State Budget Laws (LOE)

LOE is a tool that financially quantifies the resources needed to achieve the policy goals defined in the PES. The LOE predicts revenues and expenditures, and consequently serves as an economic forecast for the state’s revenue and spending over the period of a year. Planned expenditures are presented according to administrative and economic classifications (more on these in Section 3). The Ministry of Planning and Development (MPD) writes the PES, and based on that, MEF formulates the annual proposed LOE. Once the budget is published, ministries and agencies are authorised to spend money, consistent with the legal appropriations for each agencies and the categories as per the economic classification.

Budget Execution Reports (REO)

The REO is a government document that aims to review on a periodic basis the implementation of public expenditure within the period in reference. It is issued quarterly by MEF’s National Directorate of Public Accounts (Direcção Nacional de Contabilidade Pública, or DNCP). The report presents data on the execution of revenues and expenses for all public sector institutions. One of the main responsibilities of MEF is to monitor and manage in-year changes to the budget. Unforeseen circumstances or poor budgeting may make it necessary to adjust the budget (Simson, Sharma & Aziz 2011). In the course of the fiscal year, ministries and government agencies can request reprogramming to receive additional funds if a need arises. To accomplish such reprogramming requests, usually the National Directorate of Budget (DNO) has to reallocate funds among different ministries and agencies, with an eye to preserving macroeconomic stability and fiscal responsibility principles. Consequently, the REO compares the original budget allocations (LOE) with the current in-year budget allocations. REO is a snapshot that shows the situation at the time it is issued. It is worth noting that the REO could contain data that is at the time only partially processed by the electronic system e-SISTAFE, and thus could represent only partial budget execution. Also, some units are processed in e-SISTAFE before being incorporated into an expenditure item in the REO.

General State Accounts (CGE)

The CGE, also issued by MEF’s DNCP, is the government’s annual consolidated financial statement, compiled after the end of the fiscal year. This data report highlights budget implementation and financial management, containing collected revenue, paid expenses, and tables of budget execution, and these data are juxtaposed against the original budget allocations also recaptured in the CGE. The CGE, after being issued, is sent to parliament and the auditor-general. The parliament forwards these accounts to the auditor-general for external audit together with parliament’s suggestions on areas requiring attention. The auditor-general submits its report on the CGE financial statements back to parliament, following which the CGE is presented to the parliament’s plenary, where it usually is not discussed until March, i.e. about 15 months after the budget year, due to lack of capacity within parliament to analyse this report and the budget proposal for the upcoming year at the same time.

A quantitative elucidation of the relationship between LOE, REO, and CGE

We compile and analyse aggregated data from 2010 to 2013 in order to provide a quantitative elucidation of the relationship between the data across the three core reports, LOE, REO and CGE. This analysis discusses the differences among the reports, as well as factors that may explain those differences. It is important to keep in mind that differences in the numbers between the three documents exist because each report corresponds to a different stage along the budget cycle. The aggregated data for the years 2010 to 2013 with regard to the LOE, REO, and CGE are shown in the upper panel of Table 1. The LOE reports the approved predicted expenditure for each respective year (line [1]). The REO accounts for the quarterly executed budget as well as the budget reallocations. The budget reallocation constitutes the updated spending ceiling, or final budget allocation, within the period in reference (line [3]). The CGE accounts for all the budget reallocations and all the transactions, including those that within the budget year were registered neither on e-SISTAFE nor in the REO (line [6]).

Several factors explain the discrepancy between budgeted and actual values, which result in an aggregate execution rate reflecting under- or overspending, as seen in the last line of table’s upper panel. In addition to the earlier discussed reallocations, deviations from the planned disbursement of funds by donors, shortfalls or unanticipated excess in government revenue collection, and availability of supplementary funds, among others, can result in an execution rate above or below 100%. The initial allocation is first reported in the LOE [1], and repeated in the REO [2] and CGE [5] data reports. The reallocated budgets [3] and [6] did not result in any changes in some of the years, specifically in 2010 and 2012. However, in 2011, the budget was revised upwards once, and in 2013 twice (the second time reflected in the CGE), mainly due to an increase in revenues driven by foreign aid and international loans, and because the CGE report accounted for the previous year’s ending balance.

