Nick Rowe on interest rates and inflation

Nick Rowe has an excellent twitter thread on interest rates and inflation:

Here I will focus on what I see as the most important parts of his analysis. (I pretty much agree with all his points, but sometimes it’s useful to have ideas reframed in a slightly different fashion.)

So how can we resolve this, and figure out which of those theoretical possibilities is right? 1. Empirically. But CB actions don’t give us a nice RCT. “Friedman’s Thermostat” says targeting inflation is like the *worst possible* experimental design. 18/n

NeoFisherian has a certain appeal because it’s generally true that rising interest rates are associated with rising inflation, and vice versa. In the New Keynesian model this can be explained as follows. When the natural interest rate is rising, the central bank usually raises the policy rate, but more slowly (and vice versa) As a result, the gap between the natural and policy rate widens, which makes monetary policy more expansionary.

The act of raising interest rates is generally a signal of disinflationary intent, but periods of rising interest rates are often inflationary (i.e. the 1960s and 1970s). As an analogy, the act of pressing the accelerator in my car tends to cause acceleration, but during periods when I press hard (going up a steep hill) the car is often slowing down.

If the Japanese suddenly adopt a policy of depreciating the yen at 1%/ month, and also commit to maintain this policy for many decades, the immediate impact will be much higher interest rates in Japan, and double digit inflation. So why is this (NeoFisherian) example not typical of periods where the central bank raises interest rates? Here’s Nick:

But this [NeoFisherian result] requires that everyone interprets the higher r as the CB’s signal of a higher inflation target, plus 100% CB credibility. Nope.

Most central banks use interest rates (not exchange rates) as a policy signal. Raising the interest rate target (fed funds target) is a signal that the central bank intends to take other actions to reduce inflation. (AFAIK, all central banks speak in this language.) Those other actions are either less supply of money (open market sales) or more demand for money (higher IOR or higher reserve requirements.)

Even in cases where the NeoFisherian result holds in the short run, as with the Swiss decision to revalue the franc in January 2011, there were actually two signals sent out. The revaluation led to expectations of lower Swiss inflation. The concurrent cut in Swiss interest rates was a signal intended to make the revaluation smaller, and hence reduce the amount of disinflation. This combined exchange rate/interest rate policy was disinflationary, despite the lower interest rates, but only because the impact of the exchange rate signal dominated.

So here are *my* ultimate things: 1. “Never reason from an interest rate change”. NeoFisherians illustrate one example of how this can go badly wrong. Nominal (even real) interest rates are a bad measure of the tightness/looseness of monetary policy.

That’s also my ultimate thing. Keynesianism is another example of how this can go badly wrong. Not ideal New Keynesianism, but actual, real world Keynesianism. The unfortunate tendency of actual Keynesian economists to assume that when a central bank has been cutting interest rates it has also been making monetary policy more expansionary.

As in 2007-08.

PS. Why does a 1% a month (expected) yen depreciation lead to inflation? Because of PPP. In addition, this announcement would immediately raise Japanese nominal interest rates 12% above US levels due to the interest parity condition. If we assume zero IOR in Japan, then this sort of high nominal interest rate would lead to a flight from the yen, and the Japanese monetary base would quickly fall from 140% of GDP to under 10%. The nominal monetary base would also have to be reduced sharply, in order to avoid even sharper currency depreciation and hyperinflation.

Try to explain that process with MMT!

PPS. Some commenters want me to condemn Biden’s continuation of Trump’s policies of bloated government spending and tariffs on China. OK, I condemn Biden’s economic policies.



37 Responses to “Nick Rowe on interest rates and inflation”

  1. Gravatar of Philo Philo
    12. April 2021 at 22:09

    Commentators can be so unfair: governments pursue so many suboptimal policies–you can’t be expected to condemn every one!

  2. Gravatar of Matthias Matthias
    12. April 2021 at 22:27

    I would actually like to see an MMT person predict what would happen, if Japan adopted that policy, and how they would explain it.

    (I’m not sure they would have the same prediction even?)

