The twelve-month CPI and HICP rose 1.1% and 1.3% to December, and these numbers were duly headlined. But both indices have been falling in recent months, and it beats me why people use 12-month numbers in the middle of a big macroeconomic correction. Karl Whelan made a similar point here recently in the context of the quarterly national accounts.
There are small but significant seasonals in the CPI and HICP. The CSO does not adjust, but the following is based on up-to-date factors (an Excel file with the data and factors is available from firstname.lastname@example.org). Unadjusted, HICP showed small monthly changes in Sept, Oct, Nov, then fell 0.73% in Dec. The adjusted pattern was similar, but the fall in December lower, at 0.46%.
For unadjusted CPI, Sept and Oct showed only small changes, but then big falls of 0.93 and 1.21 in Nov and Dec. The adjusted falls were again smaller at 0.84 and 1.09 (but these are still very large month-on-month numbers).
Thus for the last two months, and seasonally adjusted, the CPI has dropped almost 1% per month. The HICP has been falling only for a month, and more slowly. The difference between the two is mostly about mortgage interest (see my paper in ESRI QEC for September 2007), and I think the HICP is a better index. When I expressed this view in 2007, the CPI was rising faster than the HICP, and my argument was described as academic (ie wrong) by ‘certain parties’ keen on compensation for inflation. A change of horse by these parties is confidently predicted (difficult manoeuvre at speed unless you are a Cossack).
Recent forecasts of FY 2009 CPI inflation are minus 2% (ESRI) and minus 2.5% (Pat McArdle of Ulster Bank). These numbers look well within range, but HICP could show a smaller fall than CPI if mortgage interest rates continue to drop. Either way, we are a long way away from mid-September, when the pay deal was negotiated. At that time, inflation looked set in a band around plus 4%.
There is a big shift in the price index seasonals from December to January – if the adjusted trend is zero, the unadjusted shows a significant drop. If anyone quotes the unadjusted drop next month, they incur four faults.