Saturday, April 5, 2014

Endogenous Money IS-LM Model Calls for Paradigm Shift

This is an attempt to modify the IS-LM model under endogenous money regime.  Traditional IS-LM uses the IS curve to represent the equilibrium between Investment and Savings for various levels of interest rates.  The LM curve represents the respective equilibrium in financial markets between the level of money demanded (Liquidity Preference) and the money supply M under the assumption that central banks control M.  The intersection of IS and LM sets the level of output in the economy for a specific level of interest rates.  The model is used as a simplified illustration of short-term macro-fluctuations and the impact of active fiscal and monetary policy.

The IS-LM model has been extensively criticized and even dismissed for obvious flaws such as lack of micro-foundations and disregard for inflation.  However, probably the most fundamental critique has to do with the exogenous money assumption.  Traditional IS-LM assumes that under fiat money, central banks control the money supply hence LM represents both the supply and demand for money.  As I explain in my post on endogenous money, central banks have only indirect control over money supply via their control over reserves and short-term interest rates.  When short-term interest rates are greater than zero, such indirect control is highly effective giving rise to the illusion that money is exogenous.  However, at the lower zero bound that control is greatly diminished by the dramatic rise in money demand as evidenced by the rise in excess reserves and precipitous drop in money velocity since the Great Recession.  Endogenous money undercuts the constitution of the LM curve.  By extension, the idea that banks do not need savers to fund investments nor borrowers to accommodate savers puts the IS curve on very shaky grounds as well.

That is not to say that the IS-LM model should be discarded - quite the opposite.  The real issue is that the traditional framework obscures the true meaning of the IS and LM curves.  In this post, I will attempt to present an IS-LM model under endogenous money regime.  Having to account for endogenous money forces the IS-LM framework to reveal the true link between monetary and output aggregates operating in the economy at any point in time.  Furthermore, endogenous money IS-LM provides strong factual support for another much maligned macro-economic construct.  The quantity of money equation (QP = MV) has been dismissed by many as an exercise in tautology.  However, endogenous money IS-LM reveals the true meaning of money velocity (V) as the embodiment of base money demand.  

Endogenous IS-LM has significant implications for macro-economic policy.  It offers proof that the demand for money places a constraint to both fiscal and monetary policy, the evidence of such constraint being inflation, asset booms and busts and periods of long stagnation at the lower zero bound.  It calls for a paradigm shift in macro-economic policy.  First, rather than targeting interest rates, central banks should target a balance between the monetary base and the demand for asset money.  Second, central banks need new tools in the form of consumption and investment tax credits, which will enable policy makers to control the demand for money.  Such paradigm shift for the first time in history will put policy makers in a position to manage the business cycle without introducing distortions such as asset bubbles or inflation.

Next, let's take a look at the graphical representation of the model (Chart 1).

Chart 1

First you will notice that the horizontal axis represents quantity of money rather than aggregate output.  I am attempting to use money aggregates to represent both the real economy and financial markets.  Please refer to my post on endogenous money for an explanation of Time Preference, asset money demand (Ma) and transaction money demand (Mt).  To summarize, money demand has an exchange and asset motive.  I segregate the two motives by using the quantity of money equation and the average money velocity since 1959 as the exogenous input (Vt) in the identities below: 
M2 = Mt + Ma
Q P = GDP = Mt Vt
Transaction Money (Mt) represents the exchange motive, and it correlates directly to GDP (under a traditional IS-LM framework, Mt would be equivalent to the output Y).  The asset money motive is represented by the demand for asset money  (Ma).  At any point in time, Ma corresponds to economic agents who have negative time preferences and accordingly, are willing to hold money even if interest rates were at 0%.  The LM curve combines Ma (flat portion) with Mt (positively sloped portion).

The negatively sloped IS curve represents the money supply.  It combines the exogenous component of the money supply (Monetary Base), which is provided by the Fed and could be represented as a vertical line, and the negatively sloped demand for loans, which represents the endogenous money supply (M2-Monetary Base).  This could be a bit misleading since reserves, which are included in the Monetary Base, are not counted in M2.  However, you can think of the bank liabilities offsetting such reserves as being 100% monetized by central banks.  Only bank liabilities which are not monetized by the Fed can be counted as endogenous money supply.

The real economy is represented by the demand for transaction money (Mt) and the Time Preference.  Time Preference measures expected future returns as expressed in actual spending and investment decisions.  Financial markets are represented by the Interest Rate as set by the Fed Target, the demand for money (Ma + Mt = M2) and the supply of money (Monetary Base + Endogenous Money Supply = M2).

Another way to look at the model is to segregate the individual components (Chart 2).