  3. Gravatar of postkey postkey
    13. April 2021 at 00:34

    “Warren B. Mosler #MMT
    Replying to
    These are, in general, so far out of paradigm/supported by erroneous assumptions and breaches of monetary operations logic, with quite a bit perhaps applicable to fixed exchange rates at best, and lots more. :(”

  4. Gravatar of Effem Effem
    13. April 2021 at 05:45

    Not saying this is the case for any developed country, but can a country paint itself into a corner such that there is simply no interest rate high enough to stop an inflationary spiral as the necessary interest rate would have harmful side effects?

  5. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    13. April 2021 at 06:48

    The MMTers (Bowery Boys), don’t know a debit from a credit.

    The biggest MoM jump in the CPI since June 2009 and biggest YoY jump since Aug 2018 (interest rate inversion).

    Huge differences between the taper tantrum and today’s economy. With the expansion of savings’ products in 2013 (the FDIC’s reduction in unlimited deposit insurance), real rates, and R *, rose. This time, real rates and R * fell (i.e., breakevens’ rose). In 2013 the U.S. $ rose. This time the U.S. $ fell.

    Interest rate spreads were the same:

    If the FED hadn’t continued to follow a very contractionary money policy in 2013, the acceleration in the economy would have been sustainable. With the expansion of money products this time, the FED will have to remain very easy with no interest rate inversion.

    There is no chance for inflation to subside until 2/22.

    –Michel de Nostredame

  6. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    13. April 2021 at 06:54

    There isn’t a “base” effect (and no longer a “SWEET SPOT”). Inflation will remain virulent.

    For the FED to support the economy, for the short-term roc in money flows to remain robust (or even neutral), it will have to remain virulent, i.e., we necessarily have FAIT.

    Let me remind you that lending/investing by the Reserve and commercial banks is inflationary and has a negative economic multiplier (is harmful), whereas lending/ investing by the nonbanks is non-inflationary and has a positive multiplier effect (a velocity relationship).

  7. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    13. April 2021 at 07:13

    “Disintermediation is made in Washington”

    See Barron’s:
    1) “Forgotten Man? Washington Again Is Threatening to Penalize the Thrifty” Jun. 6, 1966
    2) “Up the Down Staircase, The New Economics Doesn’t Know Whether It’s Coming or Going” Sept. 26, 1966
    3) “Ceiling Zero. The U.S. Must Take the Lid Off Money Rates” Nov. 26, 1967
    4) “Men and Money, Savers of Modest Means Deserve a Decent Return” Jan. 19, 1970
    5) “Q Marks the Spot. All Ceilings on Interest Rates Should Be Lifted” Dec. 28, 1970
    6) “Maximum Mischief, Ceilings on Interest Rates Must Go” Mar. 13, 1973
    7) “Supreme Interest. The Banking Agencies Have Finally Done Something Right” Jul. 23, 1973
    8) “No More Wild Cards, Congress Has Dealt Savers Out of the Money Game” Oct. 2, 1973
    9) “Poor Joe DiMaggio. It No Longer Pays to Save at the Bowery”” Sept. 22, 1975

    Funny – because the largest volume of disintermediation occurred during the GFC. The FED’s monetary transmission mechanism, interest rate pegs, is non sequitur.

  8. Gravatar of ssumner ssumner
    13. April 2021 at 07:47

    Matthias, MMTers don’t even believe than central banks CAN control inflation.

    Effem, You can always stop a spiral, but there may be side effects in terms of unemployment.

  9. Gravatar of art andreassen art andreassen
    13. April 2021 at 11:08

    Scott: Since American economists can’t fathom the thought that trade has any impact on employment it is no wonder that trade hasn’t entered into your discussion. A decline in the yen would increase the cost of all Japanese imports especially commodities which are priced in dollars. The depreciation would make Japanese exports cheaper increasing their demand which, along with the drop in Japanese demand for imports, would drive up Japanese employment and thus Japanese wages. Thus an increase in inflation.

  10. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    13. April 2021 at 14:13

    DXY to drop to 72 by 2023.

  11. Gravatar of Michael Sandifer Michael Sandifer
    14. April 2021 at 06:26


    Do you think nonminal short-term interest rates can sustainably be above the present NGDP growth rate, without lowering the NGDP growth rate? That’s another way of asking whether you think it ever makes sense that the netural rate should be above the NGDP growth rate.

