Monday, July 27, 2015

Rebonato on Bond-Yield Econometrics

Riccardo Rebonato (R) has a fascinating new paper, which builds most directly on important earlier work of Cieslak and Povala (2010) (CP). 

The cool thing about CP is the way it advances and blends both the spanning literature ("all information of relevance for yield prediction is embedded in the current term structure," e.g. via forward-rate tent functions as in Cochrane-Piazessi (2004)), and the non-spanning literature ("not all information of relevance for yield prediction is embedded in the current term structure," e.g. because certain macro variables seem to help predict risk premia, as in Ludvidson and Ng (2009)).

In turn, the cool thing about R is its insightful high-frequency / low-frequency interpretation of CP, with the macro predictors of primary relevance at low frequencies. 

Adapted from the R abstract:


This paper presents a simple reformulation of the restricted CP return-predicting factor which retains by construction exactly the same (impressive) explanatory power as the original one, but affords an alternative and attractive interpretation. What determines future returns, the new factor shows, is ... the distance of the yield-curve level and the slope not from fixed reference levels, but from conditional ones determined by ... long-term inflation.

Intuitively (if not exactly): We can predict the yield curve using its current deviation from its long run mean ("reference level"), but that long run mean itself varies slowly with macroecoomic conditions.

High-frequency / low-frequency decompositions have a long and distinguished history in time-series econometrics, from cycle / trend real-output decompositions in macro-econometrics (e.g., Cochrane (1988)) to short-run / long-run volatility decompositions in financial econometrics (e.g., the "component GARCH" model of Engle and Lee (1999)).

In the CP/R bond-yield context, I'm immediately reminded of key early work by Kozicki and Tinsley (2001) on market perceptions of central bank credibility providing low-frequency anchoring for long yields.

A final thought: Interestingly, recent work of Bauer and Hamilton (2015) questions the entire non-spanning literature. Perhaps more on that in a subsequent post, and its relation to CP and R (e.g., why worry about blending the spanning and non-spanning approaches if the non-spanning approach is suspect?).

Sunday, July 19, 2015

Introducing Ben Connault

I should introduce Benjamin ("Ben") Connault, Penn's newly-hired young econometrician, arriving from Princeton any day now. We're extremely grateful to Bo Honoré, Ulrich MüllerAndriy Norets, and Chris Sims for sending him our way.

Frank is the de facto required name for male Penn econometricians (as in Frank Schorfheide, Frank DiTraglia, and yours truly), but Ben somehow managed to dodge the requirement. At any rate, if his name is strikingly original by Penn's standard, so too is his research, by any standard. Very much to his credit, Ben is an independent thinker whose econometrics isn't easily catagorized. Just check him out for yourself. We look forward to great things from him.


Welcome, Ben!

Monday, July 13, 2015

What Seasonally-Adjusted U.S. Economic Data Needs Most...

...is non-adjustment. This is not a minor issue: there's not even an unadjusted U.S. GDP! 

Seasonal adjustment is sometimes desirable, but sometimes not. Sometimes it's done poorly, sometimes it's better done with extra care and transparency by the researcher, etc. And the restrictions implied by economic theory generally hold across all frequencies, not just non-seasonal frequencies. There are many, many issues. (See, for example, the discussion in Hansen and Sargent (1993) and the references there.) 


Sometimes seasonality is the center of attention, and hence of intrinsic interest, so access to unadjusted data is crucial. But even when seasonality is arguably just a "nuisance," it's valuable to have access to unadjusted series, which are more fundamental. If I have an unadjusted series, I can adjust it myself, and then you and I can have a potentially valuable discussion as to how and why I adjusted it. In contrast, if I have only an adjusted series, in general I have no way to recover the underlying unadjusted series, so you and I have no choice but to rely completely on agencies' seasonal-adjustment procedures.

Let me be clear: Both academic researchers and the data-providing agencies have made important seasonal-adjustment advances over many decades. I'm grateful and I hope they continue. I'm simply saying that we should also have access to unadjusted data. So let me add to the seasonality research "to-do list" that I recently offered


Any series provided in seasonally-adjusted form should also be provided in unadjusted form. 

Sunday, July 5, 2015

Being a Millionaire Isn't What it Used to be

One evening a few weeks ago, some friends and I wound up talking about the "roaring twenties" in the U.S., and all the "millionaires" created, and wondering just what $1 million 1925 dollars would be in 2015 dollars. Obviously the price level has multiplied greatly since 1925, but how many times? Five? Fifteen? Fifty? None of us were really sure.

The handy CPI calculator at FRB Minneapolis came to the rescue: $1 in 1925 is $13.67 in 2015. That is, you'd need $13.67 million in 2015 to have the purchasing power of someone with $1 million in 1925!

If you want to dig a little deeper, the Fed's full annual CPI data 1801-2015 appear in the table below (year, CPI price level, inflation rate). Note that the price level was stable during 1801-1913, after which it grew steadily forevermore. Quiz: Besides steady inflation, what didn't exist in the U.S. before 1913 but has been with us ever since? You know the answer.

Don't get me wrong. I'm not wishing we were back in 1880 with no Federal Reserve System. But the price level pattern certainly does suggest that the benefits delivered by central banks come at a significant cost, namely the inflation tax, which is highly regressive, borne disproportionately by the unsophisticated poor.


