
 The multiple linear regression model
Yi  b0  b1 X 1i  b2 X 2i    bk X ki   i
Xij = ith observation of the jth independent variable
N = number of observation
K = number of independent variables
 Predicted value of the dependent variable

Yˆ  bˆ0  bˆ1 Xˆ 1  bˆ2 Xˆ 2    bˆk Xˆ k

35-90
[Link]

Interpreting the Multiple Regression Results

 The intercept term is the value of the dependent variable when the
independent variables are all equal to zero.

 Each slope coefficient is the estimated change in the dependent variable for
a one unit change in that independent variable, holding the other
independent variables constant. That’s why the slope coefficients in a
multiple regression are sometimes called partial slope coefficient.

36-90
[Link]

Multiple Regression Assumptions

 The assumptions of the multiple linear regression
 A linear relationship exists between the dependent and independent
variables
 The independent variables are not random ( OR X is not correlated with
error terms). There is no exact linear relation between any two or more
independent variables
 The expected value of the error term is zero (i.e., E(εi)=0 )
 The variance of the error term is constant (i.e., the error terms are
homoskedastic)
 The error term is uncorrelated across observations (i.e., E(εiεj)=0 for all
i≠j)
 The error term is normally distributed.

37-90
[Link]

Dummy variables
 To use qualitative variables as independent variables in a regression

 The qualitative variable can only take on two values, 0 and 1

 If we want to distinguish between n categories, we need n−1 dummy

variables

38-90
[Link]

Dummy variables
 Interpreting the coefficients
 Example: EPSt = b0 + b1Q1t + b2Q2t + b3Q3t + ϵ
 EPSt = a quarterly observation of earnings
per share y x1 x2 x3
 Q1t =1 if period t is the first quarter, Q1t
EPSt Q1 Q2 Q3
=0 otherwise
EPS09Q4 0 0 0
 Q2t =1 if period t is the second quarter, Q2t
=0 otherwise EPS09Q3 0 0 1

 Q3t =1 if period t is the third quarter, Q3t EPS09Q2 0 1 0

=0 otherwise EPS09Q1 1 0 0
 The intercept term, represents the average value EPS08Q4 0 0 0
of EPS for the fourth quarter. EPS08Q3 0 0 1
 The slope coefficient on each dummy variable EPS08Q2 0 1 0
estimates the difference in earnings per share
EPS08Q1 1 0 0
(on average) between the respective quarter (i.e.,
quarter 1, 2, or 3) and the omitted quarter (the … … … …
fourth quarter in this case).

39-90
[Link]

Analysis of Variance (ANOVA)

 ANOVA Table
df SS MSS
Regression k RSS MSR=RSS/k
Error n-k-1 SSE MSE=SSE/(n-k-1)
Total n-1 SST -
 Standard error of estimate
SSE
SEE   MSE
n  k 1
 Coefficient of determination (R²)
RSS SSE
 Is R2 still reliable? R2   1
SST SST

40-90
[Link]

Adjusted R2
 R2 and adjusted R2
 R2 by itself may not be a reliable measure of the explanatory power of
the multiple regression model. This is because R2 almost always
increases as variables are added to the model, even if the marginal
contribution of the new variables is not statistically significant.
 Function of adjusted R2
SSE n  k  1  n  1  2 
adjusted R 2  1 
SST n  1
 1     1  R  
 n  k  1  
 adjusted R² ≤ R²
 adjusted R² may be less than zero

41-90
[Link]

Hypothesis Testing about Regression Coefficient

 Significance test for a regression coefficient
 H0: bj=0
bˆ j
 Test statistic: t  df = n-k-1
sbˆ
 p-value: the smallest significance level for which the null hypothesis
j

can be rejected
 Reject H0 if p-value<α
 Fail to reject H0 if p-value>α
 Regression coefficient confidence interval
j 
 bˆ  t  s ˆc bj

 Estimated regression coefficient ±(critical t-value) (coefficient standard
error)

42-90
[Link]

Regression Coefficient F-test

 An F-statistic assesses how well the set of independent variables, as a group,
explains the variation in the dependent variable.
 An F-test is used to test whether at least one slope coefficient is significantly
different from zero
 Define hypothesis:
 H0: b1= b2= b3= … = bk=0
 HA: at least one bj≠0 (j = 1 to k)
 F-statistic:
SSR
MSR k
F 
MSE SSE
(n  k  1)

