Predicting Recessions and Slowdowns: A Robust Approach Pami Dua Delhi School of Economics, University of Delhi and Economic Cycle Research Institute, New.

Predicting Recessions and Slowdowns: Predicting Recessions and Slowdowns: A Robust ApproachA Robust Approach

Pami Dua

Delhi School of Economics, University of Delhi and

Economic Cycle Research Institute, New York

Presentation to 15th Conference of

Commonwealth Statisticians

February 7, 2011

ObjectivesObjectives

Can we Predict Recessions?Can we Predict Recessions? Business Cycle Analysis: BackgroundBusiness Cycle Analysis: Background Indicator Analysis: A Robust Approach Indicator Analysis: A Robust Approach

Can we Predict Recessions?Can we Predict Recessions?

The sheer severity of the Great Recession for many The sheer severity of the Great Recession for many developed economies – most importantly the United developed economies – most importantly the United States – motivates the vital question of whether the States – motivates the vital question of whether the recession, or the crisis that triggered it, could have been recession, or the crisis that triggered it, could have been foreseen. foreseen.

Can we Predict Recessions?Can we Predict Recessions?

A few years ago, the International Monetary Fund completed a 63-A few years ago, the International Monetary Fund completed a 63-country study of the ability of economists’ consensus forecasts to country study of the ability of economists’ consensus forecasts to predict recessions. The IMF concluded:predict recessions. The IMF concluded:

““The record of failure to predict recessions is virtually The record of failure to predict recessions is virtually unblemished.”unblemished.”

The real challenge is not to identify the best model to predict The real challenge is not to identify the best model to predict recessions ex post in a specific economy over a given time frame, recessions ex post in a specific economy over a given time frame, but to identify approaches that are robust enough to perform well but to identify approaches that are robust enough to perform well in real time under diverse structural conditions.in real time under diverse structural conditions.

Can we Predict Recessions?Can we Predict Recessions?

Key is to make timely recession calls in fast-changing emerging Key is to make timely recession calls in fast-changing emerging markets as well as in mature economies undergoing material markets as well as in mature economies undergoing material structural changes.structural changes.

Thus it would do little good to develop models optimized on the Thus it would do little good to develop models optimized on the basis of past performance if the future is likely to be quite different. basis of past performance if the future is likely to be quite different.

Real time forecasting failure cannot therefore always be blamed on Real time forecasting failure cannot therefore always be blamed on “parameter drift” or “this time it’s different” as an excuse for “parameter drift” or “this time it’s different” as an excuse for forecast error. forecast error.

Can we Predict Recessions?Can we Predict Recessions?

Econometric models are “falsifiable” since these can be tested Econometric models are “falsifiable” since these can be tested against real data. against real data.

On the other hand, the leading indicator approach relies on On the other hand, the leading indicator approach relies on descriptive observations of sequences of events in the vicinity of descriptive observations of sequences of events in the vicinity of cyclical turning points that are not ‘falsifiable’. cyclical turning points that are not ‘falsifiable’.

The pioneers of business cycle research were Wesley C. Mitchell The pioneers of business cycle research were Wesley C. Mitchell and Arthur F. Burns, who in 1938 identified the very first “leading and Arthur F. Burns, who in 1938 identified the very first “leading indicators of cyclical revival” . They were later joined by Geoffrey indicators of cyclical revival” . They were later joined by Geoffrey H. Moore who developed the first ever list of “leading indicators of H. Moore who developed the first ever list of “leading indicators of cyclical revival and recession.” cyclical revival and recession.”

Can we Predict Recessions?Can we Predict Recessions?

