MFIN6210 Week 5 (Lab Sessions, in regular weekly classes, Monday, Thursday)

Objective: To test whether US acquiring firms with more anti-takeover provisions (ATPs) make poorer investment decisions.

Make sure you bring your notebook/touchscreen/computer.

Lab session: Data analysis I (manipulation and basic STATA commands). See lab folder on Moodle for additional readings and data. The Computer Lab classes in Weeks 4 and 5 are in your regular classes using “myAccess”,, which gives you access to both MS Excel and STATA-SE 14 on your own laptop/computer. It also gives you access at home, as well as at UNSW.

Useful readings:

•Masulis, R., Wang, C. and F. Xie, ‘Agency problems at dual-­‐class companies’, Journal of Finance 62, 2008, 1851-1889. (See Moodle, Lab sessions and group project)


1.Open the dataset (lab_dataset.xls) in Microsoft excel (available from Moodle). The spreadsheet contains data on US acquiring firms, comprising of: (1) acquiring firm cumulative abnormal returns; (2) accounting and industry data; and (3) governance, deal and other financial characteristics. We will follow closely the test performed by Masulis et al. (2008) in Table 6 (Panel A and B), but will substitute our proxies for governance (APTs) for their excess control rights variable (Ratio).

2.I have created a worksheet in the spreadsheet called ‘my work’. Using the raw data from the dataset, create variables in ‘my work’ that are required to perform the replication of the Masulis et al., (2008) test. You can use the @if command to create dummy variables for some of the deal characteristics (e.g., public, private, subsidiary, cash, stock, diversifying). Note that you also have industry average variables, so you could also create industry-­‐adjusted variables for alternative dependent variables using value and performance metrics (i.e., Q, ROA, etc). You should be aware of the implicit assumptions you are making when using value metrics for a sample of acquiring firms (hint – Q will be measured for the year prior to takeover).

3.Create dummy variables for high/low ATP firms for the GIM index and the BCF index. Use cut-­‐offs of GIM=>10, <10 and BCF=>3, <3 for high/low portfolios. Use these cut-­‐ offs to perform a univariate analysis of whether high/low ATP firms (using GIM, BCF, CBOARD) have significantly different deal and firm characteristics.

4.Create a new spreadsheet (CSV, comma delimited) for input to STATA. Your spreadsheet should contain the cumulative abnormal returns (CAR) for all acquiring firms plus variables (governance, deal and financial characteristics, including newly created variables) that may explain the

cross-­‐sectional differences in CARs and other value metrics. Note that you will also have to create a CAR dummy (=1, if CAR<0, 0 otherwise) and interactive dummies for public status and method-of-payment). You can create the variables in excel or use the generate command in STATA. Open STATA. Use the file-­‐import command to read the spreadsheet (text; csv) into STATA.

5.Calculate summary statistics for the variables in your worksheet (mean, median, minimum, maximum, standard deviation). What do you infer from the descriptive statistics?

6.Estimate pairwise correlation coefficients for all variables in your spreadsheet. On examination of the pair-­‐wise correlations, which variables, if any, are likely to pose multicollinearity problems (denote which are significant at the 5% level)?

7.Estimate standard OLS regressions (statistics-­‐linear models-­‐linear regression) to test whether CARs can be explained by governance, deal, acquiring firm characteristics. I suggest you estimate several models, each containing a governance variable (i.e., GIM, GIM-dummy, BCF, BCF-dummy, CBOARD). Re-estimate using the CAR dummy variable. Re-estimate for different sample specifications: (1) completed deals; (2) TV>$1m.

8.Perform standard OLS regression diagnostics (i.e., heteroskedasticity, multicollinearity) to establish if your regressions are well-­‐specified. Is there any year or industry-­‐fixed effects in your regressions? (Hint: include year and/or industry dummies). Are the standard errors robust to possible clustering by acquiring firms? (Hint: use robust cluster options for acquirer identity code (npermno)).

9.Perform additional robustness tests: (1) outliers (use the winsor command to test if outliers impact on your results). {winsor car, gen(carwin) p(0.01)}.

10.Replicate the model in Masulis et al. (2008) Table 6 (Panel B). You will need to generate a new dependent variable to denote failed and successful deals (call it ‘compm’ to denote completed as defined in Masulis et al (2008).

11.Create a STATA do file to automate the above. You can augment your do file from week 6. This do file will be useful for your project.


TobinQ : (AE-AN+AK*AR) / AE

Lev (AF + AL) / (AE-AN+AK*AR)

Inda_q TobinQ – BN
ssc install winsor