el desafio de la medida en economia dinamica

8/14/2019 El Desafio de La Medida en Economia Dinamica

1/5

The Challenge ofMeasuring andModeling a DynamicEconomyBy Alan Greenspan

Alan Greenspan is Chairman,Board of Governors, FederalReserve System and is a past presi-dent of NABE.

There is a trade-off between better data and betteranalysis in the allocationof scarce economic researchresources. Although great strides have been made ineconometric models, their use as p redictive devices doesnot seem to be superior to simple, reduced-form models,even though they have provided valuable insights intothe functioningof the economy. The financial sector, inparticular,presents unresolved problems in theory andmodeling. However, given the rapidchange in the struc-ture of the U.S. economy, it appears thatgetting betterdata may provide a better return than more sophisticatedanalysis. This is particularly ruefor prices and outputin service industries. The FederalReserve System, aswell as the Bureauof Labor Statistics, the Bureau ofEconomic Analysis, and the Census Bureau, have madegreat strides in the conceptual improvement of economicmeasurement, but much remains to be done.Presented at the Washington Economic Policy Conference of theNationalAssociationforBusiness Economics, Washington, DC ,March 27, 2001.

I am pleased to have this opportunity to address anissue of considerable importance to both businesseconomists and policymakers-that is, the chal-lenge of measuring and modeling our dynamic econ-omy. Moreover, I would like to raise the issue of howmuch of our finite resources should be directed to measur-

ing and how much to modeling.Business economics endeavors to understand thestructure of an economy-how it works and, above all, howto forecast it. It is a bedeviling job because the future, atroot, cannot be foretold. The best we can do is to constructprobabilistic models that can inform the decisions of busi-ness executives and, of course, economic policymakerswho-of necessity-will be making their decisions armedwith incomplete information.

For a while in the 1960s, we were increasingly mes-merized by the possibilities of econometric models as acrystal ball. However, history was not entirely kind to thisendeavor. For one thing, especially against the backdrop ofthe inflation of the following decade, it soon became appar-ent that our theories of the macroeconomy were woefullyinadequate. For another, even leaving aside the shortcom-ings of our theory, we soon learned that the economic struc-ture did not hold still long enough to capture its key rela-tionships. Its changing structure frustrated efforts to isolatea reasonably fixed set of coefficients. In tum, the absenceof fixed coefficients undermined the usefulness of themodel as a basis for projecting the future.Econometricians recognized many of these difficulties

The Challenge of Measuringand Modeling a Dynamic Economy Bu-siness Economics oApril 2001 5


2/5

and so developed avast and elegant liter- Should we endeavor Iature in support of thisresearch program, Or should we ... aug,covering a spectrum greaterpayoffs will cof topics ranging frommaximum-likelihood- technique.estimation techniquesto tests for coefficientstability to diagnostics for detecting undesirable propertiesof the errors of these equations. The creativity that wasapplied to this effort is all the more impressive because ittook place in the context of a computing environment thatwas, by modem standards, truly primitive.

But in time it became increasingly clear that, for alltheir theoretical advantages, these sophisticated modelsdid not reliably outperform a number of simple and far lesscostly reduced-form models, from the money supply mod-els that appeared to work well for a while during the 1970sto a structural vector autoregression models based on ahandful of lagged variables that are still employed today.

To be sure, the large econometric models have beenrefined, incorporating the fruits of later theoretical devel-opments, including perhaps most importantly the insightthat monetary policy could not permanently influence thelevel of the unemployment rate. Moreover, a larger role wasgiven to a range of financial and expectational variablesthat earlier practitioners steeped in orthodox Keynesianincome determination tended to downgrade. Liquiditypreference functions, to be sure, were included even in theearly versions of these large-scale models as necessarybuilding blocks for determining the equilibrium level ofinterest rates and income. But, overall, financial sectormodeling was primitive. Indeed, only modest progress hasbeen made in this area since the Federal Reserve began toproduce our own flow of funds accounts in 1955.

A further, and perhaps more profound, challenge to theunderlying validity of this style of modeling is the possi-bility that, whereas standard models of the real economydeternine a unique level of income, the financial systemappears to be capable of reaching myriad equilibria. Inaddition, the fundamental forces that determine which ofthese equilibria will be selected may themselves be inher-ently unpredictable.

We have built large-scale models of the United Statesand global economies at the Federal Reserve. While rec-ognizing their limitations, we do find them useful inresearch and analysis. But the experience of the last fortyyears underscores a fundamental dilemma of business eco-nomics. Should we endeavor to continue to refine our tech-niques of deriving maximum information from an existing

o continue to refine our techniques .. ?ment ourdata library .. ? I suspectomefrom more data thanfrom more

body of data? Or should we find ways to augment our datalibrary to gain better insight into how our economy is func-tioning? Obviously, we should do both, but I suspectgreater payoffs will come from more data than from moretechnique.

