Same to the decimal, twice: large-project imports in Banco de Moçambique’s August data

August 31, 2026

Wordless fine-line illustration: A customs warehouse where two identical stacks of crates, one labelled for last year and one for this year, stand side by side while an inspector with a clipboard compares the stencilled lot numbers

Two quarters of 2026 repeat the matching 2025 quarters exactly, and a revised investment figure and new number formats raise the audit stakes.

Large-project imports of goods in the first two quarters of 2026 are printed as US$693.7m and US$320.4m in Banco de Moçambique’s August 2026 statistical summary, posted on 4 September. The first quarter and second quarter of 2025 carry the same two numbers, to the first decimal. A repeated figure in a series of quarterly flows is possible but improbable, and twice running is more improbable still.

The table does not say whether the 2026 values are measured, carried forward from last year or provisional placeholders. The answer matters because the figure splits total imports into two blocks, and the block that excludes large projects feeds the Bank’s second reserve-cover ratio and carries all of the year-on-year growth the table reports.

Two quarters, two matches

The residual absorbs everything the project line does not.

Total goods imports were US$1,941.7m in the first quarter of 2026 and US$2,214.1m in the second, up 14%. Large projects account for US$693.7m and US$320.4m, which leaves US$1,248.0m and US$1,893.7m for everything else. Because the project line is constant against 2025 while total imports rise, the whole of the second-quarter increase appears in the non-project block: 51.7% quarter on quarter, and 14.4% above the second quarter of 2025’s US$1,654.8m.

O Económico reported the headline on 7 September: a goods deficit that widened from US$88.7m to US$319.6m, a fuel bill up 79.4% from US$236.4m to US$424.2m, and exports up 2.2%. Those numbers are arithmetic from the pack, and they stand. The split between projects and the rest is the part that rests on a figure the Bank has not explained.

For a firm, the size of the uncertainty is easy to state. If the true second-quarter project imports were US$150m higher than the repeated US$320.4m, the non-project block would be US$150m lower, and the 14.4% year-on-year growth in everyday imports would fall to about 5%. That is the difference between a consumer-demand story and a project-timing story, and it changes how a distributor should read the second-quarter surge.

Mechanism: why the project line matters

Large projects are a separately reported block, so a stale entry moves ratios built on both sides of the split.

The pack flags large projects as a memo item under both exports and imports, and does not define the list or its source in the pages themselves. Large construction and energy projects import heavily and unevenly, in lumps tied to equipment deliveries. If the reporting chain for these firms runs behind the main customs-based compilation, a prudent compiler might carry the last known figure forward until the return arrives. That is a plausible mechanism, offered as a hypothesis; it would explain both the repeat and the fact that the exports memo item for large projects, at US$1,471.0m and US$1,473.4m, does not repeat.

The consequence travels beyond the trade table. The July reserves footnote gives cover of 2.8 months on total imports and 4.3 excluding large projects, against 2.2 and 4.5 a month earlier. The total basis rises while the non-project basis falls, which is consistent with a smaller projected project block in the denominator. A user who does not know the project line is under review will read that as an improvement in reserve adequacy.

One quarter of investment, two numbers

Revised first-quarter investment sits in the pack, while press reports kept quoting the older figure.

Foreign direct investment for the first quarter of 2026 was printed as US$1,286.0m in the earlier editions and in the Bank’s Boletim Estatístico for May, where it appears with the financial-account sign convention as minus 1,286.0. The August edition shows US$1,328.43m, with US$589.95m for the second quarter. Against US$1,626.3m in the first quarter of 2025, the revised figure is a fall of 18.3% where the original was 20.9%. Four days after the August edition, Club of Mozambique, reporting the Bank’s statistical data, put the first-quarter figure at US$1.286bn, down 21%.

Neither number is wrong. They are different vintages, and a reader needs to know which is current. O Económico used the revised base when it reported that second-quarter investment fell 55.6% to US$590m.

Formats that confuse parsers

The August edition changed how numbers are printed in some tables and not in others.

In the investment tables, zeros that were printed as 0.0 are now dashes, and totals have gained a second decimal, as in 2,509.40 and 3,552.74, where earlier editions printed 2,509.4 and 3,552.7. The second-quarter header reads ‘II Trim20 26’. In the reserves table, the opening and closing balances of 3.463.80 and 3.436.80 use full stops as thousands separators, in a pack where the monetary tables use commas. A parser expecting one convention reads the first as 3.46, or fails.

None of these is a substantive change. All of them are the kind of change that produces silent errors in an automated workflow, and each can be caught by a rule that rejects any number whose format differs from its neighbours.

A rule that works in practice is to store every figure as text alongside its parsed value, and to quarantine any file in which the count of distinct number formats changes from one edition to the next.

Audit actions for this edition

Query the project lines, version the investment series and reject unexpected formats.

Put three questions to the Departamento de Estatística e Reporte: whether the 2026 project-import figures are measured or carried forward, which edition holds the governing first-quarter investment figure, and whether the cover ratios will be restated if the project base changes. Until the answers arrive, run reserve and import analysis on total imports and on the Bank’s reserve stock, and treat any ex-project cover ratio as provisional.

Lenders sizing foreign-currency import lines should assume that the non-project growth in the second quarter is smaller than printed if project imports turn out to be higher than the repeated figure, and should hold the Bank’s next release against that scenario before adjusting limits.