Steps in indicator analysis:Steps in indicator analysis:

Define a business cycleDefine a business cycle

Define a recessionDefine a recession

Determine the reference chronology – benchmark for Determine the reference chronology – benchmark for

determining recession-forecasting performancedetermining recession-forecasting performance

Identify leading indicators on the basis of the reference Identify leading indicators on the basis of the reference

chronologychronology

Business Cycle Analysis:Business Cycle Analysis:

BackgroundBackground

The Classical Business CycleThe Classical Business Cycle““Business cycles are a type of fluctuation found in the aggregate Business cycles are a type of fluctuation found in the aggregate economic activity of nations that organize their work mainly in economic activity of nations that organize their work mainly in business enterprises. business enterprises.

A cycle consists of expansions occurring at about the same time A cycle consists of expansions occurring at about the same time in many economic activities, followed by similarly general in many economic activities, followed by similarly general recessions, contractions, and revivals which merge into the recessions, contractions, and revivals which merge into the expansion phase of the next cycle. expansion phase of the next cycle.

This sequence of changes is recurrent but not periodic. This sequence of changes is recurrent but not periodic.

In duration business cycles vary from more than one year to ten or In duration business cycles vary from more than one year to ten or twelve years; they are not divisible into shorter cycles of similar twelve years; they are not divisible into shorter cycles of similar character with amplitudes approximating their own.”character with amplitudes approximating their own.”

-- Burns and Mitchell, 1946-- Burns and Mitchell, 1946

Recessions and ExpansionsRecessions and ExpansionsA A recessionrecession is the phase of the business cycle marked is the phase of the business cycle marked by pronounced, pervasive and persistent declines in the by pronounced, pervasive and persistent declines in the key measures of aggregate economic activity, i.e., key measures of aggregate economic activity, i.e., output, employment, income and sales. output, employment, income and sales.

An An expansionexpansion is the phase of the business cycle marked is the phase of the business cycle marked by pronounced, pervasive and persistent increases in by pronounced, pervasive and persistent increases in the key measures of aggregate economic activity, i.e., the key measures of aggregate economic activity, i.e., output, employment, income and sales.output, employment, income and sales.

Alternating expansions and recessions make up the Alternating expansions and recessions make up the business cyclebusiness cycle. .

Business Cycles, Growth Cycles, Business Cycles, Growth Cycles, Growth Rate CyclesGrowth Rate Cycles

The business cycle is a consensus of cycles in many activities, which have a tendency to peak and trough around the same time.

A growth cycle traces the ups and downs through deviations of the actual growth rate of the economy from its long-run trend rate of growth.

Growth rate cycles are the cyclical upswings and downswings in the growth rate of economic activity.

Business Cycles and Growth Rate CyclesBusiness Cycles and Growth Rate Cycles

A coincident indicator measures current economic activity. It turns down when the economy turns down and up when the economy turns up.

A leading indicator predicts future economic activity. It turns down before the economy enters a recession and up before the expansion begins.

Indicator Approach to Monitoring Indicator Approach to Monitoring Economic ActivityEconomic Activity

A composite index is constructed from a number of individual economic indicators. By construction, it has more information than any individual component of the index.

A composite coincident index comprises variables that collectively represent the current state of the economy. It indicates whether the economy is currently expanding or is in a recession.

A composite leading index includes variables that collectively anticipate turning points of the business cycle. This is used as a predictive tool to gauge if and approximately when a recession or an expansion might take place.

Composite IndexesComposite Indexes

Leading Indexes can Time TurnsLeading Indexes can Time TurnsLead

Target

Leading Index

The Recession as a Vicious CycleThe Recession as a Vicious CycleA recession occurs when a decline in some measure of aggregate A recession occurs when a decline in some measure of aggregate economic activity sets off cascading declines in the other economic activity sets off cascading declines in the other coincident measures of activity. coincident measures of activity.

Thus, when a dip in sales causes a drop in production, triggering Thus, when a dip in sales causes a drop in production, triggering declines in employment and income, which in turn feed back into a declines in employment and income, which in turn feed back into a further fall in sales, a vicious cycle results and a recession further fall in sales, a vicious cycle results and a recession ensues. ensues.

