FOREX Press I. J. of Business & Management Research
Support Open Access

Research Article |

Rwanda: Lessons from Applied Intergovernmental Fiscal Equalization Formula

Author(s) : Ludovick Leon Shirima

Publisher : FOREX Publication

Published : Article-in-Press

e-ISSN : 2347-4696

Page(s) : 101-109

This article examines the origins of the intergovernmental fiscal equalization formula proposed for Rwanda in 2003 and provides an econometric evaluation of it using 2003 data. Intergovernmental fiscal transfers were introduced as part of government reforms in the early 2000s to drive development through fiscal and financial decentralization 1 . On the one hand, the reforms aimed to improve service delivery and reduce ubiquitous poverty; on the other, they intended to strengthen national unity, promote reconciliation and power-sharing, and stimulate participation in decision-making processes that had been dismantled in the period following the 1994 Rwandan genocide.

As part of its reforms, the government introduced a mix of grants that were technically unsound, as discussed elsewhere in this article. These were conditional grants, unconditional subsidies, equalization or common share transfers, and delegated funds 2. Discussions on the effectiveness of these transfer variants are ubiquitious in the literature; for instance, Bird and Smart [3] offer a detailed description. Nevertheless, these were the precursor initiatives that underpinned the transfer system reforms of that period.

This paper is organized into eight sections: Section 2 narrates the approach used to carry out the study. Section 3 describes the data that was used in designing the formula, and their sources and study limitations, while Section 4 provides an assessment of the previous transfer system. Next, Section 5 answers the question of whether grant transfers to subnational governments were necessary for Rwanda at that time. Section 6 explains the new model in detail and its contribution to the literature. In Section 7, the designed system is econometrically evaluated to ascertain its relevancy. Finally, Section 8 presents both the conclusions and caveats.

ARC is a capital pooling vehicle regulated by the RBI for the acquisition of non-performing loans by the identified qualified institutional buyers (QIBs). AIF is regulated by the SEBI and can acquire only securities of equity or debt but not loans.

As the ARC is regulated by the RBI it can only acquire debt or it can only acquire the equity limited to the agreed debt to equity conversion. This is also in accordance with the agreement during the underlying loan acquisition by the ARC. This implies that the current regulatory framework isn't flexible enough to route the whole investment through a single pooled vehicle by a distressed fund to the investee company. As the SEBI regulates the securities market and the RBI regulates the banking system and its consequent stressed assets, these regulators don't incur the others defined responsibilities. However, for AIFs, it requires an enabling clause and for the ARCs an amendment is required to the SARFAESI Act, as the Act governs the ARCs to bring them under a single pooling vehicle. If the modifications or reforms discussed are implemented the institutional investors would get the flexibility on the capital structure of the investee company once the vehicle is chosen. However, investors need to remember that ARCs are only used for acquiring NPAs (or SMA2 defaults) and the SARFAESI Act cannot enforce the obligations for AIFs.

In the current scenario, even though, the IBC proceedings have been halted for a brief period, this may not help the large real estate projects in reality on the practical grounds other than postponing the immediate death crisis. What is required is to rejuvenate the dead projects with the infusion of funds. The funds would easily flow only when there is a flexible capital outlay structure as discussed above for the investors.

The large volumes of stakeholders present in the realty projects also require rapid changes to uplift it with suitable sops and required reforms. As a more agreed upon and flexible approach is required in the residential real estate projects to protect it from further getting into the status of limbo, a flexible regulatory framework in terms of the capital outlay needed to be sought by the regulators.

This paper describes the combination of approaches used by the study to accomplish its goals. Initially, relevant documents were gathered from ministries, departments, and agencies (MDAs), and other stakeholders who were responsible for implementing decentralization at that time. This was followed by a search for theoretical and empirical studies on intergovernmental fiscal transfers, and specifically those on the Rwandan context. The collected documents and literature were subjected to desk review in order to gain an intimate understanding of the pertinent issues.

Simultaneously, secondary actual fiscal3 and social data were collated and assembled from various sources into a portable data set4. The data were cleaned to minimize statistical errors and eliminate bias during the analysis. The formula was mathematically modeled and tested. Thereafter, simulations and analysis were carried out using Statistical Package for the Social Sciences (SPSS). Lastly, the findings were interpreted and presented schematically and in tabular form.

