Empirical Exercise 1: family firms

question, hypothesis, theory

1. Question: Does appointing a family member as CEO (after a retirement) lead to worse firm performance compared to appointing an unrelated, external candidate?

2. Hypothesis: Family CEO successions result in lower firm performance than unrelated CEO successions — especially when the incoming family CEO attended a non-selective (less prestigious) college

3. Theory: 

   1. Family CEOs have inherited firm-specific knowledge and mentorship, long-term investment focus due to financial ties, stronger commitment because of family pressure

   2. External CEOS may be more capable as they would’ve been selected from a larger talent pool, will put in more effort due to job insecurity and performance evaluations, less emotional attachment: more rational decision making

type of dataset, time period

1. Cross section of CEO transitions in non-financial, non-utility publicly traded firms founded before 1971, that have potential for family transition to occur

2. CEO transition occurred between 1981 and 2000

   1. CEO transition was:

      1. Either new CEO was related to previous CEO, member of large share-holder family or member of founder family 

         1. All classed as ‘family’ transitions 

      2. Or CEO unrelated to founder family, to the large shareholder family and to the previous CEO

         1. Classes as ‘unrelated’ transitions

      3. Family dummy= 1 if new CEO is related to founder

      4.  Family dummy= 0 if new CEO is unrelated to founder

1. Also contains other info on year in which the transitions occurred, the firm assets (in millions of 2002 dollars) and the age of the incoming CEO at the time

main variables

indpendent variables:

• Least_selective variable captures whether incoming CEO attended a US selective college eg ‘good college’ like harvard, stanford

  1. Selective when least_selective=0 and non-selective transition: when least_selective=1

     1. If the founder’s child attends bad uni but still has privileged background: not very intelligent 

• family_transition

  • value 1 if family CEO and 0 for unrelated/external CEO

dependent variables:

• Return variable is a measure of performance 

  1. Captures the two day return of shares in firm i above return of firm similar to firm

• Income_to_assets:  operating return on assets for year 2002 for all companies 

  1. Each unit of value of assets is making more income, company making more profits per unit of asset value

• Change_profits: average over 3 years, return on assets following incoming CEO joining firm i minus average over 3 years return on assets prior to incoming CEO joining firm i

  • Effect on incoming CEO on firm i profits

-335 total observations

-213 unrelated transitions and 122 family transitions

-to find out how many family  and unrelated transitions occurred each year use: transition_year family

-both tell you when you use transition_year = xxxx how many transitions occurred just in that 1 year

-to find likelihood of whether transition in a year was a family transition

-average age on incoming CEO was 47.8 and standard deviation of agein is 7.43

-age_months: average age of CEO in months, mean and standard deviation are 12x age

-testing whether the mean of return is equal to 0 for all CEO transitions 

• p value is significant at the 5% level so average of returni is statistically different from 0 when CEO transition occurs

-mean of return is not different from 0 for family transitions 

• p value is not significant at 5 or 10% level

• when there is a family transition, returni is equal to 0, so no effect of family transition on performance

-mean of return is different from 0 for unrelated transitions 

• p value is not significant at 5% level 

• when there is an unrelated transition, returni is not equal to 0, so there is an effect of an unrelated transition on firm performance

-p value is significant at 5% level so mean of return is statistically different for family transitions compared to unrelated transitions

-group 0: selective transitions

-group 1: non-selective transitions 

-p value is not significant at 10% level so there is no statistical difference in the mean of return for transitions where incoming CEO attend selective college vs attending non-selective colleges

-transitions could be family or unrelated

-whether an incoming CEO who is family or unrelated attends a selective or non-selective college has no effect on the performance of the firm

-no statistical difference for mean of return in selective family transitions compared to non-selective family transitions

• p value is 0.87, not significant 

-mean of return is statistically the same in unrelated selective transitions and unrelated non selective transitions

• p value is 0.86, not significant 

• if incoming CEO is unrelated to previous CEO, it does not make a difference on performance, whether they attended a selective or nonselective college

-no statistical diff in the mean of return in selective family transitions compared to selective unrelated transitions

• p value is 0.234

• if you went to a selective college, a related incoming CEO does not perform better than an unrelated incoming CEO

-statistical difference in the means in non-selective family transitions compared to non-selective unrelated transitions

• p value is 0.066, statistically significant at 10% level 

• if you went to a non-selective college, a related incoming CEO decreases return performance/performs worse than an unrelated incoming CEO

-compare family transitions and unrelated transitions in terms of change_profits

• unrelated transitions display a higher value of change_profits than family transitions at the 1% level

• unrelated incoming CEO has a higher effect on firm i profits than family incoming CEO

• p value is 0.007

-comparing non-selective transitions and selective transitions

• no significant difference in change_profits when incoming CEO is from a selective college compared to when incoming CEO is from non-selective college

• p value is 0.482

-comparing non-selective family transitions with selective family transitions

• selective family transitions display a higher value of change_profits at the 1% level than non-selective family transitions

• when an incoming CEO is related, if they have gone to a selective college they will have a higher impact on profits compared to if related CEO went to non-selective college 

• p-value is .0003, significant at 1% level 

-no statistical diff in change_profits between non-selective unrelated transitions and selective unrelated transitions

• p value is 0.294, not significant

• when a CEO is unrelated to founder, whether they went to a selective or non-selective college has no impact on firm’s profits

-there is a statistical diff in change_profits between non-selective unrelated transitions and non-selective family transitions

• p value is 0.00, highly significant at 1% level

• non-selective unrelated transitions display a higher value of change_profits at the 1% level than non-selective family transitions

• if incoming CEO went to a non-selective college, they have a higher impact on firm’s profits if they are unrelated to founder compared to if they are related to founder

-no statistical diff in change_profits between selective unrelated transitions and selective family transitions

• p value is 0.607

• if incoming CEO went to selective college it does not make a difference on firm’s profits, whether they are related or not to previous founder

conclusion

• Family CEOs usually lead to worse performance than unrelated CEOs

  • Especially true when they have studied in non-selective colleges

• This evidence is in terms of short-term share price performance and in terms of medium term operating return on assets performance

• It seems as if a generic external candidate would be better than a generic family candidate

internal validity

1. Selection Bias/non random assignment:

   • Firms choosing family successors may differ (e.g., culture, governance, profitability) from those that choose outsiders, confounding the CEO type and performance.

   • For example, high-performing firms might be more willing to hire an external CEO, making it hard to know whether outcomes are due to the CEO or the firm's baseline characteristics.

2. Omitted Variable Bias:

   • The analysis might omit variables influencing both CEO appointment and performance (e.g., board strength, industry shocks, internal talent pipeline), biasing estimated effects.

3. Reverse Causality:

   • Firms that expect future low performance might prefer a family CEO for loyalty or cost reasons, meaning poor performance may cause the appointment rather than result from it.

4. Short-Term Event Window:

   • The key variable returni is based on a 2-day stock market window, capturing expectations rather than actual outcomes. It assumes market efficiency and may not reflect long-term CEO impact.
     

external validity

• Time Frame:

  • The data is from CEO transitions between 1981–2000. Corporate governance, investor expectations, and executive recruitment practices have evolved, potentially limiting relevance today.

• Geographic and Institutional Scope:

  • The study is based on U.S. public firms with family involvement. Results may not generalise to:

    • Non-public firms,

    • Non-U.S. contexts

• Industry Composition:

  • The sample excludes financial and utility firms, limiting generalizability to other sectors