In this article, we examine the predictive power of capacity

Size: px

Start display at page:

Download "In this article, we examine the predictive power of capacity"

Clementine Potter
6 years ago
Views:

1 Is There a Stable Relatioship Betwee Capacity Utilizatio ad Iflatio? Keeth M. Emery Seior Ecoomist ad Policy Advisor Federal Reserve Bak of Dallas Chih-Pig Chag Graduate Studet Souther Methodist Uiversity, Dallas I this article, we examie the predictive power of capacity utilizatio for iflatio, with a focus o whether the forecastig relatioship is stable. Durig recet years, the failure of moetary aggregates as reliable guides for future iflatio has led fiacial market participats ad Federal Reserve policymakers to moitor a broad rage of ecoomic statistics. O the real side, aalysts icreasigly rely o the Phillips curve the perceived existece of a stable short-ru trade-off betwee iflatio ad real activity. Most promietly, these aalysts focus o the gap betwee the uemploymet rate ad the so-called NAIRU, or oacceleratig iflatio rate of uemploymet, which is the uemploymet rate at which iflatio is costat. Similarly, may aalysts use the Federal Reserve s idustrial capacity utilizatio rate as a idicator of future iflatio pressures. 2 Typically, utilizatio rates above 82 percet have sigaled higher future iflatio, brought o by the oset of productio bottleecks ad supply shortages. This historical relatioship is illustrated i Figure, which shows that after capacity utilizatio rates rose above 82 percet, cosumer price iflatio accelerated i most istaces over the period. Durig the past few years, however, the usefuless of both the capacity utilizatio rate ad the uemploymet rate as iflatio idicators has come uder scrutiy. Figure shows that capacity utilizatio rose above 82 percet at the ed of 993 ad that, to date, iflatio has remaied stable. Likewise, the uemploymet rate has bee below most estimates of NAIRU for some time. The respose to these recet developmets seems to fall ito two categories. First, some aalysts argue that othig has chaged ad that i due time iflatio will begi to rise. A secod ad more varied group of aalysts poits to several possible developmets to explai why iflatio has remaied stable: demographic factors have lowered NAIRU; icreasigly glo- Figure CPI Iflatio ad Capacity Utilizatio Iflatio Utilizatio Utilizatio Iflatio

2 balized labor ad capital markets have lesseed the importace of U.S. capacity utilizatio; umeasured productivity icreases have led to a rise i the U.S. ecoomy s growth potetial (relevat for gap aalyses); ad the Federal Reserve mismeasures capacity utilizatio. Whereas may studies have examied the lik betwee uemploymet ad iflatio, comparatively fewer have explored the capacity utilizatio iflatio relatioship. I this article, we examie the predictive power of capacity utilizatio for iflatio, with a focus o whether the forecastig relatioship is stable. We fid evidece that while there was a sigificat positive relatioship betwee capacity utilizatio ad chages i iflatio before 983, this relatioship has substatially weakeed sice the ed of 982. I fact, after 982, oe ca reject the hypothesis that high capacity utilizatio rates have ay predictive power at all for cosumer price iflatio. The results are similar for chages i producer price iflatio, although the deterioratio i the relatioship is ot as severe. I fact, at quarterly ad semiaual horizos, there is evidece that capacity utilizatio after 982 still has predictive cotet for chages i producer price iflatio. I the first sectio of this article, we review recet literature cocerig the capacity utilizatio iflatio relatioship. I the followig sectio, we examie the empirical evidece as it relates to the capacity utilizatio iflatio relatioship. Literature review Although the literature examiig the capacity utilizatio iflatio relatioship is relatively sparse, there are several recet studies. 3 A promiet study is Garer (994), which suggests that the relatioship is stable ad that capacity utilizatio curretly remais a reliable idicator of future chages i iflatio. Specifically, Garer uses simple ordiary least squares regressios (OLS) to show that over differet sample periods, the oacceleratig iflatio rate of capacity utilizatio (NAICU) is roughly costat i the 82-percet rage. I his article, Garer poits out the similarity betwee aalyses usig the cocept of NAIRU ad those usig the capacity utilizatio rate. Oe way to show this similarity is by replacig the uemploymet rate with the capacity utilizatio rate i a simple expectatios-augmeted Phillips curve model. Expectatios-augmeted Phillips curves posit a egative trade-off betwee levels of iflatio ad uemploymet rates for a give level of expected iflatio. The top pael of Figure 2 shows several Phillips curves associated Figure 2 Phillips Curves ad the Capacity Utilizatio Iflatio Relatioship Iflatio (Π) (Percet) Iflatio (Π) (Percet) Natural rate Natural utilizatio rate Π e = % Π e = % Π e = % Π e = % Π e = % Π e = % Uemploymet Capacity utilizatio with differet levels of expected iflatio. The atural rate i the top pael of Figure 2 is that level of uemploymet at which iflatio equals expected iflatio. 4 The bottom pael of Figure 2 shows the expectatios-augmeted Phillips curves that result whe the uemploymet rate is replaced with the capacity utilizatio rate. Similarly, the atural utilizatio rate is that rate at which iflatio equals expected iflatio. However, most aalysts posit a relatioship betwee chages i capacity utilizatio ad iflatio. To derive such a relatioship, a extra assumptio must be made about the formatio of iflatio expectatios specifically, that the ext period s expected iflatio rate (Π e ) is equal to a weighted average of lagged iflatio rates, with the weights summig to oe. Uder this Figure 3 Critical Utilizatio Thresholds i Short-Ru Phillips Curve Models Chage i iflatio NAICU Capacity utilizatio FEDERAL RESERVE BANK OF DALLAS ECONOMIC REVIEW FIRST QUARTER 997