43-90
[Link]

Regression Coefficient F-test

 Decision rule
 Reject H0 : if F (test-statistic) > F c (critical value)
 Rejection of the null hypothesis at a stated level of significance indicates
that at least one of the coefficients is significantly different than zero,
which is interpreted to mean that at least one of the independent
variables in the regression model makes a significant contribution on
the explanation of the dependent variable.
 The F-test here is always a one-tailed test.
 The test assesses the effectiveness of the model as a whole in explaining the
dependent variable

44-90
[Link]

Unbiased and consistent estimator

 An unbiased estimator is one for which the expected value of the estimator
is equal to the parameter you are trying to estimate.

 If not, called as unreliable.

 A consistent estimator is one for which the accuracy of the parameter

estimate increases as the sample size increases.

45-90
[Link]

Multiple Regression Assumption Violations

 Heteroskedasticity 异方差
 Heteroskedasticity refers to the situation that the variance of the error
term is not constant (i.e., the error terms are not homoskedastic)
 Unconditional heteroskedasticity occurs when the heteroskedasticity is
not related to the level of the independent variables, which means that it
dose not systematically increase or decrease with the change in the
value of the independent variables. It usually causes no major problems
with the regression.
 Conditional heteroskedasticity is heteroskedasticity, that is, variance of
error term is related to the level of the independent variables.
 Conditional heteroskedasticity dose create significant problems for
statistical inference.

46-90
[Link]

Multiple Regression Assumption Violations

 Effect of Heteroskedasticity on Regression Analysis
 Not affect the consistency of regression parameter estimators
 Consistency: the larger the number of sample, the lower probability
of error.
 The coefficient estimates (the bˆ j ) are not affected.
 The standard errors are usually unreliable estimates.
 If the standard errors are too small, but the coefficient estimates
themselves are not affected, the t-statistics will be too large and the
null hypothesis of no statistical significance is rejected too often (一
类错误).
 The opposite will be true if the standard errors are too large. (二类错
误)
 The F-test is also unreliable.

47-90
[Link]

Multiple Regression Assumption Violations

 Detecting Heteroskedasticity
 Two methods to detect heteroskedasticity
 residual scatter plots (residual vs. independent variable)
 the Breusch-Pagen χ² test
H0: No heteroskedasticity, one-tailed test
Chi-square test: BP = n×Rresidual², df=k
 注意：以误差项squred residuals和X做回归，Rresidual²是此回
归的决定系数
Decision rule: BP test statistic should be small (χ²分布表）
 Correcting heteroskedasticity
 robust standard errors (also called White-corrected standard errors)
 generalized least squares

48-90
[Link]

Multiple Regression Assumption Violations

 Serial correlation (autocorrelation)序列相关，自相关
 Serial correlation (autocorrelation) refers to the situation that the error
terms are correlated with one another
 Serial correlation is often found in time series data
 Positive serial correlation exists when a positive regression error in
one time period increases the probability of observing regression
error for the next time period.
 Negative serial correlation occurs when a positive error in one
period increases the probability of observing a negative error in the
next period.

49-90
[Link]

Multiple Regression Assumption Violations

 Effect of Serial correlation on Regression Analysis
 Positive serial correlation → Type I error & F-test unreliable
 Not affect the consistency of estimated regression coefficients.
 Because of the tendency of the data to cluster together from
observation to observation, positive serial correlation typically
results in coefficient standard errors that are too small, which will
cause the computed t-statistics to be larger.
 Positive serial correlation is much more common in economic and
financial data, so we focus our attention on its effects.
 Negative serial correlation → Type II error
 Because of the tendency of the data to diverge from observation to
observation, negative serial correlation typically causes the standard
errors that are too large, which leads to the computed t-statistics
too small.