As is apparent from lines [5] to [7], the CGE contains all the relevant information of the whole budget cycle. However, that only applies to public expenditures in the aggregate. If there is interest in analysing data on a specific sector or ministry along the full budget cycle, CGE and REO do not provide such a comprehensive snapshot because they do not contain the initial budget allocations by sector, so it becomes necessary for the analyst to resort to all three reports. The bottom panel of Table 1 presents data for spending on agriculture and fisheries⁴ from 2010 to 2013 available in the three data reports.

Contrary to the relatively minor differences in public expenditure along the budget cycle in the aggregate, public expenditure on agriculture varies quite significantly among the three reports. In particular, pronounced upward as well as downward adjustments are made to the budget along the budget cycle, from initial budget, to revised budgets, to final realised expenditures. Especially notable is that the execution rate of the original budget is strikingly low, ranging from about 62% to 76%, compared to the much higher execution rate of the aggregate budget. Indeed, the Public Expenditure and Financial Accountability (PEFA) evaluations have pointed out that the absence of clear and strict rules regarding budget reallocations has reduced the credibility of the institutional budgets (Lawson et al. 2016).

Administrative classification

The administrative classification links budget allocation and spending to a governmental organisation, such as a ministry, directorate, or other cost centre, thus allowing the data user to identify agencies’ responsibilities in public expenditure management. In Mozambique, the codes for the administrative classification are detailed up to nine digits. However, most of the reports use only up to the six-digit code level. Under the two-digit code level, ministry of agriculture is coded as 35, while the ministry of fisheries is coded as 37. Appendix Table A.1 presents for an overview this classification system at the two-digit code level for the whole public sector, while Table 3 provides the basic structure of this classification system up to the four-digit level (in terms of the coding system in operation through 2012) for the ministries of agriculture and fisheries. Each four digit code pertains to a distinct agency. For example, 3501 references the ‘mother’ ministry of agriculture, while 3503 pertains to the food security secretariat, 3505 the institute of irrigation, etc. These two are separate agencies, although subordinate to the mother ministry.

Until the 2012 LOE, the administrative classification up to the six-digit level linked the administrative classifier to the territorial classifiers (for an overview of the latter, see Appendix Table A.2). In other words, just adding the territorial two-digit suffix to any government agency leads to the six-digit administrative classifier. For example, the ministry of agriculture’s Provincial Directorate of Agriculture in Nampula province used to be coded as 352103, where 35 stands for the ministry of agriculture and its subordinate institutions, 21 for Provincial Directorates, and 03 for Nampula province. As of the year 2013, the administrative classification has changed, by replacing the territorial suffix with a letter after the ministry level (see Table 3, column [3]). Consequently, the 2013 code for the ministry of agriculture’s Provincial Directorate of Agriculture in Nampula province is 35D 00 0141, where D represents Nampula province and 0141 represents the primary agency for agriculture (which would be the ministry at the central level, and the directorate at the provincial level).

It is apparent that the administrative coding of public expenditures, both before and after 2013, is able to identify the distinct agency and the government tier and jurisdiction where it operates. It does not, however, identify units within the agencies. For example, the mother ministry of agriculture is subdivided into directorates, such as the directorate for veterinary services, directorate for agricultural extension, etc. The administrative classification does not provide unique codes to these directorates. This is an important shortcoming, in light of the value of obtaining public expenditure information by the different activities the various directorates of a ministry undertake.

Table 4 shows the levels of public expenditures for agriculture based on the administrative classification, and using data from the three main reports. Columns [2] and [4] account for the budget reallocations in each report, while [3] and [5] account for the executed spending reported.

As discussed in Section 2, there is relatively weak control over agricultural budgeting, resulting in realised expenditures amounting to just slightly more than 60% of the initial budget allocation. This mean aggregates a wide distribution of budget execution, with less than a third of the Agricultural Development Fund budget spent, the agricultural programme with the single largest size in the original budget, and on the other hand the mother ministry of fisheries spending 15% above the original allocation. The table shows that there are nontrivial differences in the budget figures along the budget cycle, not only between initial allocation and final expenditures, but also between the reallocated budgets in the REO versus CGE (columns [2] vs. [4]), and between realised/executed spending in the REO vs. CGE ([3] vs. [5]). Thus, analysts need to be well aware of how the figures in the columns represent different stages in the budget cycle—an issue ignored in a good deal of public expenditure analysis.

