End this corporate tax giveaway: R&D incentives

One of the bills before Congress this year makes R&D tax credits permanent at an estimated cost of $156 billion over the next decade. R&D tax credits were originally implemented as part of the Economic Recovery Tax Act (ERTA) of 1981 to reverse the dramatic decline in R&D that began in 1964. The twin goals were reviving economic growth and bolstering U.S. competitiveness against the rising threat from Japanese manufacturing.

They haven't delivered on their goal, and in an era of hotly debated corporate tax reform, this cut is easy to make: It's time to end R&D tax credits.

Victor Maffe | iStock / 360 | Getty Images

Proponents, such as TechNet—an organization of technology CEOs founded in 1997 by tech luminaries John Doerr (Kleiner Perkins), John Chambers (Cisco Systems) and Jim Barksdale (Netscape)—have been successful in sustaining the credits by focusing attention on the wrong question: Do the credits increase R&D?

The answer to that question is a resounding yes: Within four years of implementation, R&D was restored to within 10 percent of its 1964 peak (2.9 percent of sales). This ability of tax credits to increase R&D holds whether we look within the U.S. or across the globe—on average, a dollar of tax credits increases R&D by a dollar.

But we've been focused on the wrong question. When we consider whether the tax credits achieve their primary goal of stimulating economic growth, there are no studies quantifying the "social returns" to tax credits. We can attempt to answer that question by looking at nominal GDP growth. It has declined since the ERTA was implemented. So the answer would be an equally resounding no.

Read MoreMeet the CNBC RQ 50: The market's R&D all-stars

So why does TechNet continue to advocate for tax credits? Skeptics would argue these firms want the government to pick up the tab for R&D they would have done anyway. Indeed, according to the 2009 Government Accountability Office study, "The Research Tax Credit's Design and Administration Can Be Improved," these "windfalls" comprise 50 percent of tax credits.

R&D fact-check

So why aren't the tax credits delivering on their true goal? That's the critical question, and I argue it's because the economic logic of tax credits relies on two faulty assumptions:

Assumption # 1: Firms are investing optimally in R&D, so the only way to increase spending is to provide a tax incentive.

The truth: 56 percent of public firms are underinvesting in R&D relative to the amounts that would maximize their profits.

Assumption # 2: Increasing R&D in and of itself will increase growth.

The truth: GDP growth is driven by firms' R&D productivity rather than R&D spending. Mean RQ (the increase in revenue from a 10 percent increase in R&D) has dropped 67 percent since 1979.

Because tax credits rely on faulty assumptions, I believe we can eliminate tax credits and achieve even higher R&D merely by educating firms on how to identify their optimal R&D investment.

Read MoreA cheat sheet for the RQ 50 approach

Let's use McKesson—No. 1 in the 2014 CNBC RQ 50 ranking—as an example of an under-investing firm.

McKesson reported $456 million R&D investment in 2012, but they should be spending a lot more (assuming everything else about their markets and operations stays the same). The company's RQ score of 130.6 reflects what I call an "elasticity" of 0.276. It sounds complicated, but it's easier than you think to understand RQ.

Doing the math on McKesson

If McKesson increased R&D by 10 percent ($45.6 million), the company should see a revenue increase of 2.76 percent (remember the elasticity score of 0.276). That 2.76 percent increase is equal to $3.8 billion. Even with relatively slim operating margins (6.4 percent), that's still $198 million in incremental profits for McKesson. The company shouldn't need tax credits to increase its R&D. At its current P/E ratio of 35.4, shareholders would increase their wealth $7 billion if McKesson increased its R&D 10 percent.

Assuming a 10 percent increase in R&D across all 697 companies that I have identified as underinvestors, the net increase in R&D is $18.6 billion.

That's three times GAO estimate for the size of the R&D tax subsidy in 2009, which was $5.6 billion. How do we get three times the benefit of the tax credit without a government dime?

Read MoreThe market's R&D Hall of Fame

Doing the math on the R&D underinvestors

The conventional thinking is that each dollar of subsidy translates to a dollar of additional R&D (see the 2011 Ernst & Young report, "The R&D Credit: An Effective Policy for Promoting Research Spending"), which means near-$6 billion in R&D, and much less than the $18.6 billion net increase using the RQ methodology.

Even better, the corresponding shareholder wealth creation is $825 billion, meaning even the government potentially makes out on the change in R&D philosophy, to the tune of $124 billion in capital gains.

Can education make more progress than tax incentives on the real goal of economic growth? I think it can. First, by removing the tax credit we remove firms' incentives to invest in R&D that only has "social returns." This inherently increases the productivity of the remaining R&D, since the combined effects of social and private returns are 80 percent higher than social returns alone (see Griliches & Lichtenberg's 1984 paper, "Interindustry Technology Flows and Productivity Growth: A Reexamination").

Let's use Intel as an example of companies who over-invest in R&D. Intel reported $10.1 billion in R&D investment in 2012. That's 2.7 times what the company should be investing ($4 billion) if it were hewing to optimal R&D. Intel's RQ score of 94.8 reflects an elasticity of 0.078, meaning a 10 percent increase in R&D should increase revenue 0.78 percent. This sounds good, but the problem given Intel's financials is that the additional R&D ($1 billion) exceeds the expected additional revenue ($411 million). So the last thing Intel shareholders need is the company basing its R&D budget on tax credits. At the current P/E of 15 (2014 consensus earnings estimate), shareholders would increase their wealth $90 billion if Intel cut its R&D by roughly $6 billion.

Doing the math on Intel's lost-income opportunity

Intel is overspending by $6 billion, so if they cut that spending, it goes straight to net income. Multiplying the additional net income of $6 billion with the P/E ratio of 15 results in $90 billion. How I calculate the optimal R&D spend is more complicated, but it's one of the most valuable things you can do with RQ.

Intel is a single example, but even firms that are productive from an R&D standpoint can be in the R&D over-investment boat—36.4 percent, according to my RQ database. While getting these firms—497 in all—to spend optimally actually reduces R&D, multiplying the amount these firms overspend on R&D with their P/E ratio would generate $86 billion in higher profits that could be redeployed more effectively—remember, merely increasing R&D doesn't have an effect on GDP growth.

An end to the R&D tax-credit era

If the R&D emphasis were shifted from spending more to spending more effectively, firms would be able to improve their R&D productivity—ideally to the level enjoyed in 1970—by learning how productive they are, benchmarking against other firms, and then emulating the R&D best practices of higher RQ firms.

What are these best practices? While we're only beginning to learn what sets high RQ firms apart, we can generally say—based on dozens of interviews with high and low RQ companies and case studies from public documents—that R&D is more integral to their top-level strategy. High RQ firms tend to have more centralized R&D, as well as more consistent R&D spending. Firms with centralized R&D have a forty percent higher RQ score than those with decentralized R&D. R&D spending inconsistency also lowers RQ—a 10 percent increase in spending inconsistency decreases RQ by 8 percent. Identifying and quantifying the value of a broader set of R&D best practices is the goal of current research funded by the National Science Foundation.

These steps of benchmarking and best practice adoption should seem familiar. They are precisely how TQM restored manufacturing productivity and how hospital report cards improved morbidity. Improving R&D productivity requires a similar rethinking, but first, an end to the era of R&D tax incentives that haven't delivered.

By Anne Marie Knott, professor of strategy, Washington University Olin School of Business