50-90
[Link]

Multiple Regression Assumption Violations

 Detecting Serial correlation
 Two methods to detect serial correlation
 residual scatter plots
 the Durbin-Watson test
H0: No serial correlation
DW ≈ 2×(1−r)
Decision rule
Reject H0, Reject H0,
conclude conclude
positive serial Do not negative serial
Inconclusive Inconclusive
correlation reject H0 correlation
0 d1 dU 4-dU 4-d1 4

51-90
[Link]

Multiple Regression Assumption Violations

 Detecting Serial correlation
 Two methods to detect serial correlation
 residual scatter plots
 the Durbin-Watson test
H0: No positive serial correlation
DW ≈ 2×(1−r)
Decision rule

Reject H0,
conclude
positive serial Inconclusive Fail to reject null hypothesis of no
correlation positive serial correlation

0 d1 dU

52-90
[Link]

Multiple Regression Assumption Violations

 Methods to Correct Serial correlation
 adjusting the coefficient standard errors (e.g., Hansen method): the
Hansen method also corrects for conditional heteroskedaticity.
 The White-corrected standard errors are preferred if only
heteroskedasticity is a problem.
 Improve the specification of the model: The best way to do this is to
explicitly incorporate the time-series nature of the data (e.g., include a
seasonal term).

53-90
[Link]

Multiple Regression Assumption Violations

 Multicollinearity
 Multicollinearity refers to the situation that two or more
independent variables are highly correlated with each other
 In practice, multicollinearity is often a matter of degree rather
than of absence or presence.
 Two methods to detect multicollinearity
 t-tests indicate that none of the individual coefficients is
significantly different than zero, while the F-test indicates
overall significance and the R² is high.
 the absolute value of the sample correlation between any
two independent variables is greater than 0.7 (i.e., ︱r︱>0.7).
 Methods to correct multicollinearity: omit one or more of the
correlated independent variables.

54-90
[Link]

Summary of assumption violations

Assumption
Impact Detection Solution
violation
① Residual scatter plots ①robust standard errors
Conditional
Type I ② Breusch-Pagen χ²-test (White-corrected standard
Heteroskeda
/II error (BP = n× R²) errors)
sticity
② generalized least squares
① Residual scatter plots ①robust standard errors
Positive
Type I ② Durbin-Watson test (Hansen method)
serial
error (DW≈2× (1−r)) ②Improve the specification
correlation
of the model
① t-tests: fail to reject H0; ①Remove one or more
Type II independent variables
Multicollinea error F-test: reject H0; R² is high
rity ② High correlation among
independent variables

55-90
[Link]

Model Misspecification
 There are three broad categories of model misspecification, or ways in which
the regression model can be specified incorrectly, each with several
subcategories:
 1. The functional form can be misspecified.
 Important variables are omitted.
 Variables should be transformed.
 Data is improperly pooled.
 2. Time series misspecification. (Explanatory variables are correlated with
the error term in time series models.)
 A lagged dependent variable is used as an independent variable
with serially correlated errors.
 A function of the dependent variable is used as an independent
variable ("forecasting the past").
 Independent variables are measured with error.
 3. Other time-series misspecifications that result in nonstationarity.
 Effects of the model misspecification: regression coefficients are biased
and/or inconsistent

56-90
[Link]

Qualitative Dependent Variables

 Qualitative dependent variable is a dummy variable that takes on a
value of either zero or one
 Probit and logit model: Application of these models results in estimates
of the probability that the event occurs (e.g., probability of default).
 A probit model based on the normal distribution, while a logit
model is based on the logistic distribution.
 Both models must be estimated using maximum likelihood methods
（极大似然估计）.
 These coefficients relate the independent variables to the likelihood
of an event occurring, such as a merger, bankruptcy, or default.
 Discriminant models yields a linear function, similar to a regression
equation, which can then be used to create an overall score, or ranking,
for an observation. Based on the score, an observation can be classified
into the bankrupt or not bankrupt category.

57-90
[Link]

Credit Analysis
 Z – score
Z = 1.2 A + 1.4 B + 3.3 C + 0.6 D + 1.0 E
Where:
A = WC / TA
B = RE / TA
C = EBIT / TA
D = MV of Equity / BV of Debt
E = Revenue / TA
 If Z<1.8  Bankruptcy

58-90
[Link]

Reading
11

Time-series analysis

59-90
[Link]

1. Trend Models

Framework 2. Autoregressive Models (AR)

3. Random Walks
4. Autoregressive Conditional
Heteroskedasticity (ARCH)
5. Regression with More Than One Time
Series
6. Steps in Time-Series Forecasting

60-90
[Link]

Trend Models
 Linear trend model
 yt=b0+b1t+εt

 Same as linear regression, except for that the independent variable is

time t (t=1, 2, 3, ……)

yt

61-90
[Link]

