Docstoc

Hamilton Calling Recessions in

Document Sample
Hamilton Calling Recessions in Powered By Docstoc
					Calling Recessions in Real
          Time
      James D. Hamilton
     Dept of Econ, UCSD
I. Overview of some of the issues
II. Track record of alternative approaches
 Date of recession    Announcement lag

 peak       trough     peak       trough

Jan 1980   Jul 1980   5 months   12 months

Jul 1981   Nov 1982   6 months   8 months

Jul 1990   Mar 1991   9 months   21 months

Mar 2001   Nov 2001   8 months   28 months
Is our objective to:
• predict at t whether we will be in a
  recession at t + j
or
• predict at t whether we were in a recession
  at t - j

Theme: It’s very hard even to do (2) in real
 time.
Why should it be hard?
(1) recessions result in part from forecast
  errors
  (a) Fed misjudges economy
  (b) Firms misjudge markets
(2) economic relations change over time
        June labor force participation rate
               (women aged 35-44)

90
80
70
60
50
40
30
 1945     1955   1965   1975   1985   1995    2005
         June labor force participation rate
                 (men aged 45-54)

100
 90
 80
 70
 60
 50
 40
  1945     1955   1965   1975   1985   1995    2005
Why should it be hard?
(1) recessions result in part from forecast
  errors
(2) economic relations change over time
(3) data revisions
Source:
Leamer (2008)
Nonfarm payroll employment as reported on different dates
 What is the definition of a recession?
Possible answers:
 A. Ad-hoc qualitative summary of
 observable data (e.g., 2 quarters of falling
 real GDP)
 B. It’s a recession if and only if the NBER
 says so
 C. A recession is an objective but
 unobserved determinant of the data
I. Overview of some of the issues
II. Track record of alternative approaches
    A. Predicting an ad-hoc event
Ray Fair (1993)
   y 1t GDP growth in quarter t
              1 if y 1t 0 and y 1,t 0
                                     1
    St 
              0 otherwise
             1   p 
    y t c  1 y t   p y t  t
model implies
  Pr t 1|y t , y t , . . . . 
      S j             1
Stock-Watson experimental
recession index (1988-1993)
    y t    t  t
              L c u
   D ut 
       L       t

   L t    t
     c           
              c s s
   S t 1 if  t  0  B t
                     8

   B t inferred to approximate NBER
In-sample: P(t|t)
In-sample: P(t+3|t)
In-sample: P(t+6|t)
Out-of-sample: P(t+6|t)
Out-of-sample: P(t+3|t)
Out-of-sample: P(t|t)




    Recession began: July 1990
    P(t|t) > 0.5 by Nov 1990
I. Overview of some of the issues
II. Track record of alternative approaches
    A. Predicting an ad-hoc event
    B. Predicting what the NBER is going to
   say
Pr t 1|y t F t ; 
  S j            y
Choose F
         . and y t
Estimate 
                          Katayama (LSU, 2008)
Interest Rates
•   FF Federal Funds rate                   Stock Price Indices
•   3M 3-month Treasury Bill rate           •   DJ30 3-mo Δ log Dow Jones 30 average
•   5Y 5-year Treasury Bond rate            •   SP500 3-mo Δ log S&P 500 stock price
                                                index
•   10Y 10-year Treasury Bond rate
•   AAA Moody's corporate bond yield        Monetary Aggregates
•   AA Moody's corporate bond yield         •   M0 Monetary base (log-differenced)
•   A Moody's corporate bond yield          •   M1 (log-differenced)
                                            •   M2 (log-differenced)
Term Spreads
•   TS10YFF 10Y-FF Treasury term spread     Other Macroeconomic Variables
•   TS10Y3M 10Y-3M Treasury term spread     •   CLI11 Δ log composite leading indicators
•   TS10Y5Y 10Y-5Y Treasury term spread     •   CPI, all urban, all items (log-differenced)
                                            •   EXP Consumer expectation
Credit Spreads
                                            •   EXPD Changes in consumer expectation
•   CSAAA AAA - 10Y spread
                                            •   HOUSE Building permits (log-differenced)
•   CSAA AA - 10Y spread
                                            •   VENDOR performance
•   CSA A - 10Y spread
                                            •   INCOME Δ log personal income
Employment Data                             •   IP Industrial production (log-differenced)
•   EMP Δ log non-agricultural employment   •   SALES Δ log Manufacturing & trade sales
•   CEMP Δ log civilian employment
•   UICLAIM Δ log unemployment claims
•   UNEMP Unemployment rate
•   UNEMPD Change in unemployment rate
•   HOURS Δ log manufacturing hours
Evaluated with 7 different choices for F(.) by
 post-sample and leave-2-years-out cross-
 validation
Conclusion:
Improvements from F(.) with positive skew and
  excess kurtosis

