All our methods except the CMIR indicate that overseas currency flows are large and growing.
In fact, all of the methods except for that using the CMIR data from Customs suggest that a large amount of currency has gone abroad, and we are inclined to view those expansive estimates as being close to the truth.
Another part of our reason for preferring the two narrowest ranges is that they exclude the smallest observation in each year and thus give less weight to the CMIR data, which generally appear to underestimate net currency flows abroad and produce the smallest flow measure in nearly three-fourths of the periods.
Alternatively, because the CMIR flows are most often at the bottom of the range of estimates, one could diminish their influence by constructing a confidence interval ranging from the next to the smallest flow to the largest flow in any period; such a range would cover the true median about 93 percent of the time.
In the CMIR system, double counting may exist for some transactions; for example, a bank and a commercial shipper may both report the same currency shipment.
For example, the CMIR data imply that, taking the midpoint of the range of estimates, 17 percent of currency was held abroad at the end of 1995; but in that case, the implied amount overseas at the beginning of the sample (the end of 1976) would have been 67 percent.
Taking the median of the six methods excluding the CMIR method would increase the midpoint estimate of the amount held abroad slightly, from 55 percent to 57 percent.
Although banks and other firms are accustomed to filing CMIRs and probably do so fairly diligently, individuals are potentially less aware of these reports, less willing to file them, or even eager to avoid them.
Fourth, the record-keeping system for CMIRs was designed with the purpose of identifying individual transactions, not of developing accurate aggregate statistics on currency flows.
Figure 1 reveals that in the years prior to 1995, the CMIR estimates [CTI *] (adjusted for direct Federal Reserve receipts from abroad that are not required to be reported on CMIR forms) recorded a higher amount of US currency inflows to the US than the official proxy series.
The establishment of the cost efficient ECI system, however, appears to have introduced a downward reporting bias in the CMIR estimates of currency held abroad, requiring a change in CMIR reporting practices to include specifically shipments and receipts of currency dispersed and collected from the FED's ECI facilities overseas.
As such, the establishment of the ECI and the oversight of not accommodating the CMIR reporting requirement to the new institutional setting results in an understatement of outflow and inflow data in the CMIR reports for the year 1996 and beyond.