Certainly, statistical systems in the United States, bothpublic and private, are world class and, indeed, in manyrespects set the world standard. But given the rapidlychanging economic structure, one could readily argue thatmore statistical resources need to be applied to under-standing the complexities of the newer technologies thatconfront analysts.The Challenge of Measuring Price and Output

These newer technologies and the structure of outputthey have created have surfaced a set of definitional prob-lems that-although evident in a world of steel, fabrics,and grains-were never on the cutting edge of analysis. Irefer, of course, to the age-old problem of defining what wemean by a unit of output and, by extension, what we meanby price. The dollar value of sales or GD P depends, ofcourse, on the specific accounting rules chosen. And whilevalue in that context is uniquely defined, the split betweenvolume change and price change is always approximate.

In decades past, we struggled about what we meant byoutput-and hence price nonetheless an average priceof hot rolled steel sheet and a corresponding total tonnagewas precise enough for most analytic needs. By the sametoken, tons of steel per work hour in a rolling mill yieldedrough approximations of underlying productivity for mostpurposes.Output per hour in an economy dominated by suchgoods, or even services, for which the definition of a typi-cal unit of output was reasonably unambiguous, was ameaningful and relatively robust statistic. Our data sys-tems in the early post-World War II years were by and largeadequate to the task.

But over time, and particularly during the last decadeor two, an ever-increasing share of GD P has reflected thevalue of ideas more than material substance or manuallabor input. This ongoing development is imposing signif-icant stress on our statistical systems. We know, presum-

The Challenge of Measuringand Modeling a Dynamic Economy Business Economics - April 2001


3/5

ably uniquely, the dollar value of a software application.But when comparing software-application values overtime, how much of the change is volume and how much isprice? The answer, in principle, requires judgments aboutvery fundamental issues in measurement: What are theunderlying determinants of consumer value preference,and how does this good or service contribute to that pref-erence, taking account of all the other goods and servicesbeing consumed? Problems that were always latent indefining steel prices and quantities but rarely rose to thislevel of significance are threatening to seriously challengeour measurement systems in an age of the microprocessor,fiber optics, and the laser.These latent problems have emerged in full view in thepricing of medical services. Perhaps the inherent com-plexity of this undertaking is most clearly revealed by pos-ing the question, what do we mean by a standardized unitof medical output? Is it the procedure, the treatment, or theoutcome? What does the fee charged for the bundle ofservices associated with cataract or arthroscopic surgeryrepresent? How does one value the benefits to the patientof shorter hospital stays, more comfortable recoveries, andbetter physical outcomes? Clearly, the unadjusted fee for asingle medical procedure does not adequately represent its"price."

The price indexes for medical services used to be con-structed by pricing a variety of inputs-for example, anight in the hospital, or an hour of a physician's time. Afew years ago, the Bureau of Labor Statistics (BLS) beganmoving toward pricing the treatment paths of particulardiagnoses, the better to capture changes in the mix ofinputs used to treat a given disease. For example, manysurgical procedures that used to require an overnight stayin a hospital now can be performed on an outpatient basis,and the producer price index and consumer price indexare now better able to measure the price decline associat-ed with that change. Interestingly, when such techniquesare applied to individual medical procedures they appearalmost without exception to indicate falling prices at leastsince the mid-1980s. This has raised significant questionsas to whether our current measures of overall medical serv-ice price inflation are capturing the appropriate degree ofproductivity advance evident in medicine.Indeed, the level of real gross product per hour formedical services embodied in our overall productivitymeasures declined between 1990 and 1999 (the last yearfor which data are available). This is implausible and rais-es obvious questions of the validity of the price deflatorscurrently employed. Thus, while progress is visible, enor-mous measurement challenges remain in measuring pricesof medical services.

But there are deeper issues as well, associated with thevaluation of a consumer's time. Clearly the shorter stay ofa cataract patient is of value to that patient. In other words,today's techniques allow the surgeon to deliver more con-sumer value per hour of operating time, other things beingequal. A full measure of the output of the medical sectorwould take account of this reduction in recovery time andattribute it to the medical sector.Recent Approaches to Price and OutputMeasurement

While many issues of measurement arise in the contextof services, even the measurement of some goods pricespresents considerable challenges. High-technology goodsare a case in point. Academic research in this area datesto the mid-1960s, but its application in the measurementof real output gained prominence with the introduction ohedonic price indexes for computers and peripherals bythe Bureau of Economic Analysis (BEA) in 1985. Morerecently, the efforts undertaken by statistical agencies haveintensified, spurred by the accelerated pace of technologi-cal innovation, which has yielded an ever-expanding rangeof new products and product variants, as well as by the ris-ing share of these goods in our economic value added.Thus, much progress has been made by the BEACensus, and the BLS, with which I am sure you are famil-iar. This morning, I should like to alert you to some of theresearch in these areas coming from the Federal Reserve.Much of our work in this regard has been focused onimproving our published statistics on industrial produc-tion.Our staff's multiyear work in this area began in 1998with the development of new measures of the domestic out-put of semiconductors. Next, we revised our procedures forestimating the production of computers, and more recentlywe have introduced new series for an important componentof the output of the communications equipment industry-local area networking (LAN) equipment, which providesthe infrastructure critical to expanding the productive usesof information technology.In total, these high-tech goods-semiconductors, computers, and LAN equipment-currently represent lessthan eight percent of total manufacturing output. Howevertheir production, as we measure it, rose at an averageannual rate of around fifty percent in the second half of the1990s, and, taken together, they contributed two-thirds othe increase in manufacturing output between 1995 and2000. Indeed, U.S. production of semiconductors in 1996eclipsed motor vehicle assemblies as the largest four-digimanufacturing industry in nominal value-added terms.The characteristics of these goods present the range o