3 Table Regressio Results 967: 96:2 Sample R 2 Costat o CU CPI Mothly (.) 8. (.) (.) Quarterly (.) 82.6 (.) (.) Semiaual (.2) 8.7 (.) (.) PPI Mothly (.) 82.2 (.) (.) Quarterly (.) 82. (.) (.) Semiaual (.) 82. (.) (.) p values i paretheses that uses lagged iformatio to predict future chages i iflatio rates. To examie chages i iflatio at horizos loger tha oe moth, we estimate equatio usig ot oly mothly data, but also quarterly ad semiaual data that are costructed from the mothly data. The data cover the sample period Jauary 967 through February 996. For our measures of iflatio, we use the cosumer price idex (CPI) ad the producer price idex (PPI). 2 Idustrial capacity utilizatio may be more closely related Figure 4 Regressio Models: CPI (mothly), Chage i iflatio 2. assumptio, the result is a positive log-ru relatioship betwee chages i capacity utilizatio ad iflatio, depicted i Figure 3. 6 Of course, if the expectatios assumptio does ot hold, the at the least there may be istability i the relatioship. 7 Other studies of the capacity utilizatioiflatio relatioship fid mixed evidece o the issue of stability. 8 Fraz ad Gordo (993) fid that U.S. iflatio depeds more closely o the capacity utilizatio rate tha o the uemploymet rate. However, their oly stability aalysis is a compariso of the period with the period, which cocludes stability caot be rejected. Cecchetti (99), i a paper that examies a umber of iflatio idicators, fids evidece that capacity utilizatio adds sigificat iformatio to out-of-sample forecasts of iflatio before 982, but this iformatio disappears after 982. The ext sectio explores the stability of the capacity utilizatio iflatio relatioship. Empirical results A stadard OLS model. To examie more precisely the relatioship betwee capacity utilizatio rates ad chages i iflatio, we ru a series of regressios of the form () Π = A + B CU + C Π + u, t t i t i where Π t is the chage i iflatio from period t to period t ; CU is the idustrial capacity utilizatio rate; u is a error term; ad A i, B i, ad C i are parameters to be estimated. 9 Essetially, equatio is a i-sample forecastig equatio t Capacity utilizatio

