Documentation: Philadelphia Fed's Real-Time Data Set for Macroeconomists First-, Second-, and Third-Release Values

Transcription

1 Documenaion: Philadelphia Fed's Real-Time Daa Se for Macroeconomiss Firs-, Second-, and Third-Release Values Las Updaed: December 16, Inroducion We documen our compuaional mehods for consrucing ime series of firs-, second-, and hird-release values for he macroeconomic variables in he Philadelphia Fed's real-ime daa se for macroeconomiss (RTDSM). Researchers can use hese series o consruc revisions and, perhaps more imporanly, measures of he reliabiliy of early releases of he daa. We also provide caveas on using he daa. In broad erms, we consruc our series in he following manner: We begin wih he raw daa in RTDSM. These daa are expressed in levels. (For each variable in he daa se, RTDSM records he enire ime-series hisory as ha hisory appeared on various vinage daes of he pas.) For mos variables, we ransform he daa ino period-over-period growh raes (monhover-monh or quarer-over-quarer) because revisions o growh raes are ofen more ineresing han revisions o levels. We express growh raes in annualized percenage poins. For a selec few variables, we do no ransform he daa ino growh raes. We exrac he appropriae growh-raes ha represen a governmen saisical agency's firs, second, and hird releases for he observaion, forming hree ime series ha we appropriaely call Firs: The ime series for firs-release values; Second: The ime series for second-release values; Third: The ime series for hird-release values.

2 2 Finally, we also provide a ime series for he laes-revised hisorical values. These are he values of he observaions as hey appear oday (in RTDSM's mos recen vinage of daa available when we made our compuaions). We call his variable: Mos_Recen: The ime series observaions as hey appear oday (as recorded in he mos recen vinage of RTDSM). 2. Revisions Some users of he daa will wan o compue imes series for revisions. Le index he observaion dae. Using he variables defined above, we can easily define many alernaive measures of revisions (Rev). Some examples of ineres follow. Revision From Firs Release o Second Release: Rev12, Second Firs Revision From Second Release o Third Release: Rev23, Third Second Revision From Firs Release o Third Release: Rev13, Third Firs Revision From Firs Release o Mos Recen Release: Rev Mos _ Recen Firs 1 M,

3 3. Compuaional Deails 3 Sep 1. Compue Growh Raes. We begin (in mos cases) by ransforming he raw daa in RTDSM from levels o period-over-period growh raes, expressed in annualized percenage poins. Le X denoe he period- observaion for he level of variable X. The subscrip indexes eiher a quarerly or monhly observaion dae, depending on wheher he variable appears in our daa se wih quarerly or monhly observaions. The monhly and quarerly growh-rae calculaions ( GX ) are: Monh-over-Monh Growh Rae GX 12 X X 1 Quarer-over-Quarer Growh Rae GX 4 X X 1 Figure 1 shows a slice of he Philadelphia Fed's real-ime daa se for he level of real GDP. The columns show he vinages. The rows show he observaions for each vinage. Noice ha he columns are named wih a suffix (yymmm) for he vinage dae of daa hey represen. The column ROUTPUT12M7 shows he observaions for real GDP ha he U.S. Bureau of Economic Analysis made available o he public in he middle of July We highligh six numbers wih red boxes a he boom of seleced columns. These numbers are he level of real GDP on firs, second, and hird release for wo observaions: 2012 Q2 and 2012 Q3. Figure 2 shows he resuls of convering he level of real GDP (Figure 1) ino quarer-overquarer growh raes, expressed in annualized percenage poins.

