Newspapers, including the NYTimes, picked up the news on the gap between the number of those over age 100 in official records and those who seem to actually be still alive. It's big: there clearly was little effort to check local household registries and to correct obvious errors. Not just here and there, maybe 230,000 people short. That discrepancy itself is not of much concern to economists, because a population census is done in Japan (and elsewhere) exactly because of the difficulty of maintaining and then appropriately aggregating local records, made more challenging by the incentives to overreport all along the way, from local governments to the Ministry of HWL, which might have suspected that its numbers were a bit high. I pick out three additional systematic data issues below, and then generalize at the end.
Collecting data in a timely and accurate and meaningful manner is hard. In this case, tax ID numbers would help reconcile records for those who change residences to a nursing home or an ash urn. [A modest death benefit such as many systems provide does wonders for getting elderly spouses and children to report deaths, helped because hospitals and doctors also records deaths by national ID numbers.] No economy's statistical system is without weaknesses. However I think people looking at Japan tend to overlook their magnitude.
One component, discussed previously on NBR*, are sometimes seemingly minor conceptual and administrative differences. We need to remember that unemployment data are collected on a different basis in Japan, the UK, the US and the various countries of the EU. Those differences in methodology make inter-country comparisons potentially useless, absent a careful comparison of the nitty-gritty of data collection. The OECD makes a few cursory adjustments but their data remain nevertheless low in comparability, because countries simply do not routinely do supplemental surveys using the methodology of other countries that would allow systematic comparison.
GDP is an important example. The IMF has a study detailing efforts for certain types of data (e.g. GDP) across many economies; Japan has significant gaps (the Japan-specific portion runs about 200 pages...). And Japan's various statistical staff are all targets of [proposed] budget cuts, so we should worry about having any reliable numbers...they have neither staff / budget nor (as I argue below) expertise to do extra work. One of the biggest holes is the lack of consolidated data on government activities. Basics (at least to me) such as revenues and expenditures are only available 2 years after the fact, and in summary form rather than with programmatic detail. It's easy to overlook; I was always frustrated that certain GDP numbers weren't available on a quarterly basis, but I hadn't stopped to think about the implications, that no one had good consolidated government data on anywhere near a current basis. It was only when I tried to put together numbers for a couple things this summer that it dawned on my how large the gap is even with the US, which has its own data issues. [I can't speak of the EU end of things because I look at such data but once in a blue moon; my understanding however is that Japan remains very much an outlier.]
Detailed data can be worse. I wanted to know health care expenditures. Not part of GDP. I checked individual ministry sites, and the MOF budget site. While it is probably possible to put together the bigger items, the most recent consistent compilation of health care expenditures that I could find (realized, as opposed to budgeted) was for 2007, and it was only posted in the past month. The lack of consistent, timely government statistics and the very small number of categories available for consumer expenditures on a GDP basis means for example that a consistent health care share of GDP number cannot be produced on a timely manner, if at all. Instead various government accounts and the household consumption survey (kakei chosa) need to be patched together, with little assurance that there aren't both categories missed and categories that include double-counting. I was flabbergasted, because it seemed to me that in Japan of all places having a good grasp of health care expenditures would be a real priority. [To be fair, most expenditures seem to be in a couple big programs -- but those programs are administered locally, and the what- we-actually-spent numbers come out only long after the fact.]
This problem is a general one for data on services -- but Japan as with other "developed" (OECD) economies is a service economy. There are mountains of data on agriculture and fisheries, mounds of data on manufacturing, but only sparse data on services. For retailing, only a few categories are available, and they badly lag structural change in the sector. Of course at a detailed level that's an insurmountable problem, you can't ask about categories until they exist, and then you are constrained by the extent of recognition of those categories by the people who fill out the surveys or otherwise collect data. For example, in the US case there was a 1+ percentage point shift in the consumer price index in the month that Walmart was first incorporated in the set of stores from which price data were collected. But Japan is worse.
Efforts seem to be particularly constrained by the small number of staff, and by the fact that some of them are there not because of any speciality or experience but merely as a function of periodic ministerial, prefectural and city government personnel rotations.
In short, in something that is highly technical, there is a shortage of professional skills among those actually assigned to statistical functions. My observation it that the expertise issue is pervasive in Japanese government units, central and local (and in national universities, for example how IT is handled). It is thus not unique to economic data -- it's not odd that a regulator knows nothing of the industry they're supposed to regulate, whatever the ministry. MOFF staff get rotated from field extension for a particular crop [in the case I have in mind, their college specialty was just that sort of agronomy] to supervising the audits of financial management of local nokyo [not to their taste, not in their skill set, and not at their level of clout to begin addressing the slipshod practices they encountered by people in technical functions similarly removed from their past experience ... though more nefarious explanations sometimes came to mind].
It's not that the numbers are fabricated, or deliberately skewed. [Indeed, as Arthur Alexander pointed out on NBR, neither are the quarterly data unusually slow in publication or large in their revisions.] They're just, well, not all there. So we see large revisions in GDP data [with a significant change in methodology now being rolled out -- so much for comparability across time]. That's an inconvenience, for someone who tries to argue based on data.
I've not "lived on the inside" and suspect there are mechanisms to help compensate for the general bias I see towards, well, a lack of professionalism. When I was in a Japanese bank rotations were kept to about 40% a year, and the non-formalized nature of job assignments meant that low-status but experienced staff were delegated work far beyond what their title suggested, while some of the grunt work got done by purportedly senior people [or sometimes no work at all]. But basically, banking isn't that hard, and bankers shouldn't be too bright or skilled or a bank can get into trouble trying new things... [I've heard bankers claim that in the old days US banks didn't hire anyone with a stellar GPA – using themselves as examples.] A statistical agency however can't function if only a handful of staff at any given time know what they're doing. And that's true of lots of other government functions. If I'm accurate, then they are at the same time both systematically overstaffed and underperforming, and not due to laziness or ineptness, but because the staff are too new to be able to work efficiently and so stuff just doesn't get done....
Original post on NBR: September 14, 2010
* First published on the NBR Japan Forum online discussion list
maintained by The National Bureau of Asian Research
archived at http://nbrforums.nbr.org/foraui/list.aspx?LID=5.
Readers might be interested in Landefeld, J. Steven; Seskin, Eugene P; Fraumeni, Barbara M. "Taking the Pulse of the Economy: Measuring GDP." Journal of Economic Perspectives, Spring 2008, 22:2, p193-216. It is available to students at W&L under the library's subscription to journal database Business Source Complete.
Note: subsequent to my writing this the following (Japanese-language) working paper came out.
ESRI Discussion Paper Series No.249 "On Improving the Estimation Method of Japanese Quarterly GDP"
ESRI Discussion Paper Series No.249 "On Improving the Estimation Method of Japanese Quarterly GDP"
The abstract actually is translated into English; it notes that the focus is on improving the seasonal adjustments and investment data component of GDP to decrease the magnitudes of the revisions between the first "flash" release and the second revised data [and the subsequent final report]. The article itself is filled with matrix algebra, and with considerations of data sources and the treatment of inventories, changes of which are one component of investment. Additional background can be found in two working papers of the authors on the University of Tokyo Center for International Research on the Japanese Economy, as well as links to many other recent papers. (The Japanese-language portion provides details on graduate workshops, including recent and upcoming presentations, with PDFs of many of the papers -- most of which are in English.)FYI the title, authors' names and abstract follow:GDP 速報の推定法の改善について