This domino effect of the transmission of economic weakness This domino effect of the transmission of economic weakness from sales to output to employment to income, feeding back into from sales to output to employment to income, feeding back into further weakness in all of these measures in turn, is what marks a further weakness in all of these measures in turn, is what marks a recessionary downturn. recessionary downturn. This effect spreads from industry to indutry, region to region, and This effect spreads from industry to indutry, region to region, and indicator to indicator. indicator to indicator.

The Expansion as a Virtuous CycleThe Expansion as a Virtuous Cycle

At some point, the vicious cycle is broken and an At some point, the vicious cycle is broken and an analogous self-reinforcing virtuous cycle begins, with analogous self-reinforcing virtuous cycle begins, with increases in output, employment, income and sales increases in output, employment, income and sales feeding into each other – the hallmark of a business feeding into each other – the hallmark of a business cycle expansion. cycle expansion.

How the Virtuous Cycle WorksHow the Virtuous Cycle Works

Turning Points:Turning Points:Business Cycle Peaks and TroughsBusiness Cycle Peaks and TroughsBecause recessions can be characterized as vicious Because recessions can be characterized as vicious cycles and expansions as virtuous cycles, the transition cycles and expansions as virtuous cycles, the transition points between the vicious and virtuous cycles, based points between the vicious and virtuous cycles, based on the consensus of the coincident indicators (output, on the consensus of the coincident indicators (output, employment, income and sales), properly mark the start employment, income and sales), properly mark the start and end dates of recessions (peaks and troughs). and end dates of recessions (peaks and troughs).

That is also why “two down quarters of GDP” is not an That is also why “two down quarters of GDP” is not an adequate definition, nor a proper criterion, for a adequate definition, nor a proper criterion, for a recession. recession.

Business Cycle ChronologiesBusiness Cycle ChronologiesThe historical dates of business cycle peaks and troughs The historical dates of business cycle peaks and troughs are thus based on the consensus of the dates of the are thus based on the consensus of the dates of the peaks and troughs in the broad measures of output, peaks and troughs in the broad measures of output, employment, income and sales. employment, income and sales.

For 20 economies, ECRI maintains business cycle peak For 20 economies, ECRI maintains business cycle peak and trough dates based on the same approach. and trough dates based on the same approach.

The up-to-date list of business cycle dates for 20 The up-to-date list of business cycle dates for 20 countries including the U.S. is available here:countries including the U.S. is available here:

http://www.businesscycle.com

Growth Rate Cycle ChronologiesGrowth Rate Cycle ChronologiesGrowth rate cycles are made up of alternating periods of Growth rate cycles are made up of alternating periods of rising and falling economic rising and falling economic growthgrowth. .

They are based on the They are based on the growth ratesgrowth rates of the coincident of the coincident indicators whose levels relate to business cycles, andindicators whose levels relate to business cycles, anddo not rely on estimation of the current trend. Hence, do not rely on estimation of the current trend. Hence, growth rate cycles, along with business cycles, are growth rate cycles, along with business cycles, are useful for real-time monitoring of the economy. useful for real-time monitoring of the economy.

For this reason, ECRI maintains growth rate cycle For this reason, ECRI maintains growth rate cycle chronologies for 20 countries:chronologies for 20 countries:

http://www.businesscycle.com

Turning Points are Hard to PredictTurning Points are Hard to Predict

Forecast Error

Target

Consensus Forecast

Taming the CycleTaming the Cycle

Reducing Recessions:Reducing Recessions:

Raise trend rate of Raise trend rate of growthgrowth

Lower amplitude of Lower amplitude of cyclescycles

-4

0

4

8

-4

0

4

8

-4

0

4

8

-4

0

4

8

Effects of Higher Trend& Lower Volatility on Business Cycles

Recessions

Higher Trend: No Recessions

Lower Volatility: No Recessions

Hence the need to anticipate

After Chrysler’s near-death experience in early

1980’s the then Chairman, Lee Iacocca told his

Chief Economist:

"All I want from you is that you let me know six

months before the next downturn."