4 Table 2 Sample Istability: Regressios CPI 967: 82:2 Sample R 2 Costat o CU Mothly (.) 82.7 (.) (.) Quarterly (.) 83.2 (.) (.) Semiaual (.2) 82.8 (.) (.) 983: 96:2 Sample R 2 Costat o CU Mothly (.) 82. (.27) (.26) Quarterly (.) 83. (.64) (.6) Semiaual (.) 8.4 (.4) (.44) p values i paretheses Table 3 Sample Istability: Regressios PPI 967: 82:2 Sample R 2 Costat o CU Mothly (.) 82. (.) (.) Quarterly (.) 83. (.) (.) Semiaual (.4) 82.9 (.) (.) 983: 96:2 Sample R 2 Costat o CU Mothly (.) 8.6 (.) (.9) Quarterly (.) 8.2 (.3) (.3) Semiaual (.) 8.4 (.) (.) p values i paretheses for PPI iflatio because the PPI icludes oly goods prices, whereas the CPI also icludes the prices of services. Table shows both the CPI ad PPI results. For the CPI results, the model explais roughly oe-third of the overall variatio of chages i iflatio, ad there is o evidece of serial correlatio i the error terms. The lag of capacity utilizatio is very sigificat with a positive sig, which idicates that a high capacity utilizatio rate leads to risig iflatio. The magitude of the coefficiet o capacity utilizatio varies with the data used. For example, with the mothly data, a oe-percetage poit icrease i the utilizatio rate leads to a.28- percetage poit icrease i iflatio at a aualized rate. The NAICU is ear 82 percet for all three models. I other words, above 82 percet, iflatio is risig, ad below 82 percet, iflatio is fallig. These results are cosistet with Garer (994), who fids NAICUs i the 82-percet rage. For the PPI results, the explaatory power of the models is higher tha for the CPI models. Depedig o the data, the adjusted R 2 s rage from.36 to.. The coefficiets o capacity utilizatio are positive ad sigificat, ad larger tha those i the CPI models. For the mothly data, a oe-percetage poit icrease i the utilizatio rate leads to a.46-percetage poit icrease i iflatio at a aualized rate. Similar to the CPI results, the NAICUs are i the 82-percet rage for all three models. I geeral, these results are cosistet with previous work that fids a positive ad sigificat relatioship betwee capacity utilizatio ad future chages i iflatio. Stability of the utilizatio iflatio relatioship. O the basis of Figure ad because of recet assertios that the relatioship betwee the chage i capacity utilizatio ad iflatio has chaged, we test for a break i the relatioship by usig Jauary 983 as the potetial breakpoit. 3 The data idicate sigificat parameter istability. 4 Therefore, we reestimate equatio usig the two separate subsamples. Table 2 shows the CPI results, ad Table 3 shows the PPI results. For the CPI results, the NAICUs remai i the 82-percet rage for both sample periods, which is cosistet with Garer s (994) results. However, there are substatial differeces across the two samples i other aspects of the results. First, the explaatory power of the model is reduced for the post-982 period, as evideced by the much smaller adjusted R 2 s. Moreover, the margial sigificace levels (p values) for capacity utilizatio idicate that utilizatio is o loger sigificat for ay of the models i the post-982 period. Also, the poit estimates of the coefficiets o capacity utilizatio are much smaller compared with the pre-983 estimates. Figure 4 shows the trade-offs betwee chages i CPI iflatio ad capacity utilizatio for the mothly FEDERAL RESERVE BANK OF DALLAS 7 ECONOMIC REVIEW FIRST QUARTER 997

5 Figure Regressio Models: PPI (mothly), Chage i iflatio Capacity utilizatio model estimated over the etire sample, as well as the two sub-samples. 6 Not oly does the trade-off flatte durig the later sample period, but our 9-percet cofidece bads for the NAICU have wideed to the poit that ay capacity utilizatio rate is cosistet with o chage i iflatio. 7 I other words, the cofidece bads ecompass the zero chage i iflatio lie across the utilizatio rage that the U.S. ecoomy has experieced. 8 I Table 3, the differeces i PPI results across the two samples are ot as great as those for CPI. For the quarterly ad semiaual data, the adjusted R 2 s are actually higher i the later sample period. Also, the poit estimates of the coefficiet o capacity utilizatio are i the same rage for both sample periods. Moreover, the NAICUs fall i the same rage across the two subsamples ad, if aythig, are lower i the later period. However, Figure shows that the lower cofidece levels about the poit estimates of the coefficiet o utilizatio ad the costat term imply reduced cofidece about our estimate of the NAICU i the later sample period. Similar to the CPI results i Figure 4, for the mothly ad quarterly models, the cofidece bads for the NAICU i the period idicate reduced cofidece about the utilizatio rate at which PPI iflatio begis to accelerate. However, for the semiaual data (Figure 6 ), the cofidece bads remai relatively tight i the period, idicatig that PPI iflatio begis to accelerate six moths after the utilizatio rate rises above the 8- to 82-percet rage. Overall, the coclusios from the subsample results are quite strog. For chages i CPI iflatio durig the period, there is o evidece that capacity utilizatio provides ay useful iformatio about future chages i iflatio. I fact, at the 9-percet cofidece level, ay capacity utilizatio rate is cosistet with o chage i iflatio. For mothly chages i PPI iflatio, the results are similar to the CPI results. However, for the semiaual data, capacity utilizatio does have sigificat iformatio for future chages i PPI iflatio, ad the 9-percet cofidece rage for the NAICU Figure 6 Regressio Models: PPI (semi), Chage i iflatio Capacity utilizatio 8