Trend Models
 Log-linear trend model
 yt=e(b0+b1t)
 Ln(yt ) =b0+b1t+εt
 Model the natural log of the series using a linear trend
 Use the Durbin Watson statistic to detect autocorrelation

62-90
[Link]

Trend Models
 Factors that Determine Which Model is Best
 A linear trend model may be appropriate if the data points appear to be
equally distributed above and below the regression line (inflation rate
data).
 A log-linear model may be more appropriate if the data plots with a
non-linear (curved) shape, then the residuals from a linear trend model
will be persistently positive or negative for a period of time (stock
indices and stock prices).
 Limitations of Trend Model
 Usually the time series data exhibit serial correlation, which means that
the model is not appropriate for the time series, causing inconsistent b0
and b1
 The mean and variance of the time series changes over time.

63-90
[Link]

Autoregressive Models (AR)

 An autoregressive model uses past values of dependent variables as
independent variables
 AR(p) model
xt  b0  b1 xt 1  b2 xt 2  ...  bp xt  p   t
 AR (p): AR model of order p (p indicates the number of lagged values
that the autoregressive model will include).
 For example, a model with two lags is referred to as a second-order
autoregressive model or an AR (2) model.

64-90
[Link]

Autoregressive Models (AR)

 Forecasting With an Autoregressive Model
 Chain rule of forecasting
 A one-period-ahead forecast for an AR (1) model is determined in the
following manner:
  
xt 1  b0  b1 xt
 Likewise, a two-step-ahead forecast for an AR (1) model is calculated as:

  
xt 2  b0  b1 xt 1

65-90
[Link]

Autoregressive Models (AR)

 Forecasting With an Autoregressive Model, we should prove:

 No autocorrelation

 Covariance-stationary series

 No Conditional Heteroskedasticity

66-90
[Link]

Autocorrelation
 Autocorrelation in an AR model
 Whenever we refer to autocorrelation without qualification, we mean
autocorrelation of time series itself rxt , xt k rather than autocorrelation of
the error term r t , t k .
 Detecting autocorrelation in an AR model
 Compute the autocorrelations of the residual
 t-tests to see whether the residual autocorrelations differ significantly
from 0,
r t , t k -0 r t , t k
t  statistics  =
Sr 1/ n
 If the residual autocorrelations differ significantly from 0, the model is
not correctly specified, so we may need to modify it (e.g. seasonality)
 Correction: add lagged values

67-90
[Link]

Autocorrelation
 Seasonality – a special question
 Time series shows regular patterns of movement within the year
 The seasonal autocorrelation of the residual will differ significantly from
0
 We should uses a seasonal lag in an AR model
 For example: xt=b0+b1 xt-1+ b2 xt-4+εt

68-90
[Link]

Example
 Suppose we decide to use an autoregressive model with a seasonal lag
because of the seasonal autocorrelation in the previous problem. We
are modeling quarterly data, so we estimate Equation:
 (ln Salest – ln Salest–1) = b0 + b1(ln Salest–1 – ln Salest–2) + b2(ln
Salest–4 – ln Salest–5) + εt.
 Using the information in Table 1, determine if the model is correctly
specified.
 If sales grew by 1 percent last quarter and by 2 percent four
quarters ago, use the model to predict the sales growth for this
quarter.

69-90
[Link]

Table [Link] Differenced Sales

 Table 1 Regression Statistics
R-squared 0.4220
Standard error 0.0318
Observations 68
Durbin–Watson 1.8784
Coefficient Standard Error t-Statistic
Intercept 0.0121 0.0053 2.3055
Lag 1 –0.0839 0.0958 –0.8757
Lag 4 0.6292 0.0958 6.5693
Autocorrelations of the Residual
Lag Autocorrelation Standard Error t-Statistic
1 0.0572 0.1213 0.4720
2 –0.0700 0.1213 –0.5771
3 0.0065 0.1213 –0.0532
4 –0.0368 0.1213 –0.3033
70-90
[Link]

Example
 Answer
 At the 0.05 significance level, with 68 observations and three
parameters, this model has 65 degrees of freedom. The critical
value of the t-statistic needed to reject the null hypothesis is thus
about 2.0. The absolute value of the t-statistic for each
autocorrelation is below 0.60 (less than 2.0), so we cannot reject
the null hypothesis that each autocorrelation is not significantly
different from 0. We have determined that the model is correctly
specified.
 If sales grew by 1 percent last quarter and by 2 percent four
quarters ago, then the model predicts that sales growth this
quarter will be 0.0121 – 0.0839 ln(1.01) + 0.6292 ln(1.02) = e0.02372 –
1 = 2.40%.

