Quantitative Methods Module 7 hypothesis testing Examples with solution

1. Willco is a manufacturer in a mature cyclical industry. During the most recent industry cycle, its
net income averaged $30 million per year with a standard deviation of $10 million (n = 6
observations). Management claims that Willco’s performance during the most recent cycle
results from new approaches and that we can dismiss profitability expectations based on its
average or normalized earnings of $24 million per year in prior cycles.
A. With μ as the population value of mean annual net income, formulate null and alternative
hypotheses consistent with testing Willco management’s claim.
B. Assuming that Willco’s net income is at least approximately normally distributed, identify the
appropriate test statistic.
C. Identify the rejection point or points at the 0.05 level of significance for the hypothesis tested in
Part A.
D. Determine whether or not to reject the null hypothesis at the 0.05 significance level.
SOLUTION:
A. 𝐻 : 𝜇 ≤ 24 versus 𝐻 : 𝜇 > 24
B. The net income is normally distributed with unknown variance, t-test with df=6-1=5
C. One-tailed test 𝑡 . = 2.015. Reject the null if t>2.015.
𝑋 − 𝜇 30 − 24
𝑡= 𝑠 = = 0.6√6 = 1.46969 < 2.015
30
√𝑛 √6
fail to reject the null. The difference between the sample mean $30million and the
hypothesized value $24million under the null is not statistically significant.

1
Quantitative Methods Module 7 hypothesis testing Examples with solution

2. Investment analysts often use earnings per share (EPS) forecasts. One test of forecasting quality
is the zero-mean test, which states that optimal forecasts should have a mean forecasting error of
0. (Forecasting error = Predicted value of variable − Actual value of variable.)
You have collected data (shown in the table above) for two analysts who cover two different
industries: Analyst A covers the telecom industry; Analyst B covers automotive parts and
suppliers.
Number of Mean Forecast Error Standard Deviations
Forecasts (Predicted – Actual) Of Forecast Errors
Analyst A 101 0.05 0.10
Analyst B 121 0.02 0.09
A. With μ as the population mean forecasting error, formulate null and alternative hypotheses for a
zero-mean test of forecasting quality.
B. For Analyst A, using both a t-test and a z-test, determine whether to reject the null at the 0.05
and 0.01 levels of significance.
C. For Analyst B, using both a t-test and a z-test, determine whether to reject the null at the 0.05 and
0.01 levels of significance.
SOLUTION:
A. 𝐻 : 𝜇 = 0 versus 𝐻 : 𝜇 ≠ 0, 2-tailed test
.
B. For Analyst A, 𝑡𝑒𝑠𝑡 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 = = . = 0.5√101 = 5.0249
√ √
𝑎𝑡 𝛼 = 0.05, 𝑑𝑓 = 100, 𝑡 . = 1.984(𝑧 . = 1.96)
𝑎𝑡 𝛼 = 0.01, 𝑑𝑓 = 100, 𝑡 . = 2.626(𝑧 . = 2.575)
Reject the null at 0.05 and 0.01 levels of significance
.
C. For Analyst B, 𝑡𝑒𝑠𝑡 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 = = . = √121 = 2.4444
√ √
𝑎𝑡 𝛼 = 0.05, 𝑑𝑓 = 120, 𝑡 . = 1.980(𝑧 . = 1.96)
𝑎𝑡 𝛼 = 0.01, 𝑑𝑓 = 120, 𝑡 . = 2.617( 𝑧 . = 2.575)
Reject the null at 0.05 level of significance and fail to reject at 0.01 level of significance.

2
Quantitative Methods Module 7 hypothesis testing Examples with solution

3. Reviewing the EPS forecasting performance data for Analysts A and B in Question 2, you want
to investigate whether the larger average forecast errors of Analyst A are due to chance or to a
higher underlying mean value for Analyst A. Assume that the forecast errors of both analysts
are normally distributed and that the samples are independent.
A. Formulate null and alternative hypotheses consistent with determining whether the
population mean value of Analyst A’s forecast errors (𝜇 ) are larger than Analyst B’s (𝜇 ).
B. Identify the test statistic for conducting a test of the null hypothesis formulated in Part A.
C. Identify the rejection point or points for the hypothesis tested in Part A, at the 0.05 level of
significance.
D. Determine whether or not to reject the null hypothesis at the 0.05 level of significance.
SOLUTION:
A. 𝐻 : 𝜇 ≤ 𝜇 versus 𝐻 : 𝜇 > 𝜇
B. We have two normally distributed populations with unknown variances. Based on the
samples, it is reasonable to assume that the population variances are equal. The samples are
assumed to be independent; this assumption is reasonable because the analysts cover quite
different industries. The appropriate test statistic is t using a pooled estimate of the common
variance. The degree of freedom is 𝑛 + 𝑛 − 2 = 101 + 121 − 2 = 220
C. For df=220, 𝑡 . = 1.6518 (Or use the closest df=200, t=1.653)
D. We first calculate the pooled estimate of variance:
(𝑛 − 1)𝑠 + (𝑛 − 1)𝑠 (101 − 1)0.10 + (121 − 1)0.09
𝑠 = = = 0.008964
𝑛 +𝑛 −2 101 + 121 − 2
Then
(𝑋 − 𝑋 ) − (𝜇 − 𝜇 ) 0.05 − 0.02
𝑡= = = 2.3512 > 1.6518
𝑠 𝑠 0.008964 0.008964
+
+ 101 121
𝑛 𝑛
We reject the null hypothesis in favor of the alternative hypothesis that the population mean
forecast error of Analyst A is greater than that of Analyst B.

