Monday, November 24, 2014

Models Behaving Badly

LONDON – “Why did no one see the crisis coming?” Queen Elizabeth II asked economists during a visit to the London School of Economics at the end of 2008. Four years later, the repeated failure of economic forecasters to predict the depth and duration of the slump would have elicited a similar question from the queen: Why the overestimate of recovery?

Consider the facts. In its 2011 forecast, the International Monetary Fund predicted that the European economy would grow by 2.1% in 2012. In fact, it looks certain to shrink this year by 0.2%. In the United Kingdom, the 2010 forecast of the Office for Budget Responsibility (OBR) projected 2.6% growth in 2011 and 2.8% growth in 2012; in fact, the UK economy grew by 0.9% in 2011 and will flat-line in 2012. The OECD’s latest forecast for eurozone GDP in 2012 is 2.3% lower than its projection in 2010.

Likewise, the IMF now predicts that the European economy will be 7.8% smaller in 2015 than it thought just two years ago. Some forecasters are more pessimistic than others (the OBR has a particularly sunny disposition), but no one, it seems, has been pessimistic enough.

Economic forecasting is necessarily imprecise: too many things happen for forecasters to be able to foresee all of them. So judgment calls and best guesses are an inevitable part of “scientific” economic forecasts.

But imprecision is one thing; the systematic overestimate of the economic recovery in Europe is quite another. Indeed, the figures have been repeatedly revised, even over quite short periods of time, casting strong doubt on the validity of the economic models being used. These models, and the institutions using them, rely on a built-in theory of the economy, which enables them to “assume” certain relationships. It is among these assumptions that the source of the errors must lie.

Two key mistakes stand out. The models used by all of the forecasting organizations dramatically underestimated the fiscal multiplier: the impact of changes in government spending on output. Second, they overestimated the extent to which quantitative easing (QE) by the monetary authorities – that is, printing money – could counterbalance fiscal tightening.

Until recently, the OBR, broadly in line with the IMF, assumed a fiscal multiplier of 0.6: for every dollar cut from government spending, the economy would shrink by only 60 cents. This assumes “Ricardian equivalence”: debt-financed public spending at least partly crowds out private spending through its impact on expectations and confidence. If households and firms anticipate a tax increase in the future as a result of government borrowing today, they will reduce their consumption and investment accordingly.

On this view, if fiscal austerity relieves households of the burden of future tax increases, they will increase their spending. This may be true when the economy is operating at full employment – when state and market are in competition for every last resource. But when there is spare capacity in the economy, the resources “freed up” by public-sector retrenchment may simply be wasted.

Forecasting organizations are finally admitting that they underestimated the fiscal multiplier. The OBR, reviewing its recent mistakes, accepted that “the average [fiscal] multiplier over the two years would have needed to be 1.3 – more than double our estimate – to fully explain the weak level of GDP in 2011-12.” The IMF has conceded that “multipliers have actually been in the 0.9 and 1.7 range since the Great Recession.” The effect of underestimating the fiscal multiplier has been systematic misjudgment of the damage that “fiscal consolidation” does to the economy.

This leads us to the second mistake. Forecasters assumed that monetary expansion would provide an effective antidote to fiscal contraction. The Bank of England hoped that by printing £375 billion of new money, ($600 billion), it would stimulate total spending to the tune of £50 billion, or 3% of GDP.

But the evidence emerging from successive rounds of QE in the UK and the US suggests that while it did lower bond yields, the extra money was largely retained within the banking system, and never reached the real economy. This implies that the problem has mainly been a lack of demand for credit – reluctance on the part of businesses and households to borrow on almost any terms in a flat market.

These two mistakes compounded each other: If the negative impact of austerity on economic growth is greater than was originally assumed, and the positive impact of quantitative easing is weaker, then the policy mix favored by practically all European governments has been hugely wrong. There is much greater scope for fiscal stimulus to boost growth, and much smaller scope for monetary stimulus.

This is all quite technical, but it matters a great deal for the welfare of populations. All of these models assume outcomes on the basis of existing policies. Their consistent over-optimism about these policies’ impact on economic growth validates pursuing them, and enables governments to claim that their remedies are “working,” when they clearly are not.

This is a cruel deception. Before they can do any good, the forecasters must go back to the drawing board, and ask themselves whether the theories of the economy underpinning their models are the right ones.

Read more from our "In Keynes's Footsteps" Focal Point.