3.1 Data Related Issues

While designing the formula, we had to rely on secondary data from various sources that were collated into a central data set of fiscal decentralization data (FDD). The data were collected from two main sources—local governments’ (LGs) fiscal reporting systems and the central government’s data, mainly from MDAs.

The Ministry of Local Government, Information, and Social Affairs (MINALOC) provided data on the Local Authority Budget Support Fund (LABSF) and the other fiscal and financial aspects of LGs. Moreover, the Ministry of Economy and Financial Planning (MINECOFIN) supplied data on district welfare rates obtained from both, the Department of Statistics and the National Census Coordination Unit (NCCU).

Several other government agencies proved to be very supportive and provided the necessary data. First, the Rwanda Revenue Authority (RRA) contributed data on central government revenue collections, which acted as a backup for the district-level data, mainly in terms of ascertaining the revenue generated before and after the decentralization initiative. Secondly, the Central Bank of Rwanda (BNR) furnished data on the country’s macroeconomic parameters such as economic growth rate, the gross domestic product (GDP), and the inflation rate.

3.2 Study Limitations

It is worth mentioning that Rwanda, like any other developing country, faced huge problems with regard to data availability in 2003. The GoR had just started to implement fiscal decentralization, and at that time, it had not implemented systems for data collection to track the progress of its decentralization initiatives. Furthermore, the government’s MDAs lacked adequate resources and personnel for robust accounting and financial management, which are key to implementing fiscal decentralization. This, in turn, led to the fourth problem: it was impossible to design meaningful models to estimate the country’s future fiscal performance and position.

The following section examines the transfer system pre-2003 and explains the new transfer mechanism that required minimum data, was simple to implement, and was contextually relevant to Rwanda at the time.

Previously, unconditional transfers were allocated to districts and towns on the basis of four parameters: district population, equity common fund, square area, and revenue capacity5.

4.1 Inherent Weaknesses of the Previous System

There was a need to refine the applied proxy parameters as they possessed valid weaknesses. Firstly, the equity common fund proxy (the equalization portion), where each district claimed an equal share of the fund as a ‘balancing factor’, was unjust since it ignored the intended purpose of equalizing the inherent differences between the subnational jurisdictions, i.e., vertical fiscal imbalances. The implication was that this impacted other parameters, including district-specific attributes, which it was expected to address. The only rationale given for the use of this parameter was that each district had to start somewhere to provide basic public goods and services, which required financial resources. In addition, it treated all the towns and districts (urban and rural) equally, thus undermining its intended purposes. The above discussion echoes Mancur’s [8] views.

The parameter weights of the old system were determined subjectively and were highly inflated; as such, they were scaled down in the new model. Surprisingly, district revenue capacity was used, incorrectly, as a proxy for poverty (Table 6). This was unrealistic and inappropriate because it did not capture income distribution; more importantly, it did not account for variations in the population’s well-being. Next, the old transfer system omitted each district’s expenditure needs and its elements or types. Similarly, expenditure on education and health were (and still are) of paramount importance in a country like Rwanda, but were not accounted for in the old transfer system. Therefore, it is clear that this formula was fundamentally flawed, since it did not account for the fiscal gap between districts, which underpins a good transfer system.

Our statistical results showed that the allocation of the LABSF in 2002 and 2003, based on the described criteria of the previous system, led to inequity because of the system’s unreliability and unpredictability. This is demonstrated in Table 1, which shows that LABSF transfers were higher in 2002 compared to 2003, because in the latter period, districts and towns realized only three-fourths of their grant entitlement.

 

LABSF 2003

(RWF)

LABSF 2002

(RWF)

 Mean

8,644,885

14,260,726

 Maximum

13,503,396

22,274,520

 Minimum

3,405,604

5,618,011

 Standard deviation (Std)

2,279,404

3,760,084

 Observations

106

106


Table 1: LABSF actual allocation for the fiscal years 2003 and 2002 in Rwandese Francs (RWF).

The results from both fiscal periods indicate a wide disparity among the values of the entire descriptive statistics imputed. These results confirm that the LABSF transfers did not lead to equalization among subnational jurisdictions.