Best variables:
• 10Y-3M treasury spread
• S&P500 3-month growth
• employment growth
Chauvet and Potter (2002, 2005)
 Probit specification based on term spread
 allowing for serial correlation and
 structural breaks successfully predicted
 2001 recession
Wright (2006)
• F(.) ~ Normal
• 10Y-30M treasury spread
• fed funds rate
• tries to predict an NBER recession any
  time within next 12 months
Leamer (2008):
  Choose thresholds for 6-month changes
  so as to fit NBER dates
I. Overview of some of the issues
II. Track record of alternative approaches
    A. Predicting an ad-hoc event
    B. Predicting what the NBER is going to
   say
    C. Recognizing a shift in the observed
   dynamics of economic variables
y t GDP growth for quarter t
        1 if recession at t
St 
        0 if not recession at t
y t  s t  t
            
 N 
  t      0, 2

S t unobserved
Pr t j|S t ip ij
    S         1
                    Density of expansions

0.15
 0.1
0.05
                                                               = 4.7
                                                               = 3.5
  0
       -15   -12   -9   -6   -3   0   3     6   9   12   15
                             GDP growth




                    Density of recessions

0.15                                                           = -1.2
 0.1                                                           = 3.5
0.05
   0
       -15   -12   -9   -6   -3   0   3     6   9   12   15
                             GDP growth
                           Density of mixture

0.12
 0.1
0.08                                                      expansion
0.06                                                      recession
0.04                                                      mixture
0.02
   0
       -15 -12   -9   -6   -3   0   3   6   9   12   15
                                Density of mixture

0.12
 0.1
0.08                                                                      expansion
0.06                                                                      recession
0.04                                                                      mixture
0.02
   0
       -15 -12   -9       -6    -3      0      3      6    9    12   15



                                       Pr( St  2, yt )
                 Pr( St  2 | yt ) 
                                            f ( yt )
                              Pr( St  2, yt )
                 
                     Pr( St  1, yt )  Pr( St  2, yt )
                 Pr( St  2, yt )  Pr( St  2)  f ( yt | St  2)
                        Density of mixture

0.12
 0.1
0.08
0.06
0.04
0.02
   0
       -15   -12   -9   -6   -3    0    3     6   9   12   15



                   Probability of recession

  1
0.8
0.6
0.4
0.2
  0
       -15   -12   -9   -6   -3    0    3     6   9   12   15
                        GDP growth in quarter t
Filter inference:
     Pr t 1|y t , y t , . . . , y 1 
         S              1

Smoothed inference:
     Pr t 1|y T , y T , . . . , y 1 
         S                1
                    Contributions to percent change in
                       real gross domestic product

6

5

4                                                    GDP
                                                     Consumption
3                                                    Nonresidential fixed investment
                                                     Residential fixed investment
2
                                                     Change in inventories
1                                                    Exports
                                                     Imports
0                                                    Government

-1

-2
     2007:Q3   2007:Q4     2008:Q1     2008:Q2
 Chauvet and Hamilton (2006), Chauvet and Piger (2008)


                1 
Ft   t  Ft  t
        s

y rt  r Ft  rt v
v rt  r v r,t  rt
                1   
                ln sales
                ln pers income
yt 
                ln civ employ
                ln ind prod
   Month             Probability of Recession
February 2008                15.4%
 March 2008                  16.0%
 April 2008                  15.6%
 May 2008                    15.3%
 June 2008                   14.0%
  July 2008                  13.0%



 Source: Jeremy Piger, Sept. 29, 2008
Source: Jeremy Nalewaik
Source: Jeremy Nalewaik
Hamilton (2005)
y t unemployment rate
             1  2 
y t c s t  1 y t  2 y t  t
          1 if expansion
St       2 if mild recession
          3 if severe recession

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:2
posted:2/16/2012
language:
pages:61