III Quality of Data
- Marcello Caiola
- Published Date:
- August 1995
In using a country’s economic and financial data, whether to analyze economic developments in the country or to prepare a stabilization program, one must be aware of the quality and especially the deficiencies of the statistics at one’s disposal.
In general, monetary accounts are derived from fairly reliable sources—banks’ balance sheets—even though they sometimes need adjustments and explanations. For instance, information on the exchange rates used to report operations denominated in foreign currency may be needed, particularly for countries that have floating exchange rates or crawling peg systems, in which case information for those accounts may have to be requested in both local and foreign currency. Also, unclassified assets and liabilities may include advances to the government that are still pending legalization, or certain other credit operations that should be classified elsewhere. Similarly, trust deposits and credit extended from these resources may be shown as contingency accounts outside the balance sheet of the banks.
The quality of fiscal statistics varies considerably. Details on central government operations are usually available, but their coverage could be limited, inasmuch as it may exclude operations that have been financed entirely by foreign loans, certain unpaid bills (for example, unpaid interest to the central bank), or operations that for political, legal, or other reasons are excluded from the “regular” central government budget. Statistics on the operations of state enterprises and other government agencies are at times sketchy and unreliable. Enterprises may provide profit and loss statements rather than cash flows, and in addition, there may be only limited access to original sources or to officers who can clarify certain discrepancies in other sources. In many developing countries, financial data on the rest of public sector operations—excluding the central government budgetary operations—are either not available or very unreliable.
Balance of payments statistics may be derived from a variety of sources, such as customs returns, external debt tables, monetary accounts, or even rough estimates based, for example, on tourist receipts and expenditures. Balance of payments data always add up to zero because of the net errors and omissions entry, but, as will be seen in more detail later on, these errors may reflect misclassified accounts as well as contra-entries to faulty data included elsewhere in the balance of payments.
National accounts data in developing countries are usually very unreliable, particularly in regard to classification by expenditures. After all, national accounts present data on government consumption and investment even in those countries where the authorities admit that such data are not available. Investment is usually estimated on the basis of imports of capital goods, supplemented by certain other information derived from the government budget; exports and imports do not always reconcile with similar figures included in the balance of payments; and government outlays are often based on budget, rather than on actual data. However, trends in GDP (whether real or nominal) are, in many cases, reasonably reliable, even though the actual level may represent very rough estimates.
Consumer prices may be based on an obsolete basket of goods and services, or their coverage may be limited to certain sectors of the population. Also, in countries with tight (or unrealistic) price controls, the consumer price index may reflect only changes in official prices, which are not necessarily representative of the inflationary trends in the country.
The best way to decide which data are acceptable and which should be adjusted is to look at the consistency between statistics. If, for example, government statistics show a very small deficit and central bank data show a large credit expansion to the government, one can conclude that fiscal data are incomplete. And, if money and quasi-money are increasing at a fast rate in nominal terms, while the consumer price index shows only a modest increase, this could indicate that the quality of the latter may be inadequate. To conclude, the first step should be to decide which data are more reliable and should form the basis for analysis. Other data would then have to be adjusted to conform or to be consistent with the “reliable” data.
An important stage in the process of selecting data is to ensure that the coverage of data is consistent. In this context, two likely sources of discrepancy among data should be emphasized: accrual versus cash, and timing and coverage differences.
While all revenues and expenditures are passing through an accrual stage, not all will reach the cash stage. Therefore, the coverage of an accrual presentation is by far larger than that of a cash presentation. In addition, a transaction could be recorded in two different time periods and classified differently for accrual and for cash purposes. For example, if the central government delays until 1993 the payment of interest due on its external debt in 1992 (and assuming that in 1993 the central government requires a credit from the central bank to make such a payment), the recording on an accrual and on a cash basis would be as follows:
|Balance of payments|
|Changes in arrears||100||−100||—||—|
|Changes in net international reserves (increase −)||—||100||—||100|
|Changes in arrears||100||−100||—||—|
|Central bank credit||—||100||—||100|
As shown in this example, discrepancies and errors could arise if the balance of payments were on an accrual basis while the fiscal accounts were on a cash basis, and no allowances were made for such a difference.
A common error could arise when comparing national accounts and balance of payments statistics, both of which are usually on an accrual basis, with cash operations of the general government. Since domestic savings are often estimated as the difference between investment and the current account of the balance of payments, government savings could be over-estimated—and private savings underestimated—unless fiscal figures are adjusted.
Timing differences can explain discrepancies between figures derived from different sources; at times, because of seasonal factors, these may be substantial. For example, changes in government bank deposits, as reported by the treasury, may differ from changes derived from banking statistics because of checks issued and not yet cashed. Differences between treasury records and bank statistics are usually a source of error and misclassification. As bank statistics are generally more reliable than fiscal data, the government domestic financing is usually adjusted to agree with data derived from monetary accounts. The differences between the figures may include timing discrepancies as well as transactions that were not reported by the treasury. While the former should be recorded below the line, because they do not represent unrecorded operations, the latter may call for adjustments above the line, depending on the nature of the transaction. The interbank float is another form of timing discrepancy between the records of banks.
Whenever data are modified to account for differences in coverage, particular attention should be paid to the recording of the contra-entry to the adjustment. For example, if disbursements of foreign loans to the central government are derived from the balance of payments rather than from fiscal sources, the higher or lower use of these loans should be reflected in government investment or other outlays, and not as an adjustment below the line.
At times, however, large discrepancies may be good indicators that figures are deficient. For example, a large discrepancy between bank credit to the public sector, as shown by the central bank and by the treasury, is not necessarily due to a timing discrepancy but could reflect transactions that had not been reported by the treasury, such as unpaid interest on outstanding central bank credit or extrabudgetary operations that have not been reported by the bank to the treasury.