4 Sep 2. Exrac he Relevan Observaions From he Appropriae Vinages. We now wan o exrac he relevan observaions from he vinages appropriae for firs releases, second releases, and hird releases. We use he following mehod: We exrac he numbers from hree consecuively-daed monhly vinages. 4 We demonsrae he mehod in Figure 2 for firs-, second-, and hird-release values of quarerover-quarer real GDP growh. We highligh he numbers for wo observaions: 2012 Q2 and 2012 Q3 (Figure 2). For 2012 Q2: The firs-release value is percen, he second-release value is percen, and he hird-release value is percen. For 2012 Q3, he values are: percen, percen, and percen. Figure 3 shows how we arrange hese values according o he release. There is one column for firs-release values (Firs), one for second-release values (Second), and one for hird-release values (Third). The las column (Mos_Recen) gives he values as hey appear in he mos recen vinage of RTDSM. Noe ha he Mos_Recen values incorporae he cumulaive effec of all prior revisions o he observaions. 4. Imporan Caveas Imporan caveas accompany he consrucion of our values for firs-, second-, and hirdreleases. One algorihm for exracing firs-, second-, and hird-release values from he Philadelphia Fed's real-ime daa se is o search over vinages of daa for he firs, second, and hird imes each observaion appears. This algorihm has a disadvanage: When a governmen saisical agency delays he release of is daa or changes he schedule on which he daa are released over ime, his algorihm yields esimaes ha could give a misleading impression abou he dae on which he release was available o he public. In some empirical sudies, ambiguiy abou he iming of he release could yield incorrec resuls. Thus, we do no use his approach. A second, closely relaed approach is o exrac he firs-, second-, and hird-release observaions from he vinages in which hey should appear, given he iming (as we know i oday) wih which governmen saisical agencies release heir daa. We use he second approach. The wo approaches will, in general, yield nearly idenical resuls, excep as noed below.

5 5 Variables from he Naional Income and Produc Accouns. The U.S. Bureau of Economic Analysis (BEA) has changed over ime he schedule on which i released is daa. Early in our sample period (1965 Q3 o he presen), firs-releases were ofen delayed relaive o oday's schedule. In addiion, BEA did no always ry o revise is secondrelease values. These problems are more severe he more disaggregaed he variable and he earlier he observaion dae. In our files, he consequences are as follows: In observaions of he disan pas, we code many firs-release values as missing. In observaions of he disan pas, he hird-release values are ofen he same as he second-release values. Federal Governmen Shudown Variables From he Naional Income and Produc Accouns: Observaion for 1995 Q4. A shudown in he federal governmen occurred in lae This affeced he operaions of governmen saisical agencies and produced a delay in he release of daa from he naional income and produc accouns. Accordingly, we record a missing value for he 1995 Q4 firs release of real and nominal GDP and he relaed componens. Variables From he Naional Income and Produc Accouns: Observaion for 2003 Q3. BEA released a benchmark revision o he U.S. naional income and produc accouns on December 10, Prior o he release of he benchmark-revised daa, BEA released is firs esimae for 2003 Q3 U.S. real and nominal GDP and he componens. However, he benchmark revision omied he observaion for 2003 Q3. Accordingly, he daa se correcly includes he firs esimae and (correcly) omis he second esimae.

6 Consumer Price Indexes and Producer Price Indexes. 6 The U.S. Bureau of Labor Saisics (BLS) revises is headline and core consumer price indexes annually afer i re-esimaes new seasonal facors. There are no revisions o he seasonally adjused CPI ouside of hose due o re-esimaed seasonal facors. Consequenly, i makes lile sense o provide second-, and hird-release values using our mehodology: Mos observaions would show no revision o he firs-release value. Therefore, we provide only he values on firs release and he values from he mos recen vinage. Wih one excepion, he naure of revisions o headline and core measures of he producer price index for finished goods is he same as ha for he consumer price indexes. The excepion is ha producer price indexes may be revised hree o four monhs afer heir iniial release because BLS incorporaes addiional informaion from lae reporers and correcions ino is iniial esimaes. Such revisions appear in a monhly vinage ha posdaes he vinage from which we would exrac second- and hird-release values. Thus, as above, we provide only he values on firs release and he values from he mos recen vinage. Indusrial Producion Index and Capaciy Uilizaion. The saff a he Federal Reserve Board of Governors produces he well known monhly indexes of indusrial producion and relaed measures of capaciy uilizaion. The unusual feaure of he Board's saff's release schedule is ha annual revisions are released on a dae ha falls beween he daes of wo monhly repors. (In conras, many governmen saisical agencies release heir annual revisions concurren wih heir normal monhly repors.) The effec on our daa for firs-, second-, and hird-release values is as follows: In defining our values, we sick o he observaions published in he normal monhly repors. (For example, if an annual revision occurs on a dae ha falls beween he monhly vinage for he firs-release and he monhly vinage for he second release, we ake our firs- and second-release values from hose monhly vinages. We do no record as he second-release value he revised observaion published in he annual revision.)