This section answers the question of whether intergovernmental fiscal transfers were required for subnational governments in Rwanda in 2003. This study makes its case by drawing upon earlier studies on the subject including Ehdaie [9] and Martinez-Vasquez [10]. The econometric evaluation used 2003 actual data, including information on all transfers to LGs, unless mentioned otherwise. Four key aspects of the study are examined below.

5.1 District Mean Revenues

To analyze the revenue capacities of town and districts, we used district mean revenues (MREV) calculated using districts’ own sources revenues (OSRs) for 2003. Using this data, the per capita revenue (PCMREV) was calculated. This showed that districts and towns had low revenue capacities and, therefore, they needed more stringent revenue enhancement strategies. This approach is an alternative to the Australian model, which was designed for a country where data unavailability was not a pressing problem [11]. The study also showed that districts and towns in the same provinces in Rwanda had heterogeneous tax bases that led to differential tax capacities. This gap was more pronounced when considering the urban–rural divide.

Table 2 shows that PCMREV was very low in 2003, at just 88.48 Rwandese Francs (RWF). This was further confirmed by the low maximum and minimum values of 742.63 and 5.27 RWF, respectively. These results implied that districts and towns in Rwanda struggled to generate adequate revenue to meet their expenditure needs, as they worked with relatively narrow tax base.

 

PCMREV (RFW)

MREV (RWF)

Mean

88.48

6,208,417

Maximum

742.63

76,520,613

Minimum

5.27

548,641.5

SD

118.97

9,693,552

Observations

106

106


Table 2: PCMREV and MREV comparison in Rwandese Francs (RWF).

This argument was supported by the large standard deviation of 118.97 RWF. Thus, revenue assignment, as shown in table 3, was inadequate—the central government should have empowered LGs with more revenue powers.

Decentralized taxes and other local taxes (revenue assignment) were:

•        Taxes and fees of the district or town;

•        Funds obtained from certification services rendered by the district or town;

•        Funds from movable and fixed assets;

•        Profits from investments by the district or town, and interest from their own shares and income-generating activities;

•        Fines;

•        Loans;

•        Government subsidies;

•        Legal costs and proportional rights charged by Canton Courts;

•        Funds obtained from services rendered by the district or town, or from different public infrastructures;

•        Donations and legacies.

The following taxes, formerly collected by the central government, were transferred to districts, towns, and cities:

•        Property tax;

•        Trading license tax;

•        Tax on rent.

All the expected receipts by the districts or towns are incorporated into their annual budgets.


Table 3: OSRS assignment for districts and towns of Rwanda in 2003.

On examining the lower part of Box 1, it appears that too few taxes were devolved to LGs during the previous transfer regime.

5.2 District Mean Expenditures

Analogously, district mean expenditures (MEXP) from 2003 data were used to capture information on districts’ and towns’ expenditure needs. Again, when this was used to calculate per capita mean expenditure (PCMEXP), the values were observed to be quite low by all standards and did not suffice to provide the minimum national standard of public goods and services. This reveals that most of the population might have not benefited from the public spending that LABSF intended to finance.

 

PCMEXP (RWF)

MEXP (RWF)

Mean

109.00

5,652,265

Maximum

747.80

56,495,919

Minimum

16.69

1,090,845

Std. Dev.

118.28

6,22

Observations

106

106

Table 4: PCMEXP and MEXP comparison in Rwandese Francs (RWF).

The expenditure was evidently very low, with a mean PCMEXP of 109.00 RWF, and a large standard deviation of 118.28 RWF. In addition, the maximum PCMEXP stood at 747.80 RWF with a very low minimum value of 16.69 RWF, indicating that a large part of the population could not access adequate services. Apart from the two main assigned expenditure responsibilities of health and education, district-specific recurrent obligations were enormous, which led to high degrees of fund fungibility and flypaper effects.