6 is estimated to be betwee 8 ad 82 percet. Results for quarterly PPI data are itermediate. Coclusios ad Discussio Durig recet years, the reliability of the uemploymet ad capacity utilizatio rates as future iflatio idicators has come uder questio. Because the usefuless of these idicators is predicated o a stable short-ru trade-off betwee real activity ad iflatio, this scrutiy has etailed a reexamiatio of the Phillips curve. Although much of the literature has focused o the uemploymet iflatio relatioship, i this article we have examied the relatioship betwee capacity utilizatio ad iflatio. We fid evidece that although capacity utilizatio had sigificat predictive power for chages i cosumer price iflatio before 983, this relatioship has substatially weakeed sice the ed of 982. I fact, after 982 there is o evidece that high capacity utilizatio rates forecast icreases i cosumer price iflatio. For chages i producer price iflatio, we fid a sigificat positive predictive relatioship before 983 that is eve stroger tha the pre-983 capacity utilizatio cosumer price iflatio relatioship. Additioally, although there is some deterioratio i the relatioship betwee chages i producer price iflatio ad capacity utilizatio after 982, there is still evidece of a sigificat positive predictive relatioship, especially at forecast horizos of six moths. There are a umber of possible explaatios for the deterioratio i the ability of capacity utilizatio to forecast chages i iflatio, icludig potetial mismeasuremet of capacity utilizatio ad a icreasigly global ecoomy. Aother potetial explaatio for the deterioratio i the forecastig relatioship is that the coduct of moetary policy has chaged (Cecchetti 99). I fact, may aalysts have argued that the Federal Reserve has bee more forward-lookig ad quicker to brig iflatio pressures uder cotrol durig the 98s ad 99s tha durig the late 96s ad 97s (Balke ad Emery 994). If the Federal Reserve is ow quicker to tighte policy i respose to such idicators as risig capacity utilizatio rates, these idicators may o loger be followed by risig iflatio, simply because the Federal Reserve has already tighteed policy ad brought iflatio pressures uder cotrol. Importatly, however, the policy implicatio is ot that the Federal Reserve should stop moitorig the utilizatio rate. After all, it is because the Federal Reserve has moitored the utilizatio rate as a idicator of risig iflatio pressures that iflatio has remaied relatively stable. I ay case, future research should focus o establishig the validity of the moetary policy explaatio versus others that are put forward. Notes We thak Natha Balke, Carl Boham, Joh Duca, Joseph Haslag, Eva Koeig, Charles Steidel, ad a aoymous referee at the Board of Goverors for helpful commets ad suggestios. Output gap aalyses similarly reflect the belief i a stable short-ru Phillips curve. I gap aalysis, curret output above estimated potetial output sigals risig iflatio. 2 For examples, see the Board of Goverors (994), Citibak (996 a, b, ad c), Cooper ad Madiga (996 a ad b), ad Merrill Lych (996). 3 The recet literature examiig the uemploymet iflatio relatioship icludes Duca (forthcomig), Fuhrer (99), Kig, Stock, ad Watso (99), Weier (993), ad Koeig ad Wye (994). 4 The atural rate of uemploymet is also cosidered the log-ru rate of uemploymet to which the ecoomy teds over time. To see that this assumptio is ecessary, cosider a empirical Phillips curve model i which a distributed lag of past iflatio proxies for expected iflatio: Π = K + B U + Π t t λ (as i Figure 2), i t i where Π ad U are the level of iflatio ad the uemploymet rate ad E( Π i ) = λ i Π case where λ i =, the t i i = t t t i. I the special Π Π = K + B U + u( Π Π ), t i t i i where u i = j λ. I the log ru, this equatio j= has the form of Figure 3 whe lagged chages i iflatio equal zero. 6 The NAIRU ad NAICU are special cases of the atural rates where the expectatios assumptio described above is ivoked. 7 Basically, the failure of this assumptio is the Lucas critique. 8 Fi (99) is aother study that fids a positive relatioship betwee capacity utilizatio ad iflatio. However, the issue of stability is ot addressed. 9 Equatio icludes two dummy variables to cotrol for the Nixo wage ad price cotrols. It also icludes lags of chages i relative petroleum price iflatio to cotrol for eergy price shocks. Oe dummy variable equals oe for the year 972, ad the other equals oe for the years The relative price of petroleum iflatio is from the producer price idex. Coitegratio is ot a cocer, as augmeted Dickey FEDERAL RESERVE BANK OF DALLAS 9 ECONOMIC REVIEW FIRST QUARTER 997