  12. Gravatar of Michael Rulle Michael Rulle
    14. April 2021 at 06:47

    Powell will be facing a test with regard to his “rolling 2% inflation rate” or AIT. He has all but explicitly said to expect inflation to rise above 2% to maintain the target average. I wish he were a bit more explicit (we will not get NGDP targeting any time soon) as to how to assess his AIT policy. I think the markets understand better, so I won’t expect a freak out over a 2.7% increase for example. But still, I think being more specific (it cannot be precise) would be helpful.

    Your mockery of MMT is of course deserved. But your comment on Yellen and Climate Change does demonstrate how bizarre the Dems are willing to at least sound—-but maybe it is worse.

    There is far more than a zero percent chance an MMter could be named head of them Fed. Then what?

    Which is why your true but lame “I also condemn Biden’s economic policies”—-as if it need not be more persuasive——is a weak statement. You are a thought leader, as you really do know——but for some reason—for the 46th time——I tell you it is important to be more than a wise guy.

  13. Gravatar of Carl Carl
    14. April 2021 at 07:31

    I was looking at your short course on money links to review what you say about inflation and thought you might want to revise your intro,, particularly the line “Real shocks do not cause big jumps in unemployment.” Perhaps you need to make an exception for pandemics.

  14. Gravatar of ssumner ssumner
    14. April 2021 at 09:47

    art, Not sure what point you are making. How does your comment relate to my post?

    Michael Sandifer, There are times when the neutral rate is above the NGDP growth rate, but more recently it’s usually below.

    Michael Rulle, Some people are impossible to satisfy.

    You said:

    “There is far more than a zero percent chance an MMter could be named head of them Fed.”

    Yeah, a 0.0000001% chance.

    Carl, Thanks, I should change that.

  15. Gravatar of Michael Sandifer Michael Sandifer
    14. April 2021 at 14:54


    Under what circumstances in the US has the neutral interest rate been above the NGDP growth rate? And, this was sustained?

  16. Gravatar of Mark Z Mark Z
    14. April 2021 at 15:08

    Didn’t you predict Biden would *reverse* protectionist policies on China because of how much Democrats love free trade, and didn’t those of us (or just I) who pay greater attention to revealed preferences (elected officials) than stated preferences (polling results on vague questions) question this prediction?

  17. Gravatar of ssumner ssumner
    14. April 2021 at 16:02

    Michael, Perhaps in the 1980s? I’m not sure.

    Mark, I find it amusing that people “remember” me saying things that I never said.

    I did say Biden would be a lousy president, FWIW.

  18. Gravatar of Michael Sandifer Michael Sandifer
    14. April 2021 at 17:22


    I’ve seen examples of the short-term nominal rates being above the NGDP growth rate, like in Turkey, for example, presumably due to some combination of default risk and having a less-developed economy. But, in the US, it seems that the Fed Funds rate doesn’t stay above the NGDP growth rate for long, before NGDP growth slows or reverses:

    It’s tempting for me to see this as support for the idea that, at least in the case of the US, that the neutral rate should equal NGDP growth, when the money supply growth sustainably outstrips money demand. And if that’s true, one is nearly halfway toward accepting that the neutral rate should also equal NGDP growth in equilibrium, and will be below the nominal rate by a factor of twice the gap between the nominal rate and the NGDP growth rate, when the nominal rate is below the NGDP growth rate and money is tight. Of course, this assumes that the neutral rate moves 1-to-1 in an inverse fashion to changes in the value of the dollar versus output.

    But, if this is true and so simple, why hasn’t it been recognized previously? Could the path dependence of the development of macro theory be a big reason, with the single representative agent models leading to dead ends?

    On the other hand, the r* graph I produce based on this idea looks plausible, especially considering that it seems the US has very rarely had unemployment at or below it’s natural sustainable rate since ’62. Perhaps that only occurred during the Great Inflation, at points when the economy periodically overheated?