  • Contact us to secure rights


  • Hide Comments Hide Comments Read Comments (14)

    Please login or register to post a comment

    1. CommentedMatthew McCarthy

      I agree with the "Models behaving badly because the systems the model is based on have changed." I don't agree with the follow on comment "The goalposts have moved."

      The problem is in the idea that fixed models can mimic reality when reality is constantly changing. That error leads to statements like "This implies that the problem has mainly been a lack of demand for credit..." which is entirely false. Ask anyone applying for a mortgage or home equity loan how easy it is to get credit. The answer will be it is ridiculously hard. A few years ago a mortgage could be available in days with no downpayment and sometimes without proof of employment, where now a mortgage with 70% downpayment takes 2 months of paperwork - but people are still doing it. The money being retained in banks is not "a sign of low demand for credit", the demand for credit is real people requesting credit. The money retained in banks is just money retained in banks - the next step is a model which may or may not be appropriate depending on current circumstances. In today's circumstances it can actually be money purchasing government bonds at 3% with no liabilities and no maintainance required. Since it is being purchased with government loans at an effective 0% it is a reasonable investment. Credit demand is irrelevant to banks retaining money because there are risk free ways to use the money instead of lending. Expecting a relationship between money retained and credit demand is not a valid model.

    2. Portrait of Pingfan Hong

      CommentedPingfan Hong

      The problem is not that most economists underestimate the fiscal multiplier, but they misunderstand the fiscal multiplier. In the real economy, a fiscal multiplier is not a structural parameter at all: the relationship between a change in fiscal policy and a change in GDP is not fixed, but varying over time in response to changes in a number of other variables, such as investors' risk aversion, consumers' confidence, and output gaps. Assuming a constant fiscal multiplier and making projection and policies based on this assumption can be deadly wrong.

        CommentedRichard Potter

        Precisely. Like unseasoned bond traders using duration alone to forecast changes in bond prices due to interest rate movements, these macro forecasters ignore the concept of "convexity", i.e. the multiplier is a non-linear function, and not a constant.

        I wonder if Keynes realized that as well? Any thoughts, Lord Skidelsky?

    3. CommentedDaniel Tanner

      penetrating !
      "Because they're sheep, and sheep get led." - from movie "Wall Street"
      You should know, all the society systems are self-affecting.

    4. CommentedZsolt Hermann

      Models behaving badly because the system the models are based on have changed.
      The goalposts have been moved.
      Although globalization is not new, international business and economic cooperation, or even financial unions are not new, but we never had such an integral, interdependent system as we do now.
      Despite the multi-level connections before individuals, nations or companies still functioned independently, separated from each other as if there was a clear space around them, there were new, relatively untouched markets, workforce to tap into.
      Today the whole system is locked into one network, there cannot be any independent calculations, projections as even relatively small, relatively insignificant elements can change the overall balance completely. Everybody is moving together tied into the same net.
      We need completely new methods, a completely new ideology for all human interaction, one which always looks at the total sum of the system, the overall direction the system is moving towards.
      Only such plans, moves will succeed that are going alongside the best interest of the whole, regardless of individual, national, or company interests.
      The same way any negative input or direction would backfire with multiple force.
      Even today we can watch these dynamics already working.

        CommentedEdward Ponderer

        "Models behaving badly because the system the models are based on have changed. The goalposts have been moved."

        Precisely! I was also was a sucker for this twice, about 3 years apart, just like the economists -- once as an undergraduate in a course in quantum mechanics, and once as a graduate student in a course studying microwave devices, and I hope never again. The key in my cases, and the economists cases on a much grander scale, is to miss that models are local approximations on many scales. Constants are really locally slow changing variables, variables are functions of other variables, linear equations are really nonlinear equations, and nonlinear equations are really differential equations. All is built on shifting sands and the future goes into the clouds. It ends in deterministic chaos.

        Now the answer is to course correct. But the problem is there are too many dimensions, to many factors, many we can detect till -- perhaps -- after the fact.

        The key is that these are all about our ever more global world, where the interdependent coupling grows in strength and complexity with every passing hours. Second-order and tertiary effects, cross-couplings (e.g., human economy and environment), etc., continues to shift the rules. The problem is growing,, embryonic - and so is the solution.

        The only computer capable of judging and balancing this complex of human interactions is the whole of humanity itself. And it must be realized that an embryo is always born into an environment where it can now longer survive -- except in its fully newborn form.

        What form is that? Something beyond just cells phones and Internet, Facebook and Twitter, and the information superhighway in general. But there is the beginning of a sense of greater intelligence, emotion, nervous system and body here.