Robustness of Indicator Analysis Robustness of Indicator Analysis Based on an understanding of business cycle theory, Based on an understanding of business cycle theory,

Moore examined the empirical record of the behaviour Moore examined the empirical record of the behaviour of a list of indicators at business cycle peaks and of a list of indicators at business cycle peaks and troughs. This empirical testing – based on U.S. data troughs. This empirical testing – based on U.S. data from 1870 to 1938 – determined the final selection of from 1870 to 1938 – determined the final selection of Moore’s 1950 list of eight leading indicators of Moore’s 1950 list of eight leading indicators of recession and recovery. recession and recovery.

This entire process was rooted in business cycle theory: This entire process was rooted in business cycle theory: not in falsifiable statistical models, to be sure, but in a not in falsifiable statistical models, to be sure, but in a theoretical, conceptual understanding of the drivers of theoretical, conceptual understanding of the drivers of the business cycle, nevertheless. Empirical testing the business cycle, nevertheless. Empirical testing played only a secondary role in this process. played only a secondary role in this process.

Robustness of Indicator Analysis Robustness of Indicator Analysis

Nearly half a century later, Moore asked the question: Nearly half a century later, Moore asked the question: we know that the original leading indicators anticipated we know that the original leading indicators anticipated both recessions and recoveries in the late 19th and both recessions and recoveries in the late 19th and early 20th centuries, but what have they done for us early 20th centuries, but what have they done for us lately? lately?

He tested their “out-of-sample” performance, so to say, He tested their “out-of-sample” performance, so to say, in the second half of the 20th century (Moore and in the second half of the 20th century (Moore and Cullity, 1994). All the leading indicators continued to Cullity, 1994). All the leading indicators continued to lead at US business cycle turning points.lead at US business cycle turning points.

Robustness of Indicator Analysis Robustness of Indicator Analysis We recently completed a similar analysis for the US We recently completed a similar analysis for the US

from the mid 20from the mid 20thth century through the early 21 century through the early 21stst century, century, including the Great Recession. Results are similar.including the Great Recession. Results are similar.

We also gathered data on the same indicators, or their We also gathered data on the same indicators, or their closest equivalents, in all the Group of Seven (G7) closest equivalents, in all the Group of Seven (G7) economies other than the U.S. in the postwar period. economies other than the U.S. in the postwar period. Remarkably, when we compared their turning points Remarkably, when we compared their turning points with the respective business cycle chronologies with the respective business cycle chronologies established independently by ECRI on the same basis established independently by ECRI on the same basis as in the U.S., their performance held up.as in the U.S., their performance held up.

Robustness of Indicator Analysis Robustness of Indicator Analysis Next, we conducted a similar analysis, but on the basis Next, we conducted a similar analysis, but on the basis

of growth rate cycles of growth rate cycles (acceleration-deceleration cycles, (acceleration-deceleration cycles, consisting of alternating cyclical upswings and consisting of alternating cyclical upswings and downswings in economic growth)downswings in economic growth) rather than classical rather than classical business cycles.business cycles.

We found that the growth rates of the same leading We found that the growth rates of the same leading indicators continued to lead the respective growth rate indicators continued to lead the respective growth rate cycle turning points, which had been determined cycle turning points, which had been determined independently by ECRI for all the G7 economiesindependently by ECRI for all the G7 economies

Average Leads, in Months, ofAverage Leads, in Months, ofEight Leading Indicators Selected in 1950Eight Leading Indicators Selected in 1950

Before 1938:Business cycles 1948-2008:

Business cycles 1948-2008: Growth rate cycles

U.S.

