In our inv mtg today, I talked about U.S. job growth but I'm not sure the team was very receptive to my message. As I didn’t want to waste the hours I spent researching this, I will post a few brief thoughts here:
1) labour market has not been at all “robust” during this cycle
a) payrolls growth has barely surpassed population growth since end of 2001
recession, and, of course, has trended further down
b) this truly has been a jobless recovery, as payrolls growth since end of
'01 recession is way under that of '90s, '80s, '70s, etc. post-recessionary
periods
c) since end of recession (Nov01), jobs are up only 0.9% annualized
d) and since trough of payrolls (May03) (jobs kept being lost for
year-and-half even after end of recession), jobs up only 1.5% annually
e) obvious conclusion – job growth had little to do with economic strength of last four years (perhaps consumer debt, at record high, thanks to MEW, was the real key?)
2) NFP estimates not particularly reliable at turning points in economy,
says BLS itself
a) I was trying to make sense of the recent 180k print, as we discussed, as
it doesn't yet fit our expectations, so wondered if we needed to refine our
assumptions, so started investigating how NFP actually gets produced monthly (http://www.bls.gov/web/cesbd.htm)
i) First, they do a sampling of 160,000
businesses, representing about 1/3 of total jobs in the economy
ii) Second, they use their birth/death model
(estimation procedure) to impute employment for new businesses (and then use an
econometric model to estimate the residual that can’t be accounted for by using
dying businesses to estimate new businesses) --- the model assumes a seasonal
pattern, which, on net for the year from Apr06-Mar07, assumes creation of
963,000 jobs (80k/mth)
iii) Third, they take the result of the sampling and the modeling and apply
their seasonal adjustment
iv) They admit the procedure “is likely to have some difficulty
producing reliable estimates at economic turning points or during periods when
there are sudden changes in trend… [the birth/death estimation] is likely to
remain as the most problematic part of the estimation process”
v) March payrolls growth of 180,000 (seasonally adjusted) embeds an
assumption of 128,000 just from the birth/death model (prior to the SeasAdj)
b) BLS first does their monthly process as above, but over time they also
do a census to update their data; this census covers 8.8 million employers (vs
160,000), and 135.0 million F/T and P/T workers, or 98% of U.S. employees (vs
about 1/3)
i) On Apr 11 they released the results for Q3 (http://www.bls.gov/news.release/cewqtr.nr0.htm)
ii) these showed job growth in year to Sep06 of
only 1.5% (vs 1.8% from NFP) and average wage growth of just 0.9%
(vs 4.1% from NFP estimates) (http://www.bls.gov/news.release/cewqtr.t01.htm)
iii) in total for the 12mth period, according to the census, jobs increased
2.01 million – compare this to the total birth/death adjustment that NFP uses
in its estimates (0.96 mil), and the latter that they “assume” in their model
represents almost half of confirmed job growth – so the model’s validity is
crucial if the NFP data is to be taken at face value
iv) another obvious conclusion: the census #s are clearly more robust than the NFP estimates, albeit delayed, but hardly indicate that income growth will support consumption when confronted with sagging home prices, higher interest rates, higher energy bills, resetting ARMs, tighter lending standards, auto and housing recessions, rising food prices
No comments:
Post a Comment