Economic classification

The economic classification of expenditures links the budget to economic categories, such as expenditures for goods and services or capital expenditures. The economic classifier of expenditures in Mozambique aims to identify the nature of the expenditure, according to the following five levels: The first level indicates the economic categories of current and capital expenditures (see Table 5, which presents the economic classification up to the three-digit code level). The second level indicates the aggregate group of expenditure, such as personnel costs (code 11 in Table 5) or goods and services (code 12). The third, fourth, and fifth levels indicate further economic breakdown of expenditure. For instance, the retirement expenditure category is coded as 143101, where the first level (code 1) is current expenditure, second is current transfers (code 14), third is current transfers to households (code 143), fourth is civil servants’ pensions (code 1431), and finally the fifth level is retirement spending for civil servants, for a complete code of 143101. The formal economic classification system for Mozambique provides a breakdown to the level of a six-digit code (five levels).

It is important to highlight that the economic classification of expenditure, as shown in Table 5, is not to be confused with another categorisation widely used in Mozambique, which distinguishes between operational (funcionamento in Portuguese) and investment (investimento) expenditures. The ‘investment’ category should not be assumed to represent expenditures contributing to public capital formation, or capital expenditures.⁵ This ‘operational/investment’ categorisation is a long-standing approach in Mozambique and is used in all the data reports (that is, the LOE, the REO, and the CGE). It was agreed on with the International Monetary Fund and the World Bank in order to differentiate in the state budget the expenditures funded by donors, at the time typically associated with the ‘investment’ category, from those financed domestically, associated with the ‘operational’ category. However, the definition of this categorisation has been diluted over time: the ‘investment’ category began to include domestic funding sources, given government counterpart funding in donor-supported projects.

Table 6 shows the agriculture and fisheries public expenditures in 2013 by economic classification as well as by operational/investment categorisation, comparing LOE and CGE reports. While the public finance data reports disaggregated ‘operational’ expenditures by economic classification, the same is not done for the (far larger in size) ‘investment’ category. The fact that, as the table shows, it is the investment category that contributes the most to a poor budget execution rate strengthens the argument that this category would need to be further disaggregated by economic classification, in order to gain meaningful insights into the reasons for the large degree of underspending, so that the DNO can exercise more effective monitoring and spending control.

Functional classification

The functional classification (Classification of the Functions of Government, or COFOG) has shown to be of relevance to a broad variety of analyses (IMF 2014). It classifies expenditures according to their socioeconomic purposes. In Mozambique, the internationally defined COFOG system was officially adopted in 2005, although the functional presentation of the budget is not formally approved by parliament and therefore is not legally binding.

The distinction between economic and functional classifications can be summarised as the difference between what expenditures are spent on versus what expenditures are for. The economic classification described above details the particular items on which resources are expended—that is, expenditures on salaries, goods and services, capital formation, and so on—while COFOG organises public expenditures in such a way as to provide information about the purposes toward which these expenditures are undertaken, that is, expenditures to provide agricultural services, health-care, road infrastructure, and so forth. The functional classification often is close, but rarely equivalent, to the administrative structure of government. This is because public expenditures in an agency, such as the President’s office, may pertain to a sector such as agriculture. In principle, then, the functional classification would gather public expenditure data on agriculture (as one example of a function), irrespective of which agency is responsible for the budget item.

Appendix Table A.3 outlines the internationally established functional classification of public expenditures according to COFOG and includes subcategories below the first COFOG level only for functions related to agriculture expenditure: those that either constitute or encompass agriculture-related functions. There are 10 broad categories, which are at COFOG level 1. Several familiar functions, such as health (COFOG code 07) or education (COFOG code 09) are level 1 categories. Agriculture and related functions are collectively treated as a category at the second and lower levels. Specifically, agriculture is found below the level 1 category of ‘economic affairs’ (coded as 04). The main agriculture-related category (coded as 042) groups together crop and livestock agriculture, forestry, and fishing and hunting. Another level 2 subcategory under economic affairs is R&D, with code 048; under that, at level 3, is agricultural R&D (code 0482). For an easy overview of both the levels and the functions of the categories related to agriculture, level 1 categories are shaded in Appendix Table A.3, level 2 categories are unshaded, and level 3 categories are italicised. The last column of the table summarises the relation to agriculture functions.