G7 excl. U.S.0

1

2

3

4

5

6

7

8

9

Robustness of Indicator Analysis Robustness of Indicator Analysis We also constructed a composite leading index out of We also constructed a composite leading index out of

Moore’s original list of U.S. leading indicators. That Moore’s original list of U.S. leading indicators. That leading index covers 107 years and 21 recessions, leading index covers 107 years and 21 recessions, including the 1907-08 and 1920-21 depressions, the including the 1907-08 and 1920-21 depressions, the entire period of the Great Depression, and the more entire period of the Great Depression, and the more recent Great Moderation. Again, no data fitting was recent Great Moderation. Again, no data fitting was involved in creating the index. involved in creating the index.

So how did the Index of Original Leading Indicators So how did the Index of Original Leading Indicators (IOLI) perform during the Great Recession, which (IOLI) perform during the Great Recession, which caught so many by surprise? As a matter of record, it caught so many by surprise? As a matter of record, it peaked in July 2007, five months before the official peaked in July 2007, five months before the official December 2007 U.S. business cycle peak. It December 2007 U.S. business cycle peak. It subsequently troughed in March 2009, three months subsequently troughed in March 2009, three months before the June 2009 U.S. business cycle trough. before the June 2009 U.S. business cycle trough.

Robustness of Indicator Analysis Robustness of Indicator Analysis

Over its 107-year span, the IOLI exhibits a median Over its 107-year span, the IOLI exhibits a median lead of 4.5 months at US business cycle peaks and lead of 4.5 months at US business cycle peaks and three months at business cycle troughs, leading at three months at business cycle troughs, leading at 93% of business cycle turning points.93% of business cycle turning points.

In the out-of-sample post-war period, the statistics are In the out-of-sample post-war period, the statistics are similar: the IOLI has a median lead of six months at similar: the IOLI has a median lead of six months at business cycle peaks and three months at business business cycle peaks and three months at business cycle troughs, leading at 91% of business cycle cycle troughs, leading at 91% of business cycle turning points. turning points.

Robustness of Indicator AnalysisRobustness of Indicator Analysis What we have shown is that, when evaluated against What we have shown is that, when evaluated against

those objective cyclical benchmarks, the original those objective cyclical benchmarks, the original leading indicators of recession and recovery – leading indicators of recession and recovery – selected primarily on a conceptual basis – continue to selected primarily on a conceptual basis – continue to exhibit remarkably robust performance under a wide exhibit remarkably robust performance under a wide range of conditions. range of conditions.

Whether we are faced with the specific likelihood of Whether we are faced with the specific likelihood of more frequent U.S. recessions, as ECRI’s research more frequent U.S. recessions, as ECRI’s research suggests, or with unforeseeable changes in structural suggests, or with unforeseeable changes in structural conditions that may shape the next cycle, it is our conditions that may shape the next cycle, it is our belief that it would surely be prudent to rely on a time-belief that it would surely be prudent to rely on a time-tested, robust approach to business cycle forecasting. tested, robust approach to business cycle forecasting.

Way ForwardWay Forward The field of economic forecasting faces formidable The field of economic forecasting faces formidable

challenges in the years ahead. Robust leading indicators can challenges in the years ahead. Robust leading indicators can provide only a partial answer.provide only a partial answer.

Unlike econometric models, leading indicators are designed Unlike econometric models, leading indicators are designed to predict only the timing of cyclical turning points, not to predict only the timing of cyclical turning points, not forecasts of the magnitude of economic variables. Nor can forecasts of the magnitude of economic variables. Nor can they answer “what if” questions which are central to policy they answer “what if” questions which are central to policy decisions. decisions.

This approach should therefore be seen as a complementary This approach should therefore be seen as a complementary tool but one that is capable of providing invaluable guidance tool but one that is capable of providing invaluable guidance in the lead-up to recessions and recoveries, preventing in the lead-up to recessions and recoveries, preventing decision makers from being blindsided by the inevitable decision makers from being blindsided by the inevitable turning points in years to come. turning points in years to come.

Thank You!Thank You!

