Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Problem with panel data regression: fixed industry OR fixed firm effects & regression coefficients.

    Hi all,

    I'm quite new at Stata, so please bear with me, and got some questions.

    Basic info:
    My research is about the relationship between the share price (i.e. firm value) and integrated reporting (IR). For the regression, I use panel data; data of about 75 companies within 15 industries collected from 2013-2018. These are summary statistics of the variables I use:
    Total sample (N = 395) Without IR (N = 266) With IR (N = 129)
    Variabele Mean Min. Max. Mean Min. Max. Mean Min. Max.
    Share price 30.44
    [46.05]
    0.21 698 30.39
    [52.54]
    0.21 698 30.56
    [28.51]
    2.28 167.20
    IR Quality 123.70
    [54.43]
    54 199 106.36
    [51.07]
    19 194 172.37
    [27.46]
    54 199
    Book value per share 17.20
    [29.50]
    -2.63 378.11 16.20
    [34.34]
    -2.63 378.11 19.26
    [15.17]
    0.08 70.07
    Earnings per share 1.63
    [5.46]
    -9.34 95.61 1.53
    [6.30]
    -6.17 95.61 1.83
    [3.08]
    -9.34 10.07
    Leverage 3.24
    [8.94]
    -39.92 84.10 2.18
    [9.08]
    -39.92 84.10 5.44
    [8.25]
    0.07 59.24
    Loss 0.19
    [0.39]
    0 1 0.23
    [0.42]
    0 1 0.10
    [0.30]
    0 1
    LN_assets 21.82
    [2.30]
    15.21 27.71 21.23
    [2.13]
    15.21 27.71 23.05
    [2.15]
    18.21 27.62
    (IR and Loss are dummy variables)

    Questions:
    1. I chose the fixed effects regresion over pooled OLS, I definitely need to adjust for time effects, but I don't know exactly if I need to adjust for industry OR firm effects. Important to note is that the industry of the firms time doesn't change over time, so does that imply I only need to adjust for firm effects?

    2. I get somewhat questionable results, some things just don't make sense. Assuming I'm right at question one (i.e. adjust for firm effects), my code for the regression is:
    Code:
    xtset Id Year, yearly
    Code:
    xtreg SP IR BVPS EPS IR_BVPS IR_EPS Leverage Loss LN_Assets i.Year, fe
    Note: I performed the Hausman test, which implied the fixed effects model is better to use.

    I'm especially suprised about the negative regression coefficient of earnings per share (EPS), because that doesn't make sense. I have never seen a negative effect of EPS on share price. Moreover, can someone explain why the constant is very negative, and the regression coefficients are very small compared to the constant? I don't see how that could ever be positive.

    3. Regarding the somewhat unusual results, is it possible that I did something wrong? I've searched on the internet already, but it didn't help me unfortunately.

    I would be most grateful if you could help me!
    Attached Files
    Last edited by Jenna Parker; 23 May 2020, 12:56.
Working...
X