3
Quantitative Methods Module 7 hypothesis testing Examples with solution

4. Altman and Kishore (1996), in the course of a study on the recovery rates on defaulted bonds,
investigated the recovery of utility bonds versus other bonds, stratified by seniority. The
following table excerpts their findings.

Industry Group Ex-Utilities Sample

Industry Group Number of Average Standard Number of Average Standard
/Seniority Observations Price* Deviation Observations Price* Deviation
Public Utilities
32 $77.74 $18.06 189 $42.56 $24.89
Senior Unsecured
*This is the average price at default and is a measure of recovery rate.
Source: Altman and Kishore (1996, Table 5).
Assume that the populations (recovery rates of utilities, recovery rates of non-utilities) are normally
distributed and that the samples are independent. The population variances are unknown; do not
assume they are equal. The test hypotheses are 𝐻 : 𝜇 − 𝜇 = 0 versus 𝐻 : 𝜇 − 𝜇 ≠ 0, where 𝜇
is the population mean recovery rate for utilities and 𝜇 is the population mean recovery rate for
non-utilities.
A. Calculate the test statistic.
B. Determine whether to reject the null hypothesis at the 0.01 significance level without reference to
degrees of freedom.
C. Calculate the degrees of freedom.
SOLUTION:
𝑛 𝑋 𝑠 𝑛 𝑋 𝑠
32 $77.74 18.06 = 326.1636 189 $42.56 24.89 = 619.5121

( ) ( ) . .
A. 𝑡 = = . .
= 9.585

B. Since the test statistic is so large that we can confidently reject the null at 0.01 significance
level
. .

C. 𝑑𝑓 = = = 54.929
. ⁄ . ⁄

4
Quantitative Methods Module 7 hypothesis testing Examples with solution

5. The table below gives data on the monthly returns on the S&P 500 and small-cap stocks for the
period January 1960 through December 1999 and provides statistics relating to their mean
differences.
Measure S&P 500 Return (%) Small-Cap Stock Return (%) Differences (S&P 500 – Small-Cap Stock)
January 1960–December 1999, 480 months
Mean 1.0542 1.3117 –0.258
Standard Deviation 4.2185 5.9570 3.752
January 1960–December 1979, 240 months
Mean 0.6345 1.2741 –0.640
Standard Deviation 4.0807 6.5829 4.096
January 1980–December 1999, 240 months
Mean 1.4739 1.3492 0.125
Standard Deviation 4.3197 5.2709 3.339
Let μd stand for the population mean value of difference between S&P 500 returns and small-
cap stock returns. Use a significance level of 0.05 and suppose that mean differences are
approximately normally distributed.
A. Formulate null and alternative hypotheses consistent with testing whether any difference exists
between the mean returns on the S&P 500 and small-cap stocks.
B. Determine whether or not to reject the null hypothesis at the 0.05 significance level for the
January 1960 to December 1999 period.
C. Determine whether or not to reject the null hypothesis at the 0.05 significance level for the
January 1960 to December 1979 subperiod.
D. Determine whether or not to reject the null hypothesis at the 0.05 significance level for the
January 1980 to December 1999 subperiod.
SOLUTION:
A. We test 𝐻 : 𝜇 = 0 versus 𝐻 : 𝜇 ≠ 0
B. This is a paired comparisons t-test with n-1=480-1=479 degrees of freedom. At 0.05 significance level,
we reject the null hypothesis if either t>1.96 or t<-1.96. we use df=∞ in the t-distribution table under 𝛼 =
0.025 because we have a very large sample and a two-sided test. (or you can use EXCEL to get the
critical value is 1.9649)
𝑑̅ − 𝜇 −0.258 − 0 −0.258
𝑡= = = = −1.506529 𝑜𝑟 − 1.51
𝑠 4.096 0.171255
√480
At 0.05 significance level, we fail to reject the null hypothesis that the mean difference between
the returns on the S&P 500 and small-cap stocks during the entire sample period was 0.
C. This t-test now has n-1=240-1=239 degrees of freedom. At the 0.05 significance level, we reject
the null hypothesis if either t>1.972 or t<-1.972, using df=200 in the t-distribution tables.
𝑑̅ − 𝜇 −0.640 − 0 −0.640
𝑡= = = = −2.420615 𝑜𝑟 − 2.42
𝑠 4.096 0.264396
√240
At 0.05 significance level, we reject the null. During this subperiod, small-cap stocks
significantly outperformed the S&P 500.
D. This t-test has n-1=240-1=239 degrees of freedom. At the 0.05 significance level, we reject the
null hypothesis if either t>1.972 or t<-1.972, using df=200 in the t-distribution tables.
𝑑̅ − 𝜇 0.125 − 0 −0.125
𝑡= = = = 0.579962 𝑜𝑟0.58
𝑠 3.339 0.215532
√240
At 0.05 significance level, we fail to reject the null. During this subperiod, the mean
difference between the returns on the S&P 500 and small-cap stocks was zero.