Mozambique’s organisation of expenditure data by functional classification is broadly consistent with the Government Finance Statistics Manual 2014 (IMF 2014), except for minor variations at the third and fourth levels of the classification. For example, this classification scheme includes agriculture-related function codes such as 04213 ‘Prices and Agricultural Productivity’ or 04214 ‘Rural Extension.’ The functional classification and coding used in Mozambique up to level 4 is shown in Table 7. In Mozambique, expenditure by functional categorisation down to level 2 (which identifies agriculture) is reported in the REOs and CGEs, but not in the LOE. Table 7 summarises the functional classification of realised agricultural expenditures obtained from the CGEs for the years 2010 to 2013. A review of this table shows that roughly 60 percent of the total agricultural expenditure is aggregated in an ‘Other’ type of category, namely ‘Agriculture, Forestry, Fishing, and Hunting NEC’, where NEC stands for ‘not elsewhere classified’ (code 04291). This means that the existing de facto functional breakdown does not provide the meaningful insights it was intended to have.

Programmatic classification

The programmatic classification organises expenditures according to programmes that are designed to meet particular government policy objectives. A programmatic classification is intended to be supportive of the priority setting of the budget by being based on outputs (OECD 2007). Mozambique’s programmatic classification of strategic planning has the following hierarchical structure: (i) at the highest stage in this classification hierarchy is the Central Objective for the Government Five-Year Plan (Objetivo Central do Programa Quinquenal do Governo, or OCPQG), (ii) next below that are Strategic Areas (Áreas Estratégicas, or AE), then (iii) Strategic sub-areas (Subárea Estratégico, or SAE), (iv) Programmes of Government (Programas de Governo, or PG), (v) Subprogrammes of Government (Subprogramas de Governo, or SPG), (vi) and finally at the bottom of the hierarchy, Budget Actions.

The Central Objective for the Government Five-Year Plan (OCPQG) classifier aims to show the organisation of the strategic macro objectives of the government, which are defined in the PQG. This classifier captures six objectives in three-letter codes (BGD, CPD, DEC, DHS, RCI, and RDS) the descriptions of which are shown in Table 8. If a programme does not fit into one of these six defined OCPQG objectives, additional codes are used for such programmes. The Strategic Area (AE) classifier organises categories under each OCPQG objective, each Strategic Area (AE) is further broken down into Strategic Sub-Areas (SAE), and so forth (Table 8 shows codes and descriptions only up to the AE level).

A programmatic classification is rather complex to define, and even more so to apply. Because of this, this classification is not applied with as high a degree of comprehensiveness as the concept of functions of government could in principle be applied (the de facto application of the functional categories and its shortcomings has been discussed above). When a functional classification is applied properly, all of the activities of government are categorised inside one function or another, in a mutually exclusive and exhaustive fashion. The programmatic classifiers, on the other hand, refer only to certain types of public expenditure that are of particular interest to government strategy, and thus do not exhaustively capture all public expenditure related to a specific function.

According to the actually applied programmatic classification in Mozambique, in 2013, the only clearly traceable strategic area (AE) linked to the agricultural sector is that of ‘Agriculture, Livestock, Forestry, and Wildlife,’ with allocations to both central and provincial levels for a total of 5,436 million meticais, as shown in Table 9 (see code DEC-AGR). Expenditure data classified using these programmatic classification codes as shown in Table 9 are only available in the LOE. Expenditure information in the REO and CGE documents do not use this programmatic classification system anywhere, making it impossible to trace actual expenditures by these programmatic codes, or to determine budget execution rates based on these codes.