Districts assumed the duties and responsibilities assigned to them by laws and regulations, notably with regards to policy, administration, the economy, the welfare of the population, and culture. These included:

•        Agriculture, animal resources, and forestry;

•        Commercial activities in the districts;

•        The development of small-scale industries operating within the districts;

•        Sanitation in the districts;

•        Primary and secondary schools within the districts, training of school teachers, knowledge of programs conducted in these schools, and carrying out inspections of them;

•        Health centers in the districts;

•        Waterworks and their maintenance;

•        Tourism and the environment;

•        Land use, organization, and distribution of plots in the districts;

•        Associations and cooperatives in the districts;

•        Care for the needy;

•        Road networks in the districts;

•        Emergency services in the districts;

•        Graveyards and genocide memorial sites within the districts;

•        Attending to youth, culture, and leisure;

•        Organizing programs for women and gender development;

•        Attending to child rights; and

•        Planning youth and women’s programs.


Table 5: Districts’ and towns’ expenditure responsibility assignments in 2000.

Table 5 reveals that LGs were assigned huge expenditure responsibilities, which partly accounted for the colossal fiscal gaps observed in the next section. Those fiscal gaps showed wide disparities—horizontal fiscal imbalances at the inter- and intra-provincial jurisdictional levels. Finally, poor expenditure prioritization had led to a surge in fiscal needs, which translated into a huge mean fiscal gap (MFG)6 for subnational governments, as shown below.

5.3 Mean Fiscal Gap

Nevertheless, the assigned expenditure responsibilities in many occasions exceeded districts’ and towns’ ability to finance them via their OSRs. This situation caused vertical imbalances—in our study, this was reflected by a substantial district MFG. Three scenarios could explain this.

First, some MFG results were negative, which implied that districts’ expenditure responsibilities fell short of their revenue assignments. We recommended that these districts and towns not donate to the transfer pool; instead, that they should attract zero transfers [12]. Secondly, the MFG could be zero, where MREV and MEXP were exactly found to be equivalent; however, we did not find such a case in our study. Thirdly, in most of the districts we studied, districts’ expenditure responsibilities were in excess of their revenue capacity, which resulted in positive MFGs. This was expected, and highlighted the need for transfers from the central government to districts and towns. The results from the analysis showed that only in 27 of the 106 districts (25 percent), MEXP was lower than MREV. As such, these districts had a less pressing need for intergovernmental fiscal transfers. Table 4 shows that per capital mean fiscal gap (PCMFG) was about 48.25 RWF, while the maximum and minimum values were 395.03 and 0.86 RWF in 2003, respectively. This meant that districts and towns in Rwanda had insufficient revenue generation and lacked the spending power to effectively deliver social services.

 

PCMFG (RWF)

MFG (RWF)

Mean

48.25

2,977,072

Maximum

395.03

1,8134,039.5

Minimum

0.86

80,787.2

Std. Dev.

56.35

2,848,348

Observations

79

79


Table 6: PCMFG and MFG in Rwandese Francs (RWF).

Statistically, it is evident from Table 6 that the minimum values were set to equal zero by construction. This was accomplished after we had allowed for LGs with negative MFGs to imply that they can finance all their expenditure assignments using hard budget constraints and did not need transfers. Therefore, mean fiscal revenue, mean expenditure needs and mean fiscal gaps were the main rationale for intergovernmental fiscal transfers in Rwanda.

5.4 Central Government Transfers and Entitlements7

Given districts’ meager revenue assignment and their narrow local tax bases, the need for transfers to subnational governments was enormous. A comparison of the contents of Table 5 to the statistics in Table 4 revealed that both expenditure responsibilities and spending power were not sufficiently addressed under the previous system, and that most of the population lived in abysmal poverty.

Table 7 shows that the maximum welfare poverty index (WPI) rate from 2002 census data was 96.2 percentage points and the minimum was 9. The mean district poverty rate was very high, at 79 percentage points. The table demonstrates that with such high population density, the provision of public goods and services posed significant problems. Imposing expenditure reducing policies in these districts would only hurt the poor more.

 

WPI

DPDSITY

 Mean

79.34

699.28

 Maximum

96.2

19220.64

 Minimum

9.2

18.06

 SD

18.97

1921.11

 Observations

106

106


Table 7: WPI and DPDSITY illustrated.

Fiscal Year

Actual RRA Revenues

(RWF)

LABSF Allocations to Districts

(RWF)

 

2001

79,514,756,964

N/A

N/A

2002

93,949,100,000

1,331,500,000a

1,192,721,354

2003

111,201,700,000

1,243,500,000b

2,818,473,000

2004

-

800,694,316c

-


Table 8: RRA revenues and LABSF transfers to towns and districts: 2001–2004 (in RWF).