    And let’s not forget about the equity premium/risk-free rate puzzle, which seems to be neatly explained if the triple equilibrium condition I proposed holds. That is, the nominal rate = NGDP growth rate = S&P 500 earnings yield, in equilibrium, sans real shocks, and in absence of significant government default risk. Is it a coincidence that if we increase the average NGDP growth rate and S&P 500 discount rate by the difference in the average GDP growth rate and the average 1 year Treasury rate, that the equity premium comes extremely close to disappearing entirely?

  19. Gravatar of Michael Sandifer Michael Sandifer
    14. April 2021 at 17:30

    Oh, and let’s also not forget that the nominal one year rate stayed pretty close to the NGDP growth rate during the Great Moderation, as seen in the graph above.

  20. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    15. April 2021 at 07:07

    Interest is the price of loan funds. The price of money is the reciprocal of the price level. So R * is a Doppelgänger.

  21. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    15. April 2021 at 07:16

    The Wicksellian natural rate – “a real short-term rate that makes output equal to natural output with constant inflation” is bogus.

    Investment “hurdle rates” are idiosyncratic. Business expenditures depend largely on profit-expectations, and favorable profit-expectations depend primarily on cost/price relationship of the recent past and of the present. Cost/price relationships are crucial, and they are particular; they cannot be adequately treated in terms of broad-aggregates or statistical weighted “averages”.

  22. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    15. April 2021 at 07:27

    Leland James Pritchard, Ph.D., Economics Chicago 1933

    (A) The commercial banks create new money (in the form of demand deposits) when making loans to, or buying securities from the non-bank public; whereas lending by financial intermediaries simply activates existing money.

    (B) Bank lending expands the volume of money and directly affects the velocity of money, while intermediary, non-bank lending, directly affects only the velocity.

    (C) The lending capacity of the commercial banks is determined by monetary policy, not the savings practices of the public.

    (D) The lending capacity of non-banks, financial intermediaries, is almost exclusively dependent on the volume of monetary savings placed at their disposal. The commercial banks, on the other hand, could continue to lend if the public should cease to save altogether.

    (E) Financial intermediaries, NBFIs, lend existing money which has been saved, and all of these savings originate outside the intermediaries; whereas the commercial banks lend no existing deposits or savings: they always, create new money in the lending and investing process.

    (F) Whereas monetary savings received by financial intermediaries originate outside the intermediaries, monetary savings held in the commercial banks (time deposits and the saved portion of demand deposits) originate, with immaterial exceptions, within the commercial banking system. That is demand deposits, non-interest bearing deposits, constitute almost the exclusive net source of time deposits, or interest bearing deposits.

    (G) The financial intermediaries can lend no more (and in practice they lend less) than the volume of savings placed at their disposal; whereas the commercial banks, as a system, can make loans (if monetary policy permits and the opportunity is present) which amount to several times the initial excess reserves held.

    (H) Monetary savings are never transferred from the commercial banks to the intermediaries; rather are monetary savings always transferred through the intermediaries. The funds do not leave the payment’s system.

  23. Gravatar of Michael Sandifer Michael Sandifer
    15. April 2021 at 07:33

    Spencer Bradley Hall,

    Such a distinction isn’t important for the point I’m making. I accept the market monetarist view that interest rates aren’t determinative, but instead reflect the supply and demand for money.

    My first principle analysis starts at changes in the value of the unit of currency versus economic output. An unmet 2% increase in the demand for dollars causes the value of a dollar to rise 2% versus expected NGDP, and hence will slow NGDP and lower the neutral interest rate proportionally. Wages adjust more slowly than output, leading to much unnecessary unemployment.

    In a sense, real interest rates don’t change. Only the value of the unit of currency versus output changes.

  24. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    15. April 2021 at 08:52

    re: “instead reflect the supply and demand for money”

    That’s wrong. An increase in the supply of loan funds could result from a transfer of funds through the nonbanks, + loan funds, but no change in the money supply (a velocity relationship).

    Danielle Dimartino Booth’s in her book gets it backwards too: “Fed Up”, pg. 218

    “Before the financial crisis, accounts were insured up to the first $100,000 by the FDIC. That limit kept enormous sums *in the shadow banking system* [sic].