71-90
[Link]

Covariance-stationary
 Covariance-stationary series
 Statistical inference based on OLS estimates for a lagged time series
model assumes that the time series is covariance stationary.
 Three conditions for covariance stationary
 Constant and finite expected value of the time series
 Constant and finite variance of the time series
 Constant and finite covariance with leading or lagged values
 Stationary in the past does not guarantee stationary in the future
 All covariance-stationary time series have a finite mean-reverting level.

72-90
[Link]

Covariance-stationary
时间序列从长期来看，往往都带有均值回归
 Mean reversion 值时，下一个阶段数值会倾向于减小；而当小
大。以自回归
 A time series exhibits mean reversion if it AR(1)模型来看， xt  b0
has a tendency to move  b1 xt 1
towards its mean
b0 b
 mean reverting level is: xt 
For an AR(1) model, the程，就可以得到均值 。则当xt 
1  b1 1
b0
 If xt  the model predicts that
b0 x t+1 will be lower
b0 than x t,
(1  b1 ) 接近于 ；当 xt  时，AR(1)模型预
b0
1  b1 1  b1
and if xt  the model predicts that x t+1 will be higher than x t
(1  b1 )
案例 5-5，Mean-reverting level
Suppose a one-lag autoregressive model by xt  b0
16.54 and 0.65 respectively. If currently X is 42.5，
Referenced Answer
b0 16 .54
73-90
Mean Reverting level =   47 .26
1  b1 1  0.65
[Link]

Covariance-stationary
 Instability of regression coefficients

 Financial and economic relationships are dynamic

 Models estimated with shorter time series are usually more stable than
those with longer time series

 So we need to check Covariance stationary

74-90
[Link]

Random Walks
 Random walk
 Random walk without a drift
 Simple random walk: xt =xt-1+εt (b0=0 and b1=1)
 The best forecast of xt is xt-1
 Random walk with a drift
 xt=b0+xt-1+εt (b0≠0, b1=1)
 The time series is expected to increase/decrease by a constant
amount
 Features
 A random walk has an undefined mean reverting level
 A time series must have a finite mean reverting level to be covariance
stationary
 A random walk, with or without a drift, is not covariance stationary

75-90
[Link]

Unit root test

 The unit root test of nonstationarity
 The time series is said to have a unit root if the lag coefficient is equal to
one
 A common t-test of the hypothesis that b1=1 is invalid to test the unit
root, however, it is not often the case.
 Dickey-Fuller test (DF test) to test the unit root
 Start with an AR(1) model xt=b0+b1 xt-1+εt
Subtract xt-1 from both sides xt-xt-1 =b0+(b1 –1) xt-1+εt
xt-xt-1 =b0+g xt-1+εt
 H0: g=0 (has a unit root and is nonstationary) Ha: g<0 (does
not have a unit root and is stationary)
 Calculate conventional t-statistic and use revised t-table
 If we can reject the null, the time series does not have a unit root
and is stationary.
76-90
[Link]

Unit root correction

 If a time series appears to have a unit root
 One method that is often successful is to first-difference the time series
(as discussed previously) and try to model the first-differenced series as
an autoregressive time series.
 First differencing
 Define yt as yt = xt - xt-1 =εt

 This is an AR(1) model yt = b0 + b1 yt-1 +εt , where b0=b1=0

 The first-differenced variable yt is covariance stationary

77-90
[Link]

Autoregressive Conditional Heteroskedasticity

 Heteroskedasticity refers to the situation that the variance of the error term
is not constant. 多元回归中用BP test

 Test whether a time series is ARCH(1)

  t2  a0  a1 t21  ut
 If the coefficient a1 is significantly different from 0, the time series is
ARCH(1), If a time-series model has ARCH(1) errors, then the variance of
the errors in period t + 1 can be predicted in period t.

 If ARCH exists,

 the standard errors for the regression parameters will not be correct.
Generalized least squares must be used to develop a predictive model.

 we can predict the variance of the errors if we have it modeled.