7 Fuller tests idicate that the chage i iflatio ad capacity utilizatio rates are statioary. The Akaike iformatio criterio (AIC) ad the Schwartz iformatio criterio (SIC) idicate that oe year of lagged chages i iflatio ad eergy price iflatio is sufficiet. The results are qualitatively uaffected whe oly six moths of lagged iformatio is used. Oly oe lag of the capacity utilizatio rate is icluded because additioal lags are statistically isigificat. Policymakers ad fiacial market participats are more cocered with treds i iflatio rather tha the more oisy mothly chages i iflatio. The quarterly ad semiaual data are costructed from the ed-ofperiod mothly data. 2 There are o qualitative differeces whe we ru these regressios usig core-cpi iflatio, which excludes food ad eergy prices. 3 Other reasos for choosig Jauary 983 iclude a chage i the Federal Reserve operatig procedures at this time ad a chage i the behavior of iflatio (Emery 994). The qualitative ature of the results is robust with respect to other dates ear the ed of We use a likelihood ratio test to examie for istability i ay or all of the coefficiets. However, whe all of the coefficiets i the model are allowed to vary across the two samples, the data are ot strog eough to reject the hypothesis that the coefficiets o lagged capacity utilizatio are equal i both samples at the 9-percet cofidece level. 6 The trade-off is costructed from the coefficiet estimate o the costat term ad the capacity utilizatio term, settig all the other coefficiets equal to zero. 7 The cofidece bads reflect oly the ucertaity associated with coefficiet estimates o the costat term ad the capacity utilizatio term, ot the ucertaity reflected i the error term of the regressio. 8 The figures are qualitatively similar for the quarterly ad semiaual data. Refereces Balke, Natha S., ad Keeth M. Emery (994), Uderstadig the Price Puzzle, Federal Reserve Bak of Dallas Ecoomic Review, Fourth Quarter, 26. (996b), Slowig ito 997, Cool Ecoomy Will Pull Yields Dow, Ecoomic Week, Jue 24, 2. (996c), Forget the Widy Talk. The Iflatio Wids Are t Risig, Ecoomic Week, Jue, 2. Cooper, James C., ad Kathlee Madiga (996a), A Pricey Stock Market Makes the Fed s Job Tougher, Busiess Week, July, (996b), Eve the Bod Market Is Startig to Chill Out, Busiess Week, May 27, Duca, Joh V. (forthcomig), Iflatio, Uemploymet, ad Duratio, Ecoomics Letters. Emery, Keeth M. (994), Iflatio Persistece ad Fisher Effects: Evidece of a Regime Chage, Joural of Ecoomics ad Busiess 46 (August): 4 2. Fi, Mary G. (99), Is High Capacity Utilizatio Iflatioary? Federal Reserve Bak of Richmod Ecoomic Quarterly 8 (Witer): 6. Fraz, Wolfgag, ad Robert J. Gordo (993), Germa ad America Wage ad Price Dyamics, Europea Ecoomic Review, May, Fuhrer, Jeffrey C. (99), The Phillips Curve Is Alive ad Well, Federal Reserve Bak of Bosto New Eglad Ecoomic Review, March/April, 4 6. Garer, Ala C. (994), Capacity Utilizatio ad U.S. Iflatio, Federal Reserve Bak of Kasas City Ecoomic Review, Fourth Quarter, 2. Kig, Robert G., James H. Stock, ad Mark W. Watso (99), Temporal Istability of the Uemploymet Iflatio Relatioship, Federal Reserve Bak of Chicago Ecoomic Perspectives, May/Jue, 2 2. Koeig, Eva F., ad Mark Wye (994), Is There a Output Iflatio Trade-Off? Federal Reserve Bak of Dallas Southwest Ecoomy, Issue 3, 4. Board of Goverors of the Federal Reserve System (994), Moetary Policy Report to the Cogress Pursuat to the Full Employmet ad Balaced Growth Act, Federal Reserve Bulleti, March, Cecchetti, Stephe G. (99), Iflatio Idicators ad Iflatio Policy, NBER Macroecoomics Aual, Merrill Lych (996), Credit Market Memo, Jue 4. Weier, Stuart E. (993), The Natural Rate ad Iflatioary Pressures, Federal Reserve Bak of Kasas City Ecoomic Review, Third Quarter, 9. Citibak (996a), Europe Igores How Japa Fially Got Out of Its Slump, Ecoomic Week, July 8, 3. 2

This article is part of a series providing

This article is part of a series providing feature Bryce Millard ad Adrew Machi Characteristics of public sector workers SUMMARY This article presets aalysis of public sector employmet, ad makes comparisos with the private sector, usig data from