  25. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    15. April 2021 at 08:58

    The transactions concept of money velocity, Vt, has its roots in Irving Fisher’s truistic “equation of exchange”: P*T = M*Vt, where (1) M equals the volume of means-of-payment money; (2) Vt, the transactions rate of turnover of this money; (3) T, the volume of transactions units; and (4) P, the average price of all transactions units.

    The “econometric” people don’t like the equation because it is impossible to calculate P and T. Presumably therefore the equation lacks validity. Actually the equation is a truism – to sell 100 bushels of wheat “T”, at $4 a bushel “P” requires the exchange of $400 “M” once, or $200 “Vt” twice, etc.

    The real impact of monetary demand on the prices of goods and services requires the analysis of “monetary flows”, and the only valid velocity figure in calculating monetary flows is Vt. Milton Friedman’s income velocity, Vi, is a contrived figure (Vi = Nominal GDP/M) WSJ, Sept. 1, 1983. It is a “residual calculation – not a real physical observable and measureable statistic.” The product of M*Vi is obviously N-gDp.

    So where does that leave us? in an economic sea without a rudder or an anchor. A rise in N-gDp can be the result of (1) an increased rate of monetary flows M*Vt (which by definition the Keynesians have excluded from their analysis), (2) an increase in R-gDp, (3) an increasing number of housewives selling their labor in the marketplace, etc.

    The income velocity approach obviously provides no tool by which we can dissect and explain the inflation process.

    To the Keynesians, aggregate monetary demand is nominal-GDP, the demand for services (human) and final goods. This concept excludes the common sense conclusion that the inflation process begins at the beginning (with raw material prices and processing costs at all stages of production) and continues through to the end…

  26. Gravatar of ssumner ssumner
    15. April 2021 at 09:10

    Michael, You said:

    “It’s tempting for me to see this as support for the idea that, at least in the case of the US, that the neutral rate should equal NGDP growth,”

    Why? I don’t see the logic here. Because two variables are similar, they should be exactly equal?

  27. Gravatar of Michael Sandifer Michael Sandifer
    15. April 2021 at 10:21


    No, I have multiple reasons for thinking the yield on Treasuries should equal the NGDP growth rate (and yield on capital), in monetary equilibrium, sans real shocks and relatively high risk of government debt default. Here they are, enumerated:

    1. It started as an a priori concept that, assuming positive and negative real shocks are random, the risk of holding Treasuries and hypothetical securities that paid dividends proportional to NGDP growth should be equal and opposite. And, I was aware of the equity premium/risk-free rate puzzle and thought there might be a connection.

    2. Then, I noticed that one year Treasury rates pretty closely matched YOY NGDP growth during most of the Great Moderation.

    3. I then noticed that average NGDP growth since 1962 only differed from the average S&P 500 discount rate by 0.2%, but average S&P 500 index appreciation was 1.4% higher, indicating an imbalance. Under stable monetary policy, such as NGDP level targeting, I would expect the S&P 500 index to appreciate at the same mean rate as NGDP, and in-step with the earings yield, which is also the discount rate. I can explain more about why this is, if needed.

    I then performed a calculation, revealing that if the mean earnings yield on the S&P 500 increased by the difference between the mean NGDP growth rate and the mean 1 year Treasury rate, the imbalance is resolved with high precision. There would have been the same result as if one had level targeted NGDP.

    4. Just applying this approach since 2005, for example, NGDP growth would have averaged 5.3% with far less volatility, which is about what we’d want under NGDP targeting, eh?

    5. And as pointed out, NGDP growth tends to go down and usually turns negative when the 1 year Treasury rate exceeds the NGDP growth rate.

    Even if you don’t buy the explicit and implicit theory here, there’s no denying that monetary policy using my model would have been vastly better than what actually occurred.

  28. Gravatar of Michael Sandifer Michael Sandifer
    15. April 2021 at 11:11

    And if my model were applied since 1962, the average NGDP growth rate would have been ~4.8%, with far less volatility.

  29. Gravatar of Michael Sandifer Michael Sandifer
    15. April 2021 at 14:02

    I should be clear to say that my model would have produced results similar to that of a 5% NGDP level target since 1962, if you assume the model works.