Table 8 reflects the actual allocation of LABSF grants to districts and towns. In 2002, LABSF allocations to districts were 12 percentage points more than the entitlement. In 2003, it was 56 percentage points less than the expected legal entitlement. In 2003, the actual entitlement was only 43 percentage points of the legally expected entitlement. This showed that the grant system was unpredictable and had significant delays—less than the mandatory amounts were released from both the MINECOFIN and MINALOC to the towns and districts in the period between 2002 to 2004.

6.1 Features of the Designed Transfer System

The new transfer system featured a model different from those found in countries like the USA [13]; Indonesia [14-16]; China [17]; Canada [18]; and Australia [19].8 Jun Ma has given special treatment to the alternative approach applied in Korea and Japan. This study avoided replicating those models and imposing them on the GoR for three main reasons.

First, they were mathematically complex and did not offer statistical gains in terms of robustness or accuracy. Secondly, the data demands of these models were extremely high and this level of statistical analyses was unnecessary in the Rwandan context at the time. Thirdly, estimations of the two sides of the formula—i.e., expenditure needs and revenue capacity—used proxy of proxies that were applied in addition to the estimated figures. In Rwanda, it was very difficult to make accurate estimates, especially at the local level. The main concern was identifying factors specific to expenditure needs and revenue capacity; for instance, population and area were factors considered while estimating expenditure needs, but ideally, both could fit in either side.

The new model for calculating intergovernmental transfers used information on revenue capacity and expenditure needs in a plausible manner. In addition, the model was participatory—districts and towns were consulted at each stage of development, and the final formula was discussed with them before it was applied; there was transparency in the implementation.

6.2 The Study’s Contribution to Knowledge and the Literature

Any new transfer system should aim to overcome the weaknesses of its precursors [20]. The designed formula aimed to allocate grants from the central government to the provinces9 (intara), as the previous transfer system seemed to promote inequality among jurisdictions. For instance, statistical data analyses revealed that “resource endowed” provinces, such as Ruhengeri and Gitarama, attracted more resources in the period under consideration.

The reformed transfer formula incorporated all the attributes of the previous transfer system, but it addressed them in a more comprehensive way. For example, it included additional information on the districts’ fiscal performance by incorporating the fiscal performance index (FPI). The index introduced elements of fiscal gap in the analysis, which captured expenditure needs and revenue capacity. The FPI considered the difference between the mean district expenditure and mean district revenue as a proportion of mean district expenditure. As noted previously, the difference between the two means was defined as the MFG.

Next, we introduced the WPI10 derived from the census data to capture the actual living conditions of households in various LGs. In the literature, poverty has a myriad of facets, and it is variously defined [21]. Intuitively, the poverty index allowed for variations in the transfer amount based on the percentage of the district population considered to be poor; this rate was district-specific.

Scatter plot for FPI and poverty proxy

Figure 1: Scatter plot for FPI and poverty proxy.


Figure 1 shows a positive correlation between the FPI and WPI introduced in the analysis for the period under consideration. Pearson’s coefficient 11 was 0.31, which implied that as FPI increases, so does WPI, as captured by the fitted upward sloping regression line. Similarly, the scatter plot shows that most districts’ poverty levels fell within the range of 75 to 96 percentage points. At the same time, these poverty levels coincided with districts that had high FPIs, concentrated within 50 to 70 percentage points. This meant that the poorest districts and towns faced budgetary difficulties, either in generating their own revenue or in prioritizing expenditure.

The designed transfer system proposed an increase in the proportion of revenue assigned to the transfer pool rate available for the allocation of grants from the central government to districts and towns. Law no. LW33/2002 [22] stipulated that the transfer should be within the range of 3–5.3 percent of the previous fiscal year’s domestic revenue collections. It was proposed that the rate be raised from 3 to 5.3 percentage points and that it should be legally binding for a minimum of period of three years. This was because domestic revenues had been increasing continuously since 200112 . Moreover, districts’ needs and expenditures responsibilities grew significantly during the period; therefore, they required more central government transfers.