7 Uni Labor Coss and Oupu per Hour. 7 These variables are revised on an aypical basis. We provide he firs-release values and he second-release values bu no values for he hird-release which is made available o he public wo monhs (no one monh) afer he second release. Federal Governmen Shudown in All Variables Excep Housing Sars. The shudown of he federal governmen from Ocober 1, 2013, o Ocober 16, 2013, caused U.S. governmen saisical agencies o delay heir releases of new observaions for variables in he Philadelphia Fed's real-ime daa se. However, he delays were minimal for variables oher han housing sars. For variables whose releases were subjec o hese minor delays, we decided o pospone our collecion of he daa unil he new observaions became available. (The delay was a couple weeks.) This sraegy allowed us o mainain coninuiy in our ime series of firs-, second-, and hird-release values. Our sraegy for housing sars appears below. Federal Governmen Shudown in Housing Sars. As noed above, he federal governmen shudown from Ocober 1, 2013, o Ocober 16, 2013, caused major delays in he Census Bureau's release of new observaions for housing sars: The Census Bureau posponed is release of housing sars for Sepember 2013 and Ocober 2013 unil December 18, Moreover, he delayed firs-release values of hese observaions in December could incorporae source informaion no generally available a he ime of he sandard release schedule. Similar commens apply o second- and hird-release values. Accordingly, we have chosen o define missing values for he observaions of firs-, second-, and hird-release values affeced by he period of he federal governmen shudown.

8 Tom Sark Assisan Direcor and Manager Real-Time Daa Research Cener 8 Economic Research Deparmen Federal Reserve Bank of Philadelphia Ten Independence Mall Philadelphia, PA Tel: Tom.Sark@phil.frb.org

9 Figure 1. Seleced Vinages and Observaions for Real GDP Source: Philadelphia Fed's Real-Time Daa Se, Levels (Chain-Weighed Dollars) The figure shows a slice of he Philadelphia Fed's real ime daa se for he level of real GDP (ROUTPUT). Each column conains he ime series hisory for real GDP as known a he vinage dae lised in he column header. The rows lis he observaions for each vinage dae. We use a red box o highligh he las number in each column. The numbers are he Bureau of Economic Analysis's firs, second, and hird releases for he value of real GDP in 2012 Q2 and 2012 Q3.

10 Figure 2. Seleced Vinages and Observaions for Q/Q Real GDP Growh (Annualized Percenage Poins). 10 The figure shows quarer-over-quarer growh raes (annualized percenage poins) compued from he daa in levels in Figure 1. We use a red box o highligh he las number in each column. The numbers are he Bureau of Economic Analysis's firs, second, and hird releases for he value of real GDP growh in 2012 Q2 and 2012 Q3.

11 11 Figure 3. Firs-, Second-, and Third-Release Values for Q/Q Real GDP Growh (Annualized Percenage Poins) The figure shows he firs-, second-, and hird-release values for quarer-over-quarer real GDP growh over he period 2009 Q1 o 2012 Q4. We highligh he values in red boxes for 2012 Q2 and 2012 Q3. The column labeled "Mos_Recen" shows he corresponding hisorical values as hey were known in he middle of March 2013 when his documenaion was creaed.