Instructions on Final Project
Option 1: Coding based project-Backtesting a trading signal (number of students: 3-6)
The purpose of this project is to help you get in hands with the quantitative trading strategies 
in practice. You need to use the knowledge you learned in lectures to construct your own 
portfolio and trading strategies and backtest your strategy using programming language
(Python, Stata, SAS, R etc.). Submit your written report (3-5 pages), with your program code, 
program output, and other related supporting materials on or before the due date, 10:00 pm 29th
February, 2024.
Handling Data:
Step 1: Constructing the sample
1. Quant signal selection, you can use either an existing one (e.g. accruals, Piotroski F-score, 
the Magic formula), a composite score from multiple signals you learn from the class, or you 
may come up with your own trading signal (with bonus grade 3 points). 
2. Download financial data from COMPUSTAT and the market data (i.e. stock returns) from 
CRSP from 1980-2023.
3. Screen the data
e.g. Remove tiny stocks by requiring market value>1 million and/or price>1 dollar
e.g. Remove non-sensible observations with negative sales/assets (obvious data errors)
e.g. Remove firms from financial and utility industries, as they are highly regulated
Step 2: Describe your data
1. Please calculate the Mean, Median, Variance, 25% and 75% of the distribution of your 
trading signal variable and stock return variable.
** Please don  t forget to scale your trading signal variable by either asset or market 
value. Otherwise, you only pick up the scaling effect. 
2. Please present the correlations among your trading signal variable and other common 
firm characteristics, including book value of firm, market value of firm, book-to-market 
ratio, various financial ratios (such as asset growth, leverage ratio, return on assets, 
sales growth etc.).
Step 3: Backtest: Perform portfolio analysis based on your (historical) data
1. Please use market adjusted returns to proxy for abnormal returns (see class note). 
Bonus grade (1 points) will be given if you can use size and B/M adjusted returns to 
proxy for abnormal returns. 
2. Each year, sorting firms into deciles (i.e. 10 portfolios) according to your trading 
signal
3. Each year, calculating simple average abnormal return (i.e. equal weighted portfolio 
1
2
return) for each decile and the hedge return (i.e. the difference in abnormal returns 
between the long decile and the short decile, ALPHA) based on your trading signal.
a) If your trading signal is a positive return predictor   decile one is associated with 
the lowest future stock return (short) and decile 10 is associated with the highest 
stock return (long)
b) If your signal is a negative return predictor   decile one is associated with the 
highest future stock return (long) and decile 10 is associated with the lowest stock 
return (short)
4. Please plot the hedge return (alpha) each year from 1980-2022 and describe the trend 
of the hedge returns
5. Based on time-series hedge returns in step 4 (i.e. 43 observations), please calculate the 
mean average abnormal return (average alpha) and associated t-statistics (=mean 
average abnormal return/standard deviation; the higher the larger the Sharpe Ratio)
across 1980-2022.
Written report (in 3-5 pages, font 12, double spaced, the number of pages for program 
and output is unlimited)
Please explain 
1) The rationale of your trading signal. Do you expect it to be a positive or a negative stock 
return predictor? Why?
2) Please describe descriptive statistics.
3) Please describe the correlations among your trading signal and other firm characteristics.
4) Please describe the results of your portfolio tests. 
a) How many years out of 43 years can your trading signal generate positive alpha? Do 
you observe any time-trend pattern (e.g. declining magnitude of alpha)? 
b) Can your trading strategy generate statistically significant alpha in your sample period?
3
Option 2: Coding based project-Predicting fundamentals (number of students: 3-6)
The purpose of this project is to help you get in hands with the quantitative predictive analytics
in practice. Your objective is to predict a fundamental variable (i.e. earnings, revenue, cash 
flow from operation or other metrics you are interested*** Again, please scale your variable 
by either asset or market value) using programming language (Python, Stata, SAS, R etc.) for 
a large scale of data. Submit your written report (3-5 pages), with your program code, program 
output, and other related supporting materials on or before the due date.
Handling Data:
Step 1: Constructing the sample (panel data)
1. Download financial data from COMPUSTAT and the market data (i.e. stock returns) from 
CRSP (if necessary) from 1980-2023.
2. Screen the data
e.g. Remove tiny stocks by requiring market value>1 million and/or price>1 dollar
e.g. Remove non-sensible observations with negative sales/assets (obvious data error)
Step 2: Describe your data
1. Please calculate the Mean, Median, Variance, 25% and 75% of the distribution of your 
dependent (Y) and independent variables (Xs).
