0
$\begingroup$

I am given a task to perform an analysis on marketing data and to find out if the prices for the products and marketing budget are set correctly.

In order to do it, I need to find a suitable model of marketing instrumental variables and interpret its validity. The set of variables includes:

  • sales volume (number of units sold)
    • price
    • marketing budget (for advertising)
    • number of sales representatives

Data is given for 16 districts.

When I try to perform 2sls in SPSS, I put:

  • sales volume - independent;
  • price - explanatory
  • adv budget and N of sales rep - instrumental.

I get small F value (I found it should be more than 10 if I use one instrument), R squared and p value. It indicates that my model doesn't make sense.

Do I actually do it correctly, using 2SLSS straight without finding extra variables? I assumed that I need to find new variables first by using OLS, and then perform 2SLS in SPSS.

I will appreciate any help and small guidance on using 2sls in SPSS.

Update: Is there an opportunity for someone with extensive marketing mix modelling experience to provide me with more in-depth guidance (via e-mail)? I already found the material I need to use, and I know the steps, but I am afraid to omit some important details which can lead to suboptimal outcome. I need to use multiplicative regression, 2 equations for profit maximization (on price and marketing costs) and finally check how this setting will affect profit and sales. I rely on http://www.anderson.ucla.edu/faculty/dominique.hanssens/content/modeling_marketing_interactions.pdf , since I also have to consider salesforce expenditures, which are part of my marketing budget. Thank you in advance!

Improve this question
$\endgroup$
5
  • $\begingroup$ Why do you think you need 2SLS? Is there more than 1 product? Why not leverage a marketing mix model? $\endgroup$
    – Mike Hunter
    Nov 28 '15 at 15:32
  • $\begingroup$ This is given in the task. "Find a suitable model of marketing instrumental variables on sales volume and interpret the results" And I found that this can be done using 2SLS. The outcome is supposed to be a price elasticity (which is Beta). And I need to make an equation in order to find whether the budget is allocated correctly and the price is justified. $\endgroup$
    – Julia
    Nov 28 '15 at 16:11
  • $\begingroup$ Ok...but how are you motivating the use of 2SLS when the industry standard is a marketing mix model? Other issues include fitting the model to different products (assuming there is more than one), over time and also for the different regions or territories covered by the sales reps (it can't be all one territory). The single best reference I know of for fitting MMMs is Lee Cooper's free ebook Market Share Analysis available on his UCLA website ... anderson.ucla.edu/faculty/lee.cooper/MCI_Book/BOOKI2010.pdf Forget about "share analysis," it's just a great how-to for MMMs $\endgroup$
    – Mike Hunter
    Nov 28 '15 at 16:21
  • $\begingroup$ Thank you! I am a finance major, and I took this class on marketing research as it looked useful. Never did any kind of such analyses, so 2SLS was the first thing I googled :) I will of course use what is the industry standard. Can you please briefly describe what will be the outcome of my model, so I will not choose a wrong direction again. it is said : 'draw inferences 1. On the optimal marketing budget and its allocation to sales force and adv spending' 2. The optimal price and 3. effects of setting optimal budgets/prices on sales and profit. ' (use Amoroso-Robinson relation) $\endgroup$
    – Julia
    Nov 28 '15 at 17:00
  • $\begingroup$ Using one of Cooper's "baseline" models, estimating price elasticity is a huge focus of his methods. If you really want to "ace" your project, using the parameters from the model you can build a secondary function that would optimize the ad spend budget. $\endgroup$
    – Mike Hunter
    Nov 28 '15 at 17:32
1
$\begingroup$

I think you are mixing up two unrelated concepts. Instrumental variables in 2sls are variables from a multi-equation model that make it possible to estimate the model in the presence of simultaneity.

In your assignment, the term is being used to mean variables that you control that can affect the dependent variable.

Improve this answer
$\endgroup$
1
  • 1
    $\begingroup$ That's probably true, cause I had no too much theory on this subject and at the end I has been given an assignment without guidance. Still I have to do it all in SPSS (before I had experience with SAS, mainly time series data for my finance classes and I don't remember a lot about IV). My assignment is just called - Regression analysis of sales data. Thank you for the response. I will follow DJohnson suggestion regarding marketing mix concept. $\endgroup$
    – Julia
    Nov 30 '15 at 10:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.