Guest Contribution: “Rising US Real Interest Rates Imply Falling Commodity Prices”

Today, we present a guest post written by Jeffrey Frankel, Harpel Professor at Harvard’s Kennedy School of Government, and formerly a member of the White House Council of Economic Advisers. A version appeared earlier on his blog.

Real interest rates tend to have a negative effect on real prices of commodities: oil and gas, minerals, and agricultural commodities. One might hazard the prediction that US real interest rates are headed up, and therefore real commodity prices will decline.
The theory of the relationship between real interest rates and commodity prices is long-established. Personally, I like the “overshootingformulation of the theory. The econometric relationship is also pretty well-established, by statistical analyses that range from:

  1. simple correlations; to
  2. regressions that control for other important determinants such as GDP and inventories in a “Carry Trademodel; to
  3. high-frequency event studies that are not sensitive to the econometric problems of the regressions (namely, issues of causality and time series properties).

The simplest intuition behind the relationship is that the interest rate is a “cost of carrying” inventories. A rise in the interest rate reduces firms’ demand for holding inventories and therefore reduces the commodity price. Three other mechanisms operate, in addition to inventories. First, for a non-renewable resource, an increase in the interest rate increases the incentives to extract today, rather than leaving deposits in the ground for tomorrow. Second, for a commodities that have been “financialized,” an increase in the interest rate encourages institutional investors to shift out of the commodities asset class and into treasury bills. Third, for a commodity that is internationally traded, an increase in the domestic real interest rate causes a real appreciation of the domestic currency, which works to lower the domestic-currency price of the commodity.

I recently presented updated versions of my long-time research and commentary on this topic in two venues:

What does all this have to say about the future path of commodity prices as of August 2018? Even though the random walk is a good rule for prediction at a short-term horizon, I hazard a bolder guess at the medium horizon. The current US combination of loose fiscal policy (tax cuts and increased federal spending) and normalization of monetary policy (more increases in short-term interest rates from the Fed) suggests that real interest rates and the value of the dollar may go up. This monetary-fiscal mix is reminiscent of the Volcker-Reagan policy combination in 1984. The overshooting theory predicts that, as a result, real commodity prices are headed down.

This post written by Jeffrey Frankel.

40 thoughts on “Guest Contribution: “Rising US Real Interest Rates Imply Falling Commodity Prices”

  1. 2slugbaits

    a non-renewable resource, an increase in the interest rate increases the incentives to extract today, rather than leaving deposits in the ground for tomorrow.

    Of course, the one notable exception to Hotelling’s Rule seems to be the oil market. My guess (and it’s just a guess) is that Hotelling’s Rule breaks down in the oil market because a lot of oil is controlled by state entities whose internal discount rate exceeds the market interest rate. In other words, aging warlords are more concerned about living to see the sun come up tomorrow than they are concerned about optimally smoothing oil extraction according to some calculus of variations model they never understood anyway.

    1. Jeff Frankel

      To 2slugbaits:
      You are right that Hotelling’s Rule does not in practice work for oil (nor, I think, for lots of other commodities): The price does not tend on average to rise at the rate of the interest rate. But I would say the reason is that technological progress repeatedly expands the definition of what constitutes relevant deposits. Fracking is the best example of this process.
      That is the reason why I choose to base my model, not on the decisions by firms regarding how much to extract versus leave in the ground, but rather on the decisions by firms regarding what level of inventories to hold. It takes only two equations to get the “carry trade” model, in which the real commodity price depends negatively on the real interest rate (and also positively on convenience yield, negatively on inventory levels, and negatively on the risk premium).

      1. Steven Kopits

        I am not convinced that interest rates are a prime determinant of inventory levels, at least at a floor. I think minimum inventories are heavily influenced by operating considerations — the refinery doesn’t want to stock out, for example. In reality, an operator like a refinery is always looking to minimize inventory subject to operating needs, largely regardless of interest rates.

        I do think interest rates play an important role in the term structure of the futures curve, however. The cost of capital is a key determinant of total storage costs and therefore the differential between monthly futures contracts.