78-90
[Link]

Compare forecasting power with RMSE

 Comparing forecasting model performance
 In-sample forecasts are within the range of data (i.e., time period) used
to estimate the model, which for a time series is known as the sample or
test period.
 Out-of-sample forecasts are made outside. In other words, we compare
how accurate a model is in forecasting the y variable value for a time
period outside the period used to develop the model.
 Root mean squared error (RMSE): the model with the smallest RMSE
is most accurate for out-of-sample

79-90
[Link]

Regression with More Than One Time Series

 In linear regression, if any time series contains a unit root, OLS may be
invalid
 Use DF tests for each of the time series to detect unit root, we will have
3 possible scenarios
 None of the time series has a unit root: we can use multiple regression
 At least one time series has a unit root while at least one time series
does not: we cannot use multiple regression
 Each time series has a unit root: we need to establish whether the time
series are cointegrated.
 If conintegrated, can estimate the long-term relation between the
two series (but may not be the best model of the short-term
relationship between the two series).

80-90
[Link]

Regression with More Than One Time Series

 Use the Dickey-Fuller Engle-Granger test (DF-EG test) to test the
cointegration
 H0: no cointegration Ha: cointegration
 If we cannot reject the null, we cannot use multiple regression
 If we can reject the null, we can use multiple regression

81-90
[Link]

Steps in Time-Series Forecasting

画出散点图，判断序列是否有趋势
Does series have a trend?

Yes No
模
型
组
建 线性趋势 指数趋势 判断是否有季节性因素
线 a linear trend an exponential trend Seasonality?
路
图
使用DW检验判断残差是否自相关
Yes
Serial correlation?

No Yes
使用趋势模型 使用自回归模型
Use a trend model Use an AR model

82-90
[Link]

Steps in Time-Series Forecasting

序列协方差是否固定
Is series Covariance Stationary?
No Yes

以差额法重新组建序列 以AR(1)模型开始
Take First Differences 模型的估计
自
回
归 残差是否自相关 Yes 继续增加自回归数量和级数
模
型 Serial Correlation? Adding Lags
组
建 No
路
线 是否存在季节性因素 Yes 增加相应的自回归级数
图 Seasonality Present Adding Lags
No
用ARCH模型检测是否存在异质性 Yes 通过广义的最小二乘法来调
Heteroskedasticity? 整模型中的错误

No
组建完成模型，测试模型的预测能力
83-90
[Link]

Reading
12

Excerpt from“Probabilistic Approaches: Scenario Analysis,

Decision Trees, and Simulation”

84-90
[Link]

1. Simulation

Framework 2. Comparing the Approaches

85-90
[Link]

Simulation
 Steps in Simulation
 Determine “probabilistic” variables
 Define probability distributions for these variables
 Historical data
 Cross sectional data
 Statistical distribution and parameters
 Check for correlation across variables
 When two variables are strong correlated, one solution is to pick
only one of the two inputs; the other is to build the correlation
explicitly into the simulation.
 Run the simulation

86-90
[Link]

Simulation
 Advantage of using simulation in decision making
 Better input estimation
 A distribution for expected value rather than a point estimate
 Simulations with Constraints
 Book value constraints
 Regulatory capital restrictions
 Financial service firms
 Negative book value for equity
 Earnings and cash flow constraints
 Either internally or externally imposed
 Market value constraints
 Model the effect of distress on expected cash flows and discount
rates.

87-90
[Link]

Simulation
 Issues in using simulation
 GIGO
 Real data may not fit distributions
 Non-stationary distributions
 Changing correlation across inputs

88-90
[Link]

Comparing the Approaches

 Choose scenario analysis, decision trees, or simulations
 Selective versus full risk analysis
 Type of risk
 Discrete risk vs. Continuous risk
 Concurrent risk vs. Sequential risk
 Correlation across risk
 Correlated risks are difficult to model in decision trees
Risk type and Probabilistic Approaches
Discrete/ Correlated/ Sequential/
Risk approach
Continuous Independent Concurrent
Discrete Correlated Sequential decision trees
scenario
Discrete Independent Concurrent
analysis
Continuous Either Either simulations

89-90
[Link]

It’s not the end but just beginning.

Life is short. If there was ever a moment to follow your passion and do
something that matters to you, that moment is now.

生命苦短，如果你有一个机会跟随自己的激情去做你认为重要的事，那么这
个机会就是现在。

90-90