        The key in natural evolution is always crisis and altruistic response. Social networks of individuals that become something greater.

        Consider this Google Tech Talks Youtube video: "Learning from Bacteria about Social Networks," Google Tech Talk on Youtube, Note that the manner that food is detected and located, for example, is impossible from the informatin available to small subsets of individual, but as an interconnected whole, the distributed statics gel at a fantastic rate of speed.

        Try to project this concept from 70 billion bacteria, to a "megacolony" of 7 billion human beings. Understand how we can be come a true model, and a true answer. Yes, Humanity itself can do it -- if it is whole and in balance. We have all the means. What it takes now is integral education, and developing through it a sense of mutual responsibility -- to act in mutual guarantee.

    5. CommentedManfred Dix

      Lord Skidelsky says at the end of the article "There is much greater scope for fiscal stimulus to boost growth...". Really? Greece has a debt to GDP ratio of close to 200%. More scope for borrowing money? How about other countries, like Italy or Portugal with close to 100% of ratios. The good Lord seems to forget that the Eurozone countries do not print their own money.
      I guess the states in the United States which the 1840s went bankrupt at the time had it all wrong. They should have borrowed even more, for "fiscal stimulus" to "enhance growth". [Who would have loaned them the money? That question, of course, goes unanswered by Lord Skidelsky and others.]
      And such thinking passes for high quality economics being taught at prestigious universities?

    6. CommentedAvraam Dectis

      Quantitative Easing is not printing money.

      Quantitative Easing is buying debt with printed money - effectively monetization. This is highly useful in economies characterized by low inflation, high unemployment, high debt and overcapacity - it lowers the effective debt service of the government.

      Perhaps another key mistake is failing to recognize that printed money, used in other ways during similar economic duress, could be highly stimualtive.

      For example, many industrialized countries are suffering from various degrees of inadequate birth rates. That could be addresses with a Socially Positive Stimulus using printed money.

      An SPS, ( which is a stimulus aimed at a social ill), would be implemented as follows. It would be declared that a small amount of money would be given to whomever gave birth, with a bonus if they were married. That number would automatically rise until an acceptable birth rate presented itself.

      There is no way to know what sum would generate an acceptable result. Some people might be motivated at 10,000 pounds, others at 30,000. Eventually the automatic adjustment would ensure the desired effect.

      A stimulus that increases the birth rate is both a very strong short and long term stimulus. Short term because the stimulus money is spent raising the child. Long term because the child will become a taxpaying adult. Another benefit is that additional taxpayers reduce pressure on social programs.

      The automatic nature of the adjustment is an example of an Automatic Economic Stabilizer, and the same mechanism can be applied to taxing and trade.

      Perhaps the biggest key mistake is the lack of a search for new solutions.

        CommentedAvraam Dectis

        Hello Odysseas Argyriadis,

        The concerns about overpopulation are valid.

        I am only suggesting birth rate enhancement in rapidly aging societies.

        Also, it may well be that this planet could support many more people with the right technological advances.

        If a form of energy was developed that was plentiful and virtually free, many problems would go away. Better ecological management could help as well. Ultimately the answer may come down to our technological progress and world stewardship


        CommentedOdysseas Argyriadis

        Even though the argument presented is a sound one, it does not take into account the macro viability of such a move (encouraging birth rates).
        If we accept, for the sake of this argument, that the planet can support an unlimited number of humans then all is well since the global economy is stimulated artificially through birth rates.
        If, on the other hand, we accept that the planet cannot support an unlimited number of humans, then we are creating an even bigger problem through such an incentive, overpopulation, which will create even more serious problems due to diminishing resources.
        This boils down to the same essential problem that many consider a sound argument: even though we want to keep an economic system that promotes infinite growth, this is impossible due to limited resources. Unless there are plans to colonize planets in our solar system in the next century, we are just transferring the problem to the future generations.

    7. CommentedAlexandros Liakopoulos

      Excellent analysis, Professor Skidelsky! However, I cannot miss to pinpoint one question: what is out of the scope of economic forecasters due to their neoliberal training, which subjects micro to macro economics and forgets the fact the very nature of the science is all about POLITICAL ECONOMY and not about "economics" in the first place, is it also outside of the scope of political (strategic) forecasters?

      In other words, economic forecasters cannot predict and cannot provide any vital solutions in the medium and macro terms; they can only prescribe "bandages" in the "bullet trauma" the world economy is suffering, due to their training, which excludes them from being able to "see the big picture" by design. Their training as "economists" and not as "political economy scientists" forces them to deal with thousands of small burning trees while the whole forest is on fire.