        1. pgl

          “I am not convinced that interest rates are a prime determinant of inventory levels, at least at a floor. I think minimum inventories are heavily influenced by operating considerations”

          Here we go again. His paper never denied that other factors drive inventory levels. Why is it that the amateur economists here think all models should be Y = F(X) as if X were the ONLY thing that influences Y?

      2. Steven Kopits

        And one more thought: I think an interesting problem is price evolution during a period of excess inventory and over-supply, which pertained from Jan. 2015 until July 2017. I still have no clue how prices are set under the circumstances.

  2. Bruce Hall

    I remember that during 2010 when I was building a new house in the bottom of the housing market, that lumber and copper costs were rising. U.S. interest rates and demand wasn’t the issue; it was China’s demand that was twisting commodity prices in the wind.

          1. Bruce Hall

            pgl, just eyeballing the chart from the St. Louis Fed with the one from Macrotrends, they appear quite similar with the Fed chart being smoothed considerably. Regardless, for the time in question (2Q10-2Q11) the information is basically the same.

            With regard to your second comment, I’d only refer you to what I wrote: I concede, in general, that interest rates do affect commodity prices… but there are significant anomalies. So, I don’t think we have any significant disagreement, do we? I mean, nothing in a vacuum is a perfect predictor of commodity prices. But I won’t take your snide comment as anything but a reaction to past differences.

        1. pgl

          Oh good grief. Yes there are lots of factors that drive commodity prices. But that does not mean real interest rates are irrelevant. I asked Peaky to take a basic Econ 101 class so he could distinguish between movements along curves v. shifts of curves. You might want to join him.

      1. JBH

        Menzie: Between these two variables, what is the R-squared and the sign on the estimated coefficient?

          1. JBH

            Menzie: Per the quarterly data in FRED. Global price of copper vs. 5-year TIPS yields. R-squared for level form is a mere 15%. (Which of course I knew before I posted.) R-squared for first differences an even worse 8%. (Anyone could see that coming!) Eliminate the six recession quarters, which anyone familiar with OLS realizes will heavily affect the regression results because they are big plus and minus outliers. Adjusted R-squared then falls below zero. And for good measure, log first differences of price of copper vs. 5-year TIPS yields gives an R-squared of zero also.

            Moreover, price of copper going up implies industrial strength and therefore inflation going up. Which implies TIPS yields going up. Yet the estimated coefficients in the above equations are all of the wrong sign! Which of course is hardly significant since the lack of statistical significance of the overall results means there really isn’t much of a meaningful relationship between copper and TIPS. And why should there be? Copper being just a hair on the tail of a gigantic dog.

          2. Menzie Chinn Post author

            JBH: Do…you…understand…why…one…estimates…in…first…differences? Is…the…coefficient…on…TIPS yields…statistically…significant and negative?

            I run regression of PPI for copper against 10 yr Treasurys minus SPF expected 10 year inflation, in first differences, 1992-2017. Estimated slope coefficient is -14, t-stat 3.6 using HAC robust standard errors, Adj-R2 = 0.16, DW = 1.40.

          3. pgl

            “R-squared for level form is a mere 15%.”

            JBH confuses statistical significance with explanatory power. Hey – we know there are lots of factors that drive commodity prices. A regression that included only one factor is not going to have a high R-squared. No one ever said it would.

          4. 2slugbaits

            JBH Adjusted R-squared then falls below zero.

            How did you manage to get an R-square less than zero? That’s a red flag for a misspecified model…or maybe you used Excel without a constant term.

        1. pgl

          “Adjusted R-squared then falls below zero. ”

          2slug questions how one got this. Hint – if you are using Microsoft as your stats package, note it has all sorts of issues. Get a real stats package if you want to play in this league.

  3. pgl

    Frankel, Jeffrey A. “Effects of Speculation and Interest Rates in a “Carry Trade” Model of Commodity Prices.” HKS Faculty Research Working Paper Series RWP13-022, January 24, 2014 (Revised from June 2013; and November 2013).

    I remember reading this when it came out. Excellent paper. When CoRev went off on his version of this idea, we should have required him to read it.

  4. Not Trampis

    interesting implications for our country down under if we can get over getting rid of PMs!!

