The first article that caught my eye was “The Impact of IT Investments on Profits,” by academic Sunil Mithas and others. The summary presented in SMR noted research data from 400 global companies collected from 1998-2003 showing a positive effect of IT investment on firm profitability. That’s certainly not surprising, but the finding that “investment in IT had a greater impact on a company’s profits than comparable spending on either advertising or R&D” was. So I downloaded the source study paper and examined the research in more detail.
The author’s driving hypotheses are that investments in IT have a positive impact on firm profitability through both positive linkage to revenue and negative relationship to costs. In addition, he posits that IT investments have a greater impact on profitability than comparable expenditures on either advertising or R&D.
The data set used to test the hypotheses comprised 452 firms for which key IT investment and financial performance variables of interest over a six-year period. The authors used regression techniques to estimate panel (over time) models linking firm performance variables such as profitability and revenue to investment indicators and other control variables. The findings provide support for the relationship between IT investment and both increased revenue and profitability, but not for decreased costs. In addition, the computations suggest that the effect of IT investments on profitability is greater than that of advertising and R&D.
Mithas, et al. conclude: “From a top management or board perspective, IT investments should receive significant attention in governance and resource allocation processes because they appear to be more important than R&D and advertising in terms of profitability. Chief information officers can use the findings to develop a business case and justification for continued investments in IT.”
I have to admit, I was a bit underwhelmed with this analysis, but, noting the authors’ apparent statistical prowess, decided to give them another chance, opting for a quick read of a second paper by Mithis mentioned in “The Impact of IT’s” references.
For me, “From Association to Causation via a Potential Outcomes Approach” was much more interesting, especially for the statistical methodology that has applicability to business evaluation and is becoming increasingly popular in analytics circles. The author sets out to determine the effect of earning an MBA degree on IT professionals’ salaries. More specifically, he wants to know if getting an MBA “causes” positive changes in IT compensation.
Mithis’ data set, derived from the 2006 National Salary Survey conducted by InformationWeek, would just as well suit a data scientist as a business academic. Among the attributes collected from the 9,000+ respondents are total compensation in 2005, age, education level, years of experience, tenure at current position, company size, industry, hours per work week, dotcom indicator, and frequency of contact with recruiters. The non-compensation variables are considered covariates.
The big challenge with this investigation is making inferences about cause and effect with limited point-in-time “observational” data. Ideally, we’d like to know what individual X would make if she earned an MBA (the factual) and contrast that with what she’d make without one (the counterfactual). This, of course, is highly impractical.
Also impractical is the usual scientific accommodation of conducting an experiment where IT professionals are randomly assigned to either an “earn-an-MBA” group or a “not-earn-an-MBA” one. Without randomization that assures within statistical limits that “other factors are equal” between the treatment groups, however, any comparison of MBA versus non-MBA salaries might be biased and open to the alternative explanations of skeptics. It could well be the case, for example, that IT professionals who opt to pursue the MBA are simply more motivated and would earn higher compensation regardless of attaining an advanced degree.
Mithis’s methodological solution is to use a potential outcomes-based propensity score model to estimate the causal impact of the MBA. With this approach, a propensity score that summarizes an individual’s probability of getting an MBA given her values on the other collected attributes (covariates) is first estimated using logistic regression techniques. Then, each MBA group case is matched with one or more non-MBA cases based on similar propensity scores. This matching accomplishes after the fact statistically what randomization does before – assuring within limits that other factors are equal. Finally, the mean compensations of MBAs are compared to non-MBAs within categories of similar propensity scores.
The findings? Even after bias-cleansing propensity score adjustments, IT professionals with an MBA earn significantly more than their non-MBA colleagues. Apparently, the MBA pays off for IT. A left-unanswered question: is the MBA worth more than other advanced degrees in the IT world?