Chart 2

From Chart 1 and Chart 2, it becomes immediately apparent that the Time Preference in the real economy and the Interest Rate in financial markets do not have to match.  I've made this observation repeatedly when discussing the business cycle and endogenous money demand.  Banks can meet the demand for loans by simply taking on a liability thus expanding the money supply (new loan assets are immediately offset by a deposit liability).  In a similar fashion, banks can accommodate the demand for savings by expanding  reserves (new deposit liabilities are immediately offset by new reserves).

In the example in Chart 1 the Fed has set the Fed Rate Target below the Time Preference by increasing the Monetary Base above Ma.  This causes the economy to expand from M20 to M2.  If the economy is below capacity, the monetary expansion will result in real economic growth.  If the economy is at capacity, the shift from M20 to M2 will result in inflation.  Please note that by increasing reserves, the Fed does not directly increase M2 (after all, reserves are not even counted in M2).  Instead reserve expansion acts to shift money demand to the right.  In effect, when the Fed buys securities, money holdings in the economy increase above the level the public would have demanded otherwise.  This pushes interest rates lower, which drives up loan demand and ultimately the money supply to M2.

The reverse scenario is also possible.  If the Fed were to set the Fed Rate Target above time preference, money demand will shift to the left contracting money supply and the real economy with it.  However, if you refer to my post on time preferences, where I derive a data set of time preferences over the last 55 years, you will see that this has never been the case at least during that period.

Another scenario, which until recently was relegated to the by-gone years of the Great Depression, has come back with a vengeance.  Since the Great Recession of 2008, we've been stuck at the lower zero bound.  Chart 3 illustrates ZLB - a condition where the Fed simply can't lower interest rates low enough as to push the economy back to full capacity.

Chart 3

As illustrated in Chart 3, a drop in time preferences results in a precipitous down-shift of the IS curve (dashed tan line).  At low time preferences, the demand for asset money (Ma) increases rapidly.  People attempt to satisfy their desire to hold cash by cutting spending and increasing savings basically shifting money from the Mt category into the Ma category. If the central bank does not react quickly to meet the excess demand for money, the economy will suffer a violent contraction represented by M20. Instead, what a central bank should do is expand reserves rapidly such that the excess demand for Ma is satisfied through a growing monetary base rather than a reduction in Mt.

By growing the monetary base and supplying excess reserves, a central bank can prevent declines in Mt and avoid the worst-case scenario.  However, it is unable to get Mt to grow and by extension the economy.  Any additional reserves simply end up in the Ma category shifting all points and solid curves to the right with no impact on Mt.  The reason Ma seems to have infinite capacity to grow at ZLB is because people expect negative returns and holding cash at 0% is definitely better than losing money. It is a simple arbitrage that can go on for a very, very long time.  Only when time preferences lift back up into positive territory would the IS curve shift back up resulting in Mt expansion.

It is notable that under the gold standard the monetary base was fixed and during downturns, central banks could not expand the money supply to meet excess Ma.  Instead, people attempted to meet their liquidity preference by cutting spending and liquidating assets (internal devaluation).  In a closed economy, this is a futile exercise, which simply initiates self-fulfilling deflationary pressures and violent contractions (point M20 on Chart 3).  In an open economy, internal devaluation can be effective as long other countries inflate.  However, if the downturn is global, internal devaluation becomes a race to the bottom as in the case of the Great Depression. 

Now, let's attempt to examine the historical record and derive evidence of the relationships as predicted by the model.  Below, I derive the individual components from widely available macro aggregates (all data from FRED).  Also, here is a link to the historical time preference data set I referred to earlier.  Please note that inflation acts to increase the real time preference - if you expect 2% real growth and 4% inflation, real time preference is 6% since the utility of future spending will be 6% lower than the utility of current spending.  Accordingly, all charts use real time preference.

Chart 4

First let's look at money demand (Chart 4).  To the right, I plot changes in US Asset Demand (Ma) to Time Preference.  The reverse relationship is quite clear.  Also, the experience post-2000 is quite notable.  The rapid increase in Ma with Time Preference approaching zero confirms the infinite liquidity theorem and the shape of the Ma curve.  To the left, I examine the relationship between the two money motives.  The reverse relationship is also quite clear.  Again, very notable is the 2009 Recession when nominal GDP and its derivative Mt were declining while the demand for money (Ma) was increasing rapidly.

Chart 5

Chart 5 above examines the relationship between Time Preference and Transaction Money Demand (Mt) for evidence of the down-shift effect due to active Fed policy.  The chart clearly identifies four periods of Fed easing which coincide with down-shift in Mt.  Also, post 2008 the data points become very compressed, which is a testament to constraints imposed by ZLB.  What is even more remarkable is the startling similarity between the chart to the left and the Fed Funds Rate (graph to the right) indicating that the Fed has a great degree of control over the LM curve.