      However, this situation is predictable for a strategic analyst and forecaster. Therefor, within the Cold Economic War Era we are currently undergoing at the world level, economists' assumptions, analysis and ongoing work can very well be predicted, "channeled" and "handled" according to the specific interests of the Very Serious Power(s), which cover the medium and long terms while dealing with the short term subjects.

      What economists' cannot do, the Very Serious Power(s) can do. After all, the US government acts on a "50 years forward orientation" since 1997 and the speech of the CIA director at the time. Therefor, economists' incapacity to deal with the crisis, which you highlight so excellent, could also serve as a policy instrument, "designed" to provide an excellent excuse to the Very Serious House for boosting its own agenda and change the World Order in a US Pantocracy Model.

      People behave in a certain matter when "crisis" changes everything; economists are also expected to act and think in the way they are trained; whoever knows which way they are trained, also knows what to expect; if expectations can be incorporated in a wider (and deeper in time) plan. Being so, should we blame the economists for behaving according to their (bad - one sided - anti-scientific) training, or should we blame those people abusing this very well known characteristic of world economists to cover their own involvement for "shooting world economy with a bullet" in the first place?

      Year 2012: Six years after the collapse of Mac n Mae and 4 years after Lehman - which "triggered" every single "focal financial crisis" on the planet - we all talk about "Greece's financial problems", "South's financial problems", "the Divided States of Europe's financial problems", "China's slowdown", "Japan's recession", etc. We also talk about "US financial cliff" and not "crisis", as also other "little agendas". However, what we miss to talk about is Capitalism as we knew it exists no more, it collapsed and ended!

      We live after the end of the previous world order, in between with the next one. Global Regulations from US government cover the vacuum of power after every "too-big-to-fail" (state or company) does fail and looks for "protection"; Old World Order gives its space to a New One, step-by-step and "tower-after-tower" (of Power). "Game theories" apply better in such Eras, in such Historic Times as the current years. However, "game theories" are hardly in the courses of the economists'. On the other hand, they serve as an everyday tool for political analysts.

      Today, Geography and Politics revenge Economy and prove that Power and money are two distinctive things: while money do serve as a means of Power, real Power can destroy money all together and - by doing so - it can absorb the territory, the legal frameworks, even the political systems of every "falling" entity. This is both foreseeable and very well known, especially among Very Serious Power(s), House(s), People and Institution(s). Even more, this is also "treat-able", which is a much more SERIOUS (seriousness and "dangerous-ness on world politics are mostly the same thing) "aspect of the world politics of today.

      Economists' Inability to forecast is very well-forecasted, nowadays! Actually, "everyone's else" inability to forecast is expected, as people tend to make assumptions and form their expectations according to information provided. Whoever controls the information flows, also control "others' ability to foresee independently". And this is the exact problem of our world, while it collapses step-by-step, "according to plan", exactly as predicted!

    8. CommentedFrank O'Callaghan

      Skidelsky is right about many things but he leaves out some important issues.

      Quantitative easing and fiscal austerity act on different people. Higher taxes on the low paid, cutting services and increasing their point-of-issue cost hits most people but 'printing money' helps only those to whom that money is given.

      How did we get into this mess? Fraudulent accounting by and large. Those who directed and benefited from thee creation of the crisis are insulated from the damage.

      For quantitative easing to work, equal checks should be given to everyone to spend as they wish (spaced out over time). It would be interesting to see what the multiplier and the tax yield would be.

    9. CommentedProcyon Mukherjee

      Robert Skidelsky offers many insights that raise further questions. First of all is it at all about forecasting errors, or is it that results have followed the intent where forecasting is just a means?

      Let me take only two examples, first that monetary transmission while continuing at the zero lower bound have sometimes forced investments back into those assets that may not generate employment, very typical is buy-back of stocks, or merger and acquisition, or simply buying less risky assets from buyers while selling even more risky assets (sometimes leveraging on one side has negated the deleveraging by the other); any amount of monetary release if it does these would have little impact on the employment position or GDP.

      The second is on fiscal stimulus, where we have reasons to believe that federal fund flow in a range of sectors that were intended to generate employment in the public sector, actually showed a trend downhill in the period 2009 to late 2011, which begs the question that stimulus does not simply turn into jobs. All of the above have less to do with forecasting, it is a mix of what the monetary transmission is actually being used for and the lack of will to execute actions for specific delivery of targets like employment.

      Procyon Mukherjee