5
Quantitative Methods Module 7 hypothesis testing Examples with solution

6. During a 10-year period, the standard deviation of annual returns on a portfolio you are
analyzing was 15 percent a year. You want to see whether this record is sufficient evidence to
support the conclusion that the portfolio’s underlying variance of return was less than 400, the
return variance of the portfolio’s benchmark.
A. Formulate null and alternative hypotheses consistent with the verbal description of your
objective.
B. Identify the test statistic for conducting a test of the hypotheses in Part A.
C. Identify the rejection point or points at the 0.05 significance level for the hypothesis tested in
Part A.
D. Determine whether the null hypothesis is rejected or not rejected at the 0.05 level of significance.
SOLUTION:
A. 𝐻 : 𝜎 ≥ 400 versus 𝐻 : 𝜎 < 400
B. The test statistic is chi-square with 10-1=9 degrees of freedom
C. The rejection point is found across degrees of freedom of 9, under the 0.95 column, 3.325,
we will reject the null hypothesis if we find that 𝜒 < 3.325.
D. The test statistic is calculated as
(𝑛 − 1)𝑠 9 × 15 2025
𝜒 = = = = 5.0625 > 3.325
𝜎 400 400
We fail to reject the null hypothesis.

6
Quantitative Methods Module 7 hypothesis testing Examples with solution

7. The following table shows the correlation matrix for various US debt returns and US large and
small company stock returns using monthly data from January 1926 to December 2012.

Can we reject a null hypothesis that the underlying or population correlation coefficient of US
long-term government bonds and monthly returns to T-bills equals 0 at the 0.05 and 0.01 levels
of significance?
Solution:
To test the null hypothesis 𝐻 : 𝑟 = 0 versus alternative hypothesis 𝐻 : 𝑟 ≠ 0
We use t-test. There are 1,044 months during the period January 1926 to December 2012. The
degrees of freedom is 1044-2=1.42. At 0.05 level of significance, the critical level 𝑡 = 1.96. At
0.01 level of significance, the critical level 𝑡 = 2.576. We use the following test statistic:
𝑟√𝑛 − 2 0.18√1044 − 2
𝑡= = = 5.9069 > 2.576 > 1.96
√1 − 𝑟 √1 − 0.18
so we can reject the null hypothesis of no correlation in the population at both the 0.05 and 0.01
levels. (This example shows that, in large samples, even relatively small correlation coefficients can
be significantly different from zero.)

7
Quantitative Methods Module 7 hypothesis testing Examples with solution

8. Suppose we observe the following frequency table of 1,594 exchange-traded funds (ETFs)
based on two classifications: size (that is, market capitalization) and investment type (value,
growth, or blend).

A. What are the expected frequencies if the market capitalization and the investment type are
independent?
B. Using a 5% level of significance, determine whether the market capitalization and the
investment type are independent of one another.
Solution:
A.

B.
𝐻 : ETF size and investment type are not related, so these classifications are independent;
𝐻 : ETF size and investment type are related, so these classifications are not independent.
the test statistic is
(𝑂 − 𝐸 )
𝜒 = ~𝜒( )×( )
𝐸
( )
where is summarized in the following table

With the level of significance 5%, a one-sided test critical value is 9.4877. We reject the null
hypothesis if the calculated 𝜒 statistic is greater than 9.4877.
Calculate the test statistic. 𝜒 = 32.08025>9.4877. We reject the null hypothesis: these
classifications are not independent.