Annualized Growth in U.S. Coincident IndicatorsAnnualized Growth in U.S. Coincident Indicatorsin Postwar U.S. Expansions (%)in Postwar U.S. Expansions (%)

49-5354-57

58-6061-69

70-7375-80

80-8182-90

91-0101-07

Employment

GDP

Personal Income

Mfg &Trade Sales

Industrial Production0

2

4

6

8

10

12

Three Key Aspects of the EconomyThree Key Aspects of the Economy

Economic Growth

Inflation

Em

ploy

men

t

The State of the ArtThe State of the Art

Economic GrowthEconomic Growth

InflationInflation

Em

ploy

men

tE

mpl

oym

ent

Long

Lea

ding

, W

eekl

y Le

adin

g,

Sho

rt L

eadi

ng &

Coi

ncid

ent

Ser

vice

s

Manufacturing

Construction

DomesticDomesticForeignForeignTradeTrade

Exports

Imports

Trade Balance

Future Inflation Gauge

Home Prices

Em

ploy

men

t

Non-Financial

Financial

Non

-Mfg

Mfg

Crack of the BullwhipCrack of the Bullwhip Around the globe, recession is being Around the globe, recession is being

transmitted in amplified form to the more transmitted in amplified form to the more export-oriented economiesexport-oriented economies

Ruth Mack, a colleague of ECRI founder Dr. Ruth Mack, a colleague of ECRI founder Dr. Geoffrey H. Moore, uncovered this link in a Geoffrey H. Moore, uncovered this link in a study of shoe, leather and hidesstudy of shoe, leather and hides

When Mack did her study, shoes were not When Mack did her study, shoes were not impulse buys but expensive products that impulse buys but expensive products that consumers would buy in good timesconsumers would buy in good times

In not so good times, consumers would get In not so good times, consumers would get their shoes repaired and postpone the their shoes repaired and postpone the purchase, implying that shoe demand was purchase, implying that shoe demand was moderately cyclical moderately cyclical

Crack of the BullwhipCrack of the Bullwhip An increase in inventories of shoes and shoe leather An increase in inventories of shoes and shoe leather

due to a drop in demand resulted in shoemakers due to a drop in demand resulted in shoemakers reducing production and orders for leatherreducing production and orders for leather

Thus slowdown in shoe demand would result in an Thus slowdown in shoe demand would result in an actual decline in the demand for leather, which is actual decline in the demand for leather, which is made from cattle hidesmade from cattle hides

This would trigger a sharp plunge in the demand for This would trigger a sharp plunge in the demand for hides hides

Thus small shifts in demand growth at the consumer Thus small shifts in demand growth at the consumer level are amplified through the supply chain into big level are amplified through the supply chain into big swings in demand as we move up the supply chain swings in demand as we move up the supply chain away from the consumeraway from the consumer

This is the BULLWHIP EFFECT because a little flick of This is the BULLWHIP EFFECT because a little flick of the wrist produces a big arc at the end of the whipthe wrist produces a big arc at the end of the whip

Shoe-Leather-Hide Sequence:Shoe-Leather-Hide Sequence:The Bullwhip EffectThe Bullwhip Effect

Leather Demand:Level Falls

Shoe Demand:Growth Slows

Hides Demand:Level Plunges

Frontline demand is Frontline demand is cyclicalcyclical

Midline demand is Midline demand is more cyclicalmore cyclical

Demand early in the Demand early in the supply-chain is most supply-chain is most cyclical but supply is cyclical but supply is relatively insensitive, relatively insensitive, so so pricesprices are highly are highly cyclicalcyclical

Bullwhip Effect Bottom LineBullwhip Effect Bottom Line

Contractions in the global economy are Contractions in the global economy are concentrated in the developed concentrated in the developed countriescountries

Economies that are heavily involved in Economies that are heavily involved in the exports of manufactured goods will the exports of manufactured goods will be lashed by the Bullwhip Effect and be lashed by the Bullwhip Effect and their suppliers – especially the their suppliers – especially the producers of industrial commodities, producers of industrial commodities, including oil – will be in even worse including oil – will be in even worse predicamentspredicaments