Chart 6

Next let's study how the Fed's control over the LM curve translates to the IS curve.  The Fed already controls the Monetary Base.  The other component is the endogenous money supply (M2 - Monetary Base).  Chart 6 studies the relationship between transaction money demand and endogenous money supply.  Again, the relationship is very, very strong. It is not 100% predictive (slight deviations from the trend line to the right) suggesting a transmission mechanism, possibly interest rates but also potentially, nominal GDP, itself.  The direction of the causality needs to be studied more, but it is possible that it works both ways.  Nonetheless, the conclusion is that endogenous money supply and the demand for transaction money (Mt) are highly correlated, which provides the Fed with control of three out of the 4 components of the IS-LM model.  The missing piece is asset money demand (Ma).  When time preferences are positive, Ma in absolute terms is small giving the illusion that money is exogenous under the control of central banks.  However, when time preferences begin to approach zero or become negative, Ma becomes the dominant force lifting the veil and revealing the true endogenous nature of money. 

Yes, a central bank can determine the absolute level of M2, but it cannot determine the balance between Ma and Mt.  And here-in lies the constraint that the demand for asset money places on both fiscal and monetary policy.  If inflation expectations are high, changes in asset money demand will be plowed directly into the real economy driving up inflation pressures as was the case in the 1960's and 70's.  If inflation expectations are low, excess money supplies begin to churn in stock and real-estate markets giving rise to asset bubbles.  Finally, a dramatic rise in asset money demand will render "money printing" efforts by central banks akin to pushing on string with little impact on the real economy and inflation, which is where we find ourselves 5 years into the Great Recession.

The profound implication is that central banks have to target the monetary base.  Since endogenous money supply creates its own demand for transaction money (Mt) and vice versa, a mismatch between the monetary base and asset money demand will result in the negative effects I describe above.  The problem is that neither fiscal nor monetary policy is concerned with monetary aggregates. In the case of monetary policy, central banks target interest rates as to meet price stability goals.  As predicted, this has caused distortions in the form of inflation and asset bubbles (Chart 7).

Chart 7

Another implication is that the Phillips Curve  may not be describing causation between inflation and unemployment.  Rather, both inflation and unemployment are dependent on asset money demand.  When trust in the monetary base is strong such as under the gold standard, the Phillips Curve will hold.  However, the Phillips Curve will be undermined when trust in the monetary base as a safe asset is low (stagflation).  This is often the case in emerging markets, when increased appetite for safe assets is usually directed toward FX reserves limiting the capacity of easy monetary policy to affect the business cycle.

Endogenous money IS-LM also provides a clear explanation of the pro-cyclical nature of the gold standard.  By definition the monetary base was fixed, which introduced an inherent instability resulting in boom and bust cycles as the demand for asset money changed in response to changing time preferences.  Also, the Gibson Paradox is no longer a paradox.  Under a fixed base regime, changes in interest rates truly represent changes in time preferences.  Low interest rates mean low time preferences hence high demand for base money. Since money is denominated in itself, excess demand for base money over supply causes other prices to come down.  On the flip side, high interest rates mean high time preferences and low demand for base money.  The excess supply of base money leads to higher prices for everything else.

The zero lower bound presents its own challenges.  Even if a central bank were to supply the base money demanded by the public, it cannot induce real growth unless improving time preferences push up the IS curve. Some would argue that fiscal policy is more effective at ZLB.  Fiscal deficits act to shift the dashed tan line (Chart 3) to the right by expanding the demand for borrowing.  I would argue that fiscal and monetary policy are equally ineffective.  Again, we are dealing with a massive increase in asset money demand (the proverbial infinite liquidity preference).  The question is whether you want to meet that demand with money created as a result of fiscal deficits or excess reserves supplied by the Fed.  Both fiscal and monetary policy act to mitigate the economic damage due to rising asset money demand, but neither can compel the economy to break free from the lower zero bound unless we are talking about something on a super massive scale like WWII or QE I, II and III.

Instead, what policy makers need are new tools that can target the core problem, namely low or negative time preferences.  That's why I have repeatedly advocated for the hybrid approach to government intervention which calls for equipping central banks with consumption and investment tax credits.  Such credits give central banks unlimited capacity to increase the utility of current spending and the expected returns on current investment, which will give an instant boost to the Time Preference itself.  With a degree of control over time preferences, central banks will be able to control the demand for asset money, which to this day is the missing piece to macro-economic policy (the Euro-zone can reap even further benefits as I explain here).

Endogenous money IS-LM reveals the true link between real economic activity and monetary aggregates.  It provides the full picture of the enormous impact money has on our lives.  More importantly, it calls for a paradigm shift in policy response to the business cycles.  Rather than relying on outsized moves in interest rates, which could fuel asset bubbles, or the 2% inflation target as cushion against the lower zero bound, central banks have to target the monetary base and use precision tools when time preferences become negative.

No comments:

Post a Comment