# Interpreting of hypothesis prediction after reversed relationship

I am currently facing a big issue with regard to my master thesis. I developed following hypotheses:

Hypothesis 1: Managers’ investment in innovation increases when the organization is judged in decline state (positive relationship between organizational decline and investments in innovation)

Hypothesis 2: The relationship between organizational decline and investments in innovation is positively moderated by managerial ownership: Managers’ with higher company ownership will invest more in innovation when the organization is judged in a decline state

Hypothesis 3: The relationship between organizational decline and investments in innovation is positively moderated by institutional ownership: Firms with higher institutional ownership will invest more in innovation when the organization is judged in a decline state

Here you can find my conceptual model: http://imageshack.us/photo/my-images/29/kvji.png/

This is the outcome of the regression:

$log(Investments in Innovation) = 6.984 - 0.852Decline - 4.703Managerialown + 3.030ManagerialownDecline-1.870Institionalown + 1.14 Institionalown x Decline$

My issue is now at the beginning I predicted a positive moderation in Hypothesis 2 and 3 based on a positive relationship in Hypothesis 1.

Now the return shows that there is a negative relationship in Hypothesis 1 (so managers invest less), yet both moderators Managerialown and Institutionalown mitigate this relationship (so managers invest less at a slower rater with increasing ownership)

Do I confirm my Hypothesis 2 and 3 or reject them? As I understand, a positive moderation intensifies the main effect. However, now the moderator lessens the main effect of Hypothesis 1, yet my second description of the hypotheses is still somehow valid (Managers’ with higher company ownership will invest more in innovation when the organization is judged in a decline state)

Thank you very much for your help.

Tom

I think the term "positive moderation" - if it means anything - means that the moderation term has a positive coefficient. You explain what you mean in the second part of H 2 and 3, so I would say you have confirmed them.

I hadn't seen "positive moderation" or "negative moderation" in this sense - I don't think they are used much (probably because of precisely the confusion brought on by the issues you raise).

To confirm this, I first Googled the two words "positive" and "moderation" and I got what I thought I would - lots of articles about moderation in the statistical sense. When I Googled the phrase "positive moderation" however, I got mostly stuff unrelated to statistics.

• Thanks Peter. Do you think it would be perhaps wise to rephrase the second and third hypothesis to a "neutral" statement? e.g. "The relationship between organizational decline and investments in innovation is moderated by managerial ownership: Manager's with higher company ownership will invest more in innovation when the organization is judged in a decline state"
– Tom
Commented Aug 10, 2013 at 12:21
• If your committee lets you rephrase them, yes. Commented Aug 10, 2013 at 12:44
• Well, it is up to my own judgement. So would you agree that the neutral statement is more accurate / more scientific correct? One last thing. I really appreciate your help Peter. Without people like you, lots of students would be lost!
– Tom
Commented Aug 10, 2013 at 12:56
• I would just delete the phrase "positive moderation" and leave the rest. Commented Aug 10, 2013 at 13:02