    My point is to point out that this in perhaps an unlikely mere coincidence, as is the closeness of the mean NGDP growth rate to the mean S&P 500 yield, and that the gap between those rates and the mean Treasury yield since 1962, if treated as subtractive from optimal NGDP, would eliminate the equity premium puzzle, and equalize S&P 500 earnings yields, earnings growth, and the appreciation of the index.

  30. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    16. April 2021 at 06:38

    Bernanke in, BIS Review 49/2007 “wave of innovation”, stated that:

    The rapid pace of financial innovation creates challenges for policymakers.

    That shifting changes the relationship of “yield on Treasuries should equal the NGDP growth rate”.

  31. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    16. April 2021 at 06:44

    The BEA uses quarterly accounting periods for R-gDp and the deflator. The accounting periods for gDp should correspond to the specific economic lag, not quarterly (apples to oranges).

    Because the lags for gDp data overlap Roc’s in M*Vt, the statistical correlation between the two is somewhat degraded. However the statistical correlation between Roc’s in M*Vt, & for example, the bond market is unparalleled…

  32. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    16. April 2021 at 06:49

    And we knew this already:

    In 1931 a commission was established on Member Bank Reserve Requirements. The commission completed their recommendations after a 7 year inquiry on Feb. 5, 1938. The study was entitled “Member Bank Reserve Requirements — Analysis of Committee Proposal” its 2nd proposal: “Requirements against debits to deposits”

    After a 45 year hiatus, this research paper was “declassified” on March 23, 1983. By the time this paper was “declassified”, Nobel Laureate Dr. Milton Friedman had declared RRs to be a “tax” [sic].

    We can now expect the FED to lose control of the money stock in the longer-term, as the FED reduced the reserve requirement ratio to zero percent across all deposit tiers, effective March 26, 2020.

  33. Gravatar of Michael Sandifer Michael Sandifer
    17. April 2021 at 01:14

    Spencer Bradley Hall,

    You replied:

    ‘That shifting changes the relationship of “yield on Treasuries should equal the NGDP growth rate”.’

    All I care about is whether a model fits empirically. Then, build deeper theory that fits the empirical results.

    I could certainly be missing something, but it increasingly looks to be like I’m right. If I am, macroeconomics becomes something much more like an exact science, taking it out of the dark ages.

    Sure, the monetary policy my approach recommends is no better than NGDP level targeting right now, but it could be superior to it if we enter an era of increasing rates of increasing productivity, for example. That may or may not happen in my lifetime.

    Also, my approach offers good ways to target NGDP, such as targeting the nominal growth of the S&P 500 index, one year out.

  34. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    18. April 2021 at 08:25

    Since as Dr. Philip George says:

    So your demand function might work for a long while.

  35. Gravatar of Michael Sandifer Michael Sandifer
    18. April 2021 at 17:38

    Spencer Bradley Hall,

    I’m considering studying for a Masters in econ, but still haven’t decided. That said, it would probably be the best thing to do if I actually want to see if this could lead to a published paper one day.

    But, I’m actually more interested in going beyond publishing a peer-reviewed paper, and seeing how these ideas influence the development of my Exact Macro software. Much better than just publishing a paper, though necessary, is demonstrating the value of the model through software that actually has utility. I think this is the future of economics, and science in general.

    If I’m wrong, it’s no big deal. If I’m right though, this is a hell of an approach to macroeconomics.

  36. Gravatar of Michael Sandifer Michael Sandifer
    18. April 2021 at 17:42

    One possible application of this model is to calculate the possible upside for major stock indexes, like the S&P 500, after economic shocks. Also, it might improve market-based forecasting in various ways.

    The more seeming evidence there is for the equilibrium conditions I propose, the more confidence I have to move forward and develop applications. I’m now working on more comprehensive equations to better explore and encapsulate these ideas, so that they can be applied.

  37. Gravatar of Michael Sandifer Michael Sandifer
    18. April 2021 at 17:48

    Also, best case scenario, if the software works one day, I could develop a monetary policy dashboard for central banks, that could be used with and without discretion. I think monetary policy should be run by software, rather than committees, so it might be easier to sell a government somewhere on a rules-based approach with software if the software exists.

Leave a Reply