Priority-sector classification

While the programmatic classifier is not present in the REO and CGE, a similar-in-spirit classifier is reported in the REO and CGE documents. This is an organisation of expenditures that are considered ‘priority sectors’ according to poverty reduction strategies of the government. The priority sectors according to which expenditures are reported in, for example, the CGE 2013 are: Education; Health; Infrastructure; Millennium Challenge Account; Agricultural and Rural Development; Governance, Security and Judicial System; and Other Priority Sectors. In this case, the priority programmes do not have designated codes, but only titles (e.g. Education, etc. as denoted above). Then, under each such title/sector are organised public expenditures by administrative classification, with the corresponding administrative codes.

For example, the agriculture and rural development priority-sector expenditure is meant to correspond to the sum of spending related to the agricultural sector and rural development undertaken within all institutions, not only in the ministries of agriculture and fisheries. Table 10 presents the expenditures with their corresponding administrative codes that are considered to fall under the ‘agriculture and rural development’ priority-sector.

Within the administrative classifiers categorised under the Office of the President, there is spending by government through local development funds by the district administrations), in which agriculture related spending may be significant (see Table 10). At the district level, institutions are not represented using the traditional administrative classifiers found at central or provincial levels. Instead, at this local level the key agencies are referred to as ‘services’, and one service may cater to more than one sector. For example, agricultural activities are undertaken both under District Services for Economic Activities (SDAE) as well as under District Services for Planning and Infrastructure (SDPI). Both services also undertake non-agricultural activities. This, of course, can make it difficult to know how much goes just for agriculture at the district level, given that district expenditure data are reported primarily at the level of the services.

The priority-sector classification system, as it is used in Mozambique, is fairly nonformal in its nature. Despite changes in the accounting rules over the past few years, this classification system still does not provide a reliable instrument to comprehensibly aggregate agricultural expenditure. For example, since 2012, agricultural expenditure by priority sector classification has shown a greater than three-fold increase (see Table 10). Indeed, district administrations (code 0122), through the district development fund, made an extreme jump to more than 139-fold from 2011 to 2012. Both the increase of agricultural spending from one year to the next by over 250%, driven by the increase of one component by about 13,000%, raises serious questions about whether this is a genuine increase in spending of given categories, or a reclassification to boost the reported aggregate agricultural spending figures.

Promises and pitfalls of using public expenditure data along Mozambique’s classification systems

Mozambique’s public financial management system has improved substantially since the 1990s, with the introduction of e-SISTAFE. One of the outcomes of reform efforts is the abundant availability of highly detailed public finance data to any Mozambican analyst, through free online access of data reports over multiple years. But the existence of so many classifiers reduces the public expenditure data reports’ usability. First, there are a number of classifiers within Mozambique’s accounting system that are not in fact actively used in the official reports (see Table 2). Second, even some of those classifiers on the basis of which official reports present spending data are only very partially applied—for example, only for budgets but not realised expenditures (or vice versa), or only to a subsection of the overall budget. Consequently, in order to precisely define the expenditure of the ministry of agriculture and its agencies by administrative and economic classification, one must reconstruct public expenditures across hundreds of report pages. This renders the task of quantifying agricultural public expenditures unwieldy for analysts who are not technical staff in MEF or MPD, or otherwise do not have the capacity and time to dedicate to developing a very intimate acquaintance with the detailed methods in the LOE, REO and CGE reports.

In particular, the functional classification of government expenditures would be of significant interest to a wide range of policy analysts. However, Mozambique agricultural expenditure organised along COFOG lines provides a breakdown of agricultural spending for less than half of what it considers to fall under this function. This may not be a barrier to those analysts who are solely interested in deriving the totality of agricultural spending. However, even for them it is not possible to verify what type of agricultural activities over 60% of these expenditures support. As a result, it is not optimal for use in an aggregation process for policy or analytical purposes.