  5. JBH

    Raw CRB commodity prices topped in March. March was the breakout month for trade negotiations. Commodity prices are falling because of the perception that global trade is in decline not because of interest rates. Given that and what’s happened on the margin since March, real rates are secondary or tertiary at best to the generalized near-term outlook. In fact, January through August saw virtually no variation in the real funds rate. Start -.46, end -.42. Yet copper broke trend in June, and has plunged steeply ever since. This is all about president Trump shutting off the high-speed gravy train that’s run from the US to the rest of the world for all too long. At a unique juncture like this, the meat will be found “high-frequency event studies not sensitive to the econometric problems of regression.”

  6. Steven Kopits

    I come out with an inverse relationship of real short term interest rates (3 mo) v 3 mo movements in the real (or nominal) WTI oil price, but the R-squared is only 0.08. Real interest rates do not tell us much about oil price movements.

    What’s more, nominal oil prices rise with nominal interest rates (just as they have in the last year), which makes sense, as both tend to be pro-cyclical. But again, the correlation is weak, with an R-squared of less than 0.01.

    The divergence between real and nominal interest rate correlations is problematic in the real world, because we can observe nominal oil prices and interest rates, but can only estimate real values ex-post and somewhat imprecisely.

    In practice, we see that oil traders are cognizant of exchange rates, which feed into oil prices very quickly — almost real time. Relative interest rates are, of course, a key determinant of exchange rates, but in practice, oil prices move on exchange rate news, much less on developments in interest rates.

    1. JBH

      Kopits Real short rates were pretty much relegated to the wastebasket by the dwellers of ZIRPLAND. Much better to ponder on the plethora of deeper more meaningful relationships revealed to Alice as she tripped through Wonderland.

      1. Steven Kopits

        I believe Prof. Frankel used three month T-bills in his analysis, so that’s what I used.

        1. JBH

          Stephen You may want to do the work to convince yourself there is not a shred of difference between the funds rate and the 3-mo bill rate when using monthly data for the oil market purposes you apply it to. Daily events analysis, of course, would be a different story.

          1. Steven Kopits

            First, it’s Steven with a ‘v’, as you can plainly see from reading my name literally hundreds of times.

            Second, I have no problem with using either the FF rate or the 3 mo rate. Again, I used the 3 mo rate because Prof. Frankel did, and I wanted to keep things reasonably apples-to-apples. In either case, I think the R-squared is small.

            The difficult part is not which nominal rate to use, but rather estimates of underlying real rates, which involve some subjective assessments of current inflation and which cannot be known in real time.

            If I were to critique the Prof Frankel analysis, it would be these items:

            1. Nominal oil prices are positively correlated with nominal interest rates. So if US rates go up, then expect oil prices to go up. But the correlation is weak, and the commonality runs through the business cycle, rather than necessarily from interest rates to oil prices. Having said that, low interest rates tend to support higher asset values, and commodity inventories are assets, after all.

            2. It’s not clear how good the real model really is. Here are some pairs of oil prices and real interest rates as I calculated them (first number is nominal WTI, second is real 3 mo interest rate):

            2013 Sept: $106, -0.8%
            2014 Nov: $76, -0.97%
            2016 Jan: $32, -0.89%
            2017 Sept: $50, -0.93%
            2018 Feb: $62, -0.85%

            You can see that a wide range of recent WTI prices are compatible with essentially similar real short term interest rates. So how much can we lean on short term real rates, really?

            3. As noted above, we can only observe nominal rates and prices, so the utility of a real model for decision-making purposes is limited.

            4. Current stocks of oil (to keep the discussion narrow) are a function of current supply / demand conditions, both of which are subject to very small short run elasticities. I don’t think interest rates have much to do with how much inventory oil markets players store in the short run. They store the difference between current demand and supply, which at the, say, 90-180 days horizon is largely independent of interest rates barring some catastrophic event.