2. Please present the correlations among your Y and X variables.
Step 3: In-sample prediction 
You may try different combinations of predictive variables you expect to forecast Y 
For each trial, please specify Y (target) and Xs (predictive variables) of your predictive 
model (i.e. Y=a+b1X1+b2X2+b3X3  ). 
1. For each trial, please run OLS regression. Any other methods are encouraged. 
2. For each trial, please describe the economic (i.e. magnitude of the coefficients on your 
predictive variables) and statistical significance (i.e. t-stat and p-value associated with 
the coefficient of each predictive variable).
3. For each trial, please describe the overall fitness of the model (i.e. adjusted R-square). 
Written report (in 3-5 pages, font 12, double spaced the number of pages for program
and output is unlimited): 
Besides describing results mentioned above. 
1) Please explain the rationale of your choice of each predictive variable. Do you expect it to 
be a positive or a negative Y predictor? Why?
2) Please describe descriptive statistics.
3) Please describe the correlations among X and Y variables.
Option 3: Non-coding based project (number of students: 3-6 students)
You will prepare a detailed analysis of earnings quality of a company. Select a firm in which 
you have a particular interest. Assess its attractiveness from the perspective of a hedge fund 
specializing in US equity and focusing on value investing. Concluding your findings in the 
form of an Analyst Recommendation (Buy, Sell, or Hold).
DESCRIPTION 
The team will be analyzing a company  s annual report for consecutive 5 years. The team must 
prepare a prospectus of not more than 12 pages (font 12, double spaced), including title page, 
which describes the team  s analyses and recommendations. A prospectus is a short description 
of the analyses and must include the following sections: 1. Executive Summary 2. Earnings 
Quality Analysis 2. Conclusions/Recommendations 
SECTION 1: EXECUTIVE SUMMARY 
This section provides a brief overview of the company. Participants are not limited but, at a 
minimum, should provide the following information for both companies:
 Official name of the corporation 
 Stock symbol of the corporation and the exchange on which it is traded 
 Date of the annual report (10-K) filing according to the financial statements provided 
 The primary products(s) and/or services (s) of the corporation 
 The major competitors of the corporation
SECTION 2: EARNINGS QUALITY ANALYSIS 
Your project should demonstrate evidence of detailed financial analysis of firm data. 
Integration of tools and concepts acquired in this and other courses (e.g., financial ratio, vertical 
analysis, horizontal analysis, cash flow analyses and analyst forecasts) will enhance your grade. 
Your project should also include some form of   benchmarking,   that is, comparison to other 
similar firms or circumstances.
Please analyze major accounts on the Balance Sheet and Income Statement, and also read the 
Cash Flow Statement to see whether you can locate the trace of possible earnings misstatement. 
Hints are provided, but not limited, below: 
Incentives: 
 Any change of auditors, CEO/CFO? 
4
 Earnings are just above an important benchmark (such as analyst forecast, zero or last 
year  s earnings)? 
 Issuance of capital (equity or debt) or any M&A transaction?
Balance Sheet:
 Abnormally high accrual or low accruals for each individual accruals account? 
 Any   cookie jar reserve   or reverse of the reserve?
 Any evidence on   big bath   accounting? Such as   not well explained   big scale write?off/down? 
 Excess capitalization?
Income Statement and Comprehensive Income   
 Big change in R&D, advertising intensity or SG&A? 
 Big amount of gains or losses? 
 Big amount of dirty surplus? 
 Big book-tax difference? 
Quantitative Analysis
Please calculate the following quantitative scores each year to help you access the financial 
status of the company and its competitors (at least two). 
 Piotroski F-score (overall financial health status)
 Beneish-M score (likelihood of earning manipulation)
 Benford score for Balance Sheet and Income Statement (hint of earnings manipulations)
 Altman Z-score (likelihood of financial distress)
 Dechow et al.-F score.(likelihood or earnings manipulation)
Other: 
 Any changes in accounting policy? 
 Off-Balance Sheet Transactions? 
 Adequate Disclosure? 
 Enough information for contingency? 
 Reasonable assumptions for the discount rate applied to pension assets/liabilities? 
SECTION 3: CONCLUSIONS/RECOMMENDATIONS Draw conclusions from the data 
and your analysis, what recommendation would you make to current and potential private or 
organizational investors of the company? 


請(qǐng)加QQ：99515681  郵箱：99515681@qq.com   WX：codinghelp