In the case of a small individual business, it is unlikely that an expenditure of funds, an input or output, or an income from production will be counted twice. If it happens, that's usually just bad accounting (a math error), or else a case of fraud.
But things are more complicated when we aggregate the accounts of many enterprises, households and government agencies ("institutional units" or transactors in social accounting language). Here, a conceptual problem arises.
The basic reason is that the income of one institutional unit is the expenditure of another, and the input of one institutional unit is the output of another.
If therefore we want to measure the total value-added by all institutional units, we need to devise a consistent system for grossing and netting the incomes and outlays of all units. Lacking such a system, we would end up double-counting incomes and expenditures of interacting units, exaggerating the quantity of value-added or investments.
The system of gross and netting actually used, is ultimately based on a value theory, which specifies what may generally count as:
In other words, we cannot relate, group and aggregate prices in different ways without making some value-based assumptions that enable valid comparisons. Without those value assumptions, the aggregates themselves would be meaningless. Thus, when economists focus on market-prices, value assumptions are always in the back of their mind, even if they are not aware of that, and regard value theory as metaphysical.
Neo-classical economics rejects any value theory other than subjective marginal utility preferences, but social accountants who provide the empirical data for their economic science cannot regard value as simply subjective. Otherwise, anything can count as anything, according to subjective preference, and any old computation is permissible.
Once the principles of the value theory are established, categories and counting units can be exactly and logically defined, as a basis for mathematical operations to aggregate the flows of incomes and expenditures. All flows can then be allocated to their appropriate category, without counting the same flow several times.
In fact, the value theory applied in national accounts is nowadays strongly influenced by the valuation principles of ordinary business accounts and the prevailing social relations governing economic exchange, often fixed by law. Thus, for example, it is argued that no new value can result from a unilateral transfer of funds, i.e. where funds are provided without anything being provided in return.
The implicit assumption made in national accounts, is that the account at the macro-level must be similar to that at the micro-level. Economic relations are regarded as broadly the same at the micro-level and the macro-level. An individual business buys and uses up inputs and produces outputs for sale; it has costs and revenues. Thus, in social accounting all transactors are treated in a similar way ("as if" they were a business). The accounts can be criticised for being eclectic in some ways, but that is not necessarily a problem; the aim of the exercise is to identify and categorise all flows, and the user can then reaggregate them in different ways.
However, even if a consistent system of accounting rules is devised that conceptually eliminates double-counting, double-counting may technically still occur to some extent.
The first and most obvious reason is that, in actual accounting practice, boundary problems arise, because a flow of expenditures might be interpreted in different ways, from an accounting point of view. Sometimes, it will not be altogether clear which category a flow of expenditure belongs to exactly, it may not "fit" exactly into a category, or, it is technically impossible to separate out different flows in financial data, in such a way that is required by the social accounting system. This may mean that a flow is, in part or as a whole, inadvertently counted twice, because of difficulties with the data sources.
We might, for example, be easily able to identify an expenditure, yet this expenditure may not tally with the corresponding income that should exist, insofar as we can identify it (or vice versa). In that case, we have to make some assumptions or imputations based on what we do know, or can observe. Yet, some statistical discrepancies may remain.
Another reason has to do with the complexities of trade, in particular trade in services and international trade. Not only can it be difficult to correctly identify, survey and allocate particular financial incomes and expenditures, but also revaluations of assets occur, creating problems of how to value goods and services as such.
At the highest level, due to the expansion of foreign trade, a fraction of local value-added may consist of the local inflation of foreign-produced value-added, simply because imported foreign products are resold locally, at inflated prices, without any corresponding additional local production occurring. This may not necessarily create problems of double-counting locally, but if we want to estimate world GDP, we may face double-counting problems of some kind.
Interestingly, while world Gross Domestic Product (GDP) and world Gross National Product (GNP) are conceptually identical values for social accountants, the World Bank valuation of each of these aggregates differs by about one trillion US dollars, and the difference grows year by year.
The reason here however is not a double-counting error, but that different valuations of national currencies are used. Thus, the World Bank applies "ppp" (purchasing power parity) valuations to GDP, but an "Atlas Method" to estimate GNI. The argument is apparently that if world GDP is treated as an income, it will shrink (the World Bank cites 2004 world GNI of $39.8 trillion and a world GDP of $40.9 trillion, a discrepancy of $1.1 trillion).
One result of these different valuation methods, critics point out, is that it becomes impossible to compare GDP and GNI internationally with respect to the net international transfer of factor-income, which is excluded from national GDP, but included in national GNI.