There is greater disaggregation in the organisation of government public accounts along administrative lines, with expenditures grouped by the ministries and agencies undertaking the expenditures. Also, the administrative classification is applied fully along the budget cycle, from the initial budget allocation to realised expenditures in audited financial statements. Perhaps the most significant shortcoming of this classification in Mozambique’s data reports is that the breakdown stops at whole agencies, and does not continue to major units or divisions within agencies. To see why this may be a problem, the National Directorate for Promotion of Rural Development (DNDPR), which is part of the Ministry of State Administration, should be accounted for in agriculture, given its large focus on providing agricultural extension to farmers. However, since it is a directorate within a different ministry that does not otherwise focus on agriculture, there is no administrative classification that identifies DNDPR’s activities. So, adding a budgetary allocation code at the sub-agency level would allow one to clearly identify functionally differentiated actions within an agency, as well as to explicitly define transparent rules to aggregate expenditure. Consequently, strengthening the administrative classifier by further disaggregating it to directorates within agencies seems a reasonably simple and plausible way to enhance the quality and usability of the public expenditure data.

The economic classification, in use across all of Mozambique’s major public expenditure data reports, is helpful for gaining insights on how expenditures on functions or sectors break down by salaries, capital equipment, operational expenditures such as goods and services, etc. However, the active use in Mozambique of the ‘operational/investment’ classifier (funcionamento and investimento in Portuguese), in parallel to the more conventional economic classifier, functions as a source of high potential confusion for analysts, given that it seems close to the recurrent/capital distinction of the economic classification, but this is not so. If the operational/investment categorisation is to be retained, it should be reconceived as a category that can be directly mapped into the existing economic classification, to avoid the potential confusion. Furthermore, when examining the data underlying the operational and investment categories, the latter is currently the largest source of underspending in agriculture, which clearly impacts midterm planning. The proper mapping of the operational and investment categories into the economic classification may allow MEF and the legislative body to increase their control over budget execution by increasing clarity about the concrete sources of low budget execution, and thus reduce budget underspending.

In the absence of administrative coding of directorates within ministries, as discussed above, one source for identifying these additional agricultural expenditures is the so-called priority sectors classifier, which was detailed in Table 10, and includes, for example, spending by government through local development funds, where agriculture-related spending may be significant. However, caution is warranted in the process of extracting agricultural expenditure data undertaken by other ministries in this way, since doing so relies on the assumption that the government has properly identified such public expenditures as being properly associated with agriculture. Both the trends observed in Table 10, such as the more than 130-fold increase in the expenditures of district administrations claimed to be associated with agriculture from one year to the next, as well as the possibility that policy dialogue and external mandates to spend more on agriculture could create incentives to assign more budget items to the sector, are examples of why care is needed in relying on the priority-sector categorisation in correctly identifying agricultural public expenditures.

Figure 1 gathers and compares figures on total realised expenditures on agriculture obtained from CGE reports, calculated based on the functional, administrative, and priority-sector classification systems—as well as agricultural expenditure figures obtained from two major cross-country public expenditure databases, namely the International Monetary Fund’s Government Finance Statistics, and the International Food Policy Research Institute’s SPEED database (Statistics on Public Expenditures for Economic Development) (IFPRI 2015). The latter two sources did not have data available for the last two years, 2012 and 2013, at the time of this research. Overall, we see that the figures vary across the different classification systems used, in some cases wildly so. For example, while the lowest figures in 2010 and 2011 were 82% and 83%, respectively, of the highest figures—reflecting a nontrivial variation, in 2012 and 2013 the analogue shares were more dramatic, 30% and 35% respectively. The outsized estimates derive from the priority-sector classification, which jumped drastically from its two previous years. A more in-depth examination beyond the scope of this analysis would be warranted to determine whether this was a real increase on agricultural spending at the local government level, or a reclassification of previously existing spending that used to not be considered as agriculture, possibly in response to the imperative from the Africa-wide CAADP (Comprehensive Africa Agriculture Development Programme) process to spend more on agriculture.

It should be expected that the agriculture COFOG includes expenditures from agencies besides the ministries of agriculture and of fisheries. This is because expenditures on a particular function are often undertaken by government agencies that have a partial mandate for the particular function. Agriculture is a good example of the dispersion of functional spending across various agencies. For instance, the Ministry of Water Resources invests in irrigation for crop production, and the Ministry of Education funds agricultural training. Consequently, COFOG’s agricultural expenditure should be larger than the expenditure of the ministries of agriculture and fisheries combined. However, contrary to expectation, the functional-classification based figures are lower than the administrative-classification based figures. In fact, the figures based on the functional classification are the lowest across all measures.