            4. Interest rates do matter for the structure of the futures curve. Because producers hedge a significant part of their output and use the hedges to underpin bank loans to finance further exploration and production activities, interest rates can very much matter for future oil production. I think Goldman or one of the other big banks had a study on the impact of interest rates on US shale production a couple o years back. (Maybe it was PIRA, or maybe Citi.) It was material. Thus, we should be looking at the impact of interest rates with a lag of 6-12 months on the shale front, probably more for the IOCs. That’s, I think, the underlying linkage.

            5. I think there are more interesting problems to study. As I mentioned, the behavior of prices during conditions of both excess inventory and excess supply would make a very interesting research piece and a valuable contribution to our understanding of commodity price movements.

            In addition, I would be very interested to better understand the real movements of inventories depending on the slope of the futures curve. So, if you look at my weekly DOE report (, on slide 5, you can see the incentive to store oil (which does incorporate interest rates). Now, you can see the market believes inventories will draw all through next year. But here’s the question: If the incentive to store is -$1.00 / month, versus -$0.50 per month, does that mean inventory will drain faster? Put another way, can we make inferences about balances from the structure of the futures curve? Does a steepening curve imply larger surpluses or deficits? You would think so, but I don’t know that for a fact. That would be a very interesting economic analysis, I think.

  7. pgl

    Maybe Menzie can post on this variation of the Trump-Wilbur Ross swamp:

    Bad News for U.S. Papers, but Tariffs Are Paying Off for One Rock Capital
    Private-equity firm headed by a Washington and Wall Street veteran pushed for the tariffs on behalf of its North Pacific Paper and hope they are affirmed in a coming trade-commission vote

    When the Trump administration applied tariffs on imports of newsprint earlier this month, it brushed aside opposition from the Canadian government, the U.S. newspaper industry, printing companies, and a long list of lawmakers, including Democrats and Republicans. The tariffs, though, have been cause for celebration at private-equity firm One Rock Capital Partners LLC.

    1. Menzie Chinn Post author

      Steven Kopits: It’s analogous to our “deal” with N. Korea: They accelerate (estimated) production of fissile materials, and we pretend the problem has been solved. Here, we pretend we have a change in NAFTA, when we have only renegotiated components of the US-Mexico portion with hopes of leveraging Canada into a quick agreement. If not done within a very short time period, NAFTA “revision” is dead for the year; and dead perhaps longer if Democrats control the House come January.

      1. Steven Kopits

        My take is even worse.

        The ‘re-negotiated’ terms with Mexico have yet to be litigated in the US media. I suspect the terms will be picked apart when they are better understood.

        Otherwise, I agree with you.

    2. 2slugbaits

      Steven Kopits We don’t know a lot of the nitty gritty details, but based on what I’ve heard so far it’s primarily a return to where we were before Trump launched his foolish trade war, plus a little less free trade. For example, Trump’s NAFTA 2.0 effectively increases the amount of domestic content in autos, which is a step backward in terms of freer trade. It also has a sunset clause, which is not something you’d want to do if you’re really concerned about reducing long term uncertainty.
      It strikes me as a missed opportunity. There were some valid reasons for renegotiating around the edges; e.g., the original NAFTA was done before the digital age, so bringing NAFTA up to date technologically would have been a good thing, albeit not something with the kind of sex appeal that would appeal to Trump’s base.

      1. pgl

        I watched Trump on the phone. He could not even get the call right. Why on earth does anyone think he has negotiated some new trade arrangement. All show – no substance.

  8. Steven Kopits

    I am inclined to agree, Slugs.

    I personally think the sunset clause will prove a deal killer.

    1. pgl

      “There’s still lots of work to be done — the administration is calling the deal with Mexico a “preliminary agreement in principle,” and the U.S. is still hopeful Canada will sign on.”

      This says it all – we have no deal here. A lot of meaningless bluster and no substance.

    2. 2slugbaits

      Under the new deal, Mexico committed to “specific legislative actions” to recognize workers’ rights to collective bargaining…

      Does Gov. Scott Walker know about this? Apparently GOP politicians think collective bargaining agreements are great for Mexican workers, but bad for American workers.

      1. Steven Kopits

        Strange, isn’t it? It feels like we’re having a complete political inversion, similar to the period after Johnson’s civil rights legislation, which essentially sent southern Democrats over to the Republican side.

Comments are closed.