1801
50
-2.0%
1802
43
-14.0%
1803
45
4.7%
1804
45
0.0%
1805
45
0.0%
1806
47
4.4%
1807
44
-6.4%
1808
48
9.1%
1809
47
-2.1%
1810
47
0.0%
1811
50
6.4%
1812
51
2.0%
1813
58
13.7%
1814
63
8.6%
1815
55
-12.7%
1816
51
-7.3%
1817
48
-5.9%
1818
46
-4.2%
1819
46
0.0%
1820
42
-8.7%
1821
40
-4.8%
1822
40
0.0%
1823
36
-10.0%
1824
33
-8.3%
1825
34
3.0%
1826
34
0.0%
1827
34
0.0%
1828
33
-2.9%
1829
32
-3.0%
1830
32
0.0%
1831
32
0.0%
1832
30
-6.3%
1833
29
-3.3%
1834
30
3.4%
1835
31
3.3%
1836
33
6.5%
1837
34
3.0%
1838
32
-5.9%
1839
32
0.0%
1840
30
-6.3%
1841
31
3.3%
1842
29
-6.5%
1843
28
-3.4%
1844
28
0.0%
1845
28
0.0%
1846
27
-3.6%
1847
28
3.7%
1848
26
-7.1%
1849
25
-3.8%
1850
25
0.0%
1851
25
0.0%
1852
25
0.0%
1853
25
0.0%
1854
27
8.0%
1855
28
3.7%
1856
27
-3.6%
1857
28
3.7%
1858
26
-7.1%
1859
27
3.8%
1860
27
0.0%
1861
27
0.0%
1862
30
11.1%
1863
37
23.3%
1864
47
27.0%
1865
46
-2.1%
1866
44
-4.3%
1867
42
-4.5%
1868
40
-4.8%
1869
40
0.0%
1870
38
-5.0%
1871
36
-5.3%
1872
36
0.0%
1873
36
0.0%
1874
34
-5.6%
1875
33
-2.9%
1876
32
-3.0%
1877
32
0.0%
1878
29
-9.4%
1879
28
-3.4%
1880
29
3.6%
1881
29
0.0%
1882
29
0.0%
1883
28
-3.4%
1884
27
-3.6%
1885
27
0.0%
1886
27
0.0%
1887
27
0.0%
1888
27
0.0%
1889
27
0.0%
1890
27
0.0%
1891
27
0.0%
1892
27
0.0%
1893
27
0.0%
1894
26
-3.7%
1895
25
-3.8%
1896
25
0.0%
1897
25
0.0%
1898
25
0.0%
1899
25
0.0%
1900
25
0.0%
1901
25
0.0%
1902
26
4.0%
1903
27
3.8%
1904
27
0.0%
1905
27
0.0%
1906
27
0.0%
1907
28
3.7%
1908
27
-3.6%
1909
27
0.0%
1910
28
3.7%
1911
28
0.0%
1912
29
3.6%
1913
29.7
2.4%
1914
30.1
1.3%
1915
30.4
0.9%
1916
32.7
7.7%
1917
38.5
17.8%
1918
45.2
17.3%
1919
52.1
15.2%
1920
60.2
15.6%
1921
53.6
-10.9%
1922
50.3
-6.2%
1923
51.2
1.8%
1924
51.5
0.4%
1925
52.7
2.4%
1926
53.2
0.9%
1927
52.2
-1.9%
1928
51.6
-1.2%
1929
51.6
0.0%
1930
50.2
-2.7%
1931
45.7
-8.9%
1932
41.0
-10.3%
1933
38.9
-5.2%
1934
40.2
3.5%
1935
41.2
2.6%
1936
41.7
1.0%
1937
43.2
3.7%
1938
42.3
-2.0%
1939
41.8
-1.3%
1940
42.1
0.7%
1941
44.2
5.1%
1942
49.1
10.9%
1943
52.0
6.0%
1944
52.9
1.6%
1945
54.1
2.3%
1946
58.6
8.5%
1947
67.1
14.4%
1948
72.2
7.7%
1949
71.5
-1.0%
1950
72.3
1.1%
1951
78.0
7.9%
1952
79.8
2.3%
1953
80.4
0.8%
1954
80.7
0.3%
1955
80.5
-0.3%
1956
81.7
1.5%
1957
84.4
3.3%
1958
86.7
2.7%
1959
87.6
1.0%
1960
88.9
1.5%
1961
89.8
1.1%
1962
90.9
1.2%
1963
92.0
1.2%
1964
93.2
1.3%
1965
94.7
1.6%
1966
97.5
3.0%
1967
100.2
2.8%
1968
104.5
4.3%
1969
110.2
5.5%
1970
116.7
5.8%
1971
121.7
4.3%
1972
125.7
3.3%
1973
133.4
6.2%
1974
148.2
11.1%
1975
161.7
9.1%
1976
171.0
5.7%
1977
182.1
6.5%
1978
196.0
7.6%
1979
218.1
11.3%
1980
247.6
13.5%
1981
273.2
10.3%
1982
290.0
6.1%
1983
299.3
3.2%
1984
312.2
4.3%
1985
323.2
3.5%
1986
329.4
1.9%
1987
341.4
3.7%
1988
355.4
4.1%
1989
372.5
4.8%
1990
392.6
5.4%
1991
409.3
4.2%
1992
421.7
3.0%
1993
434.1
3.0%
1994
445.4
2.6%
1995
457.9
2.8%
1996
471.3
2.9%
1997
482.4
2.3%
1998
489.8
1.6%
1999
500.6
2.2%
2000
517.5
3.4%
2001
532.1
2.8%
2002
540.5
1.6%
2003
552.8
2.3%
2004
567.6
2.7%
2005
586.9
3.4%
2006
605.8
3.2%
2007
623.1
2.9%
2008
647.0
3.8%
2009
644.7
-0.4%
2010
655.3
1.6%
2011
676.0
3.2%
2012
689.9
2.1%
 2013
700.0
1.5%
 2014
711.4
1.6%
 2015*
720.3
2%