The Challenge of Measuringand MVodeling a Dynamic Economy Business Economics o April 2001


4/5

complexities that one faces in measuringquality-adjusted prices. First, many are The chiwholly new products: For example, switch-es-the largest single segment of LAN hard- the supware-did not enter the market until 1993.And even "older" high-tech products, such Commuas computers, are now bundled together in help deways that offer an enormous variety of com-binations of characteristics related to speed, agenciememory, networking capability, and graphicscapability, to mention just a few. For all ofthese goods, product cycles are truncated by rapid innova-tion. For instance, in 1995, ten megabit-per-secondEthernet switches dominated that market; last year, the twomost popular switches operated at rates of 100 and 1,000megabits per second. Product lives for semiconductors andcomputers can be even shorter; some computer modelshave remained on the market for only a couple of months.

In such an environment, the availability of detailedmicro-level data describing the attributes of these goods iscrucial. One means of defining the unit of output is tounbundle the characteristics of a high-tech product and toprice each of them separately. This so-called "hedonic"technique-now applied by the BEA to items that accountfor eighteen percent of GDP-is one approach. (Landefeldand Grimm, 2000) In our work at the Federal Reserve, wehave developed hedonic price indexes for network routersand switches using, in the case of the former, data fromproduct catalogs and, in the case of the latter, privately pro-duced reports evaluating the performance of the products.

However, hedonics are by no means a panacea. Mostimportant, the measured characteristics may be acting onlyas proxies for the qualities of the services that buyers ulti-mately value. This, again, raises the difficult issue of theappropriate scope for value measurement and poses thequestion of whether the correct approach may be to movetoward directly pricing the services we obtain from ourinformation processing systems rather than pricing sepa-rately the individual hardware components and the soft-ware.The Federal Reserve staff has found that, whendetailed data are available on prices and on quantities, wecan produce results that are comparable to those based onhedonics, using the conceptually simpler "matched model"approach. Indeed, we have taken this approach in con-structing quantity and price indexes for several high-techitems. (Aizcorbe, Corrado and Doms, 2000) In the case of

semiconductors, we relied on data from three private ven-dors for information on nearly 100 unique microprocessors,more than 200 types of memory chips, and more than 80other chips. We also acquired nominal sales and unit value

illenge that lies ahead .. will requireport of the business andacademicnities to supply the informationand tovelop the tools that our statisticales require.

data for about 1,100 distinct computer models.Neither hedonic nor matched-model techniques aresufficient to deal with the introduction of wholly new prod-ucts that differ fundamentally in their characteristics fromtheir predecessors. This will continue to be one of ourmajor ongoing challenges.I am encouraged by the progress that economists and

economic statisticians have been making to date in tack-ling the daunting task of measuring real output and pricesin a rapidly changing economy. The challenge that liesahead is, indeed, large, and to meet it will require the sup-port of the business and academic communities to supplythe informnation and to help develop the tools that our sta-tistical agencies require.The information revolution, itself, will also surely playan important role. For example, high-tech information sys-

tems might some day allow statistical agencies to tap intoa great many economic transactions on a basis close to realtime. More generally, I am certain that the possibilities forcreatively harnessing technology for the improvement ofeconomic measurement are much broader in scopealthough, as in many other areas of endeavor, the precisedirections those advances will take are difficult to predict.If we had the appropriate database, of course, who knows?REFERENCES

Aizeorbe, Ana,Carol Corrado, and Mark Doms. July 26, 2000."Constructing Price and Quantity Indexes for High TechnologyGoods." Industrial Output Section, Division of Research and Statistics,Board of Governors of the Federal Reserve System.

Landefeld, J. Stephen and Bruce T.Grimm. December 2000. "ANote on the Impact of Hedonics and Computers on Real GDP." Surveyof Current Business.

The Challengeof Measuringand Modeling a DynamicEconomy Business Economics o April 2001


5/5

COPYRIGHT INFORMATION

TITLE: The challenge of measuring and modeling a dynamic

economy

SOURCE: Business Economics 36 no2 Ap 2001

WN: 0109104079001

The magazine publisher is the copyright holder of this article and it

is reproduced with permission. Further reproduction of this article in

violation of the copyright is prohibited.

Copyright 1982-2001 The H.W. Wilson Company. All rights reserved.

el desafio de la medida en economia dinamica

Documents