Finally, the study specified a new set of plausible weights and normalization parameters to strengthen the FPI until more data became available. Alternatively, those parameters could be confirmed through sensitivity analyses or by conducting regression analyses.13

6.3 Mathematical Construct of the Proposed Transfer Formula

Analytically, let the symbols used be defined as:

hi = district ith population, N = total population of all districts, m = 106, that is, the total number of districts that were included in the analysis, gi = district ith square area, W = total transfer pool derived from the previous year’s mobilized domestic revenues,Wi = total transfer funds available for the ith district, Âi = fiscal capacity for the ith district that captures the mean of all OSRs, Ái = fiscal expenditure needs for the ith district, li = equal lump-sum amount for the ith district, and F = poverty incidence of the ith district imputed from the census data.

Then let ak stand for the shares of the five parameters used; and “k” take values from 1 to 5, reflecting the proposed five parameter model.

The study begins by defining how to derive the total transfer pool:

Where t stands for the predetermined percentage of the central government’s previous year’s mobilized domestic revenues, which is legally binding. Analogously, bt-1 is the central government mobilized domestic revenues lagged one year,

Where aiF = the ith district poverty incidence share, ail = the district ith balancing or lump-sum share, aiz = district ith land area share, aiP = district ith fiscal gap share, and aih = district ith population share.

Mathematically, the proposed formula can be modeled as follows:

Equation (2) represents the funding available for all the districts:

Where, Equation (6) is the entire district fiscal gap.

It follows then, if “full equalization” is anticipated, then the social planner should set

Equation (7) above says that if each district/town is to attain the minimum national standard of expenditure, then it should be given LABSF just equivalent to its fiscal gap.

Equation (8) captures the fiscal performance behavior of district officials and acts as a basis for monitoring and evaluating the management’s decision-making in districts and towns.

Rearranging and using Equation (8), we can specify the mean fiscal gap ratio as:

Recall the five parameters for this model, specified elsewhere in this paper, i.e., fiscal gap index, poverty incidence index (WPI), population, district land area, and equal lump-sum amount. At this juncture, weights are assigned to each parameter “k” and denoted as wk , where “k” takes values from 1 to 5, as shown earlier.

Notwithstanding the foregoing mathematical representation, we imputed the transfer shares of each individual district from the predetermined grant pool via Equation (2):

Equations (15), (16), and (17) say that the eligible districts receive transfers from the central government that are a function of the five elements defined above and weighted by plausible parameters to minimize regional resource disparities. These equations form the main result of the current study.

The proposed transfer formula was evaluated using an econometric approach to justify its superiority over the previous model, in terms of efficiency and equalizing power; Lewis [23] and Brodjonegoro and Martinez-Vazquez [24] used a similar technique. Table 9 provides the ordinary least square (OLS) regression results for the two models. In Model 1, LABSF allocations for 2003 (LABSF3) under the previous system were regressed on its determinants—population (POP), square area (AREA), and mean district revenues (MREV). The results showed that all the coefficient signs turned out as expected and the coefficients were highly significant at the 1 percent level. However, the coefficient of determination was 33 percent, which showed that the old transfer system explanatory variables did not effectively explain the variations observed in LABSF 2003. Surprisingly, despite being highly significant, the standard errors appeared too large.

Model 2 presented the simulation results of the proposed transfer system using the same transfer pool. That is, LABSF3S was regressed using the new proxies introduced: i.e., FPI, WPI, AREA, and POP. The findings indicated that all variables’ signs turned out as expected and coefficients were significant at the 1 percent level. In addition, the standard errors were very small, confirming that the newly designed system was highly equalizing. The results showed that when the district surface area increased by 1 percent, then the total transfer claims rose by 35 percentage points. This was because large area districts required more resources to work on their vast lands. Population carried a high level of importance, with the largest coefficient contribution to the transfer demand—if it increased by 1 percent, transfer needs increased by 45 percentage points. This is because higher population growth leads to a greater demand for public goods and services.

Similarly, the WPI and FPI indices portrayed the same trend: that is, if they increased by 1 percent, both triggered increases in LABSF3S of 41 and 40 percentage points, respectively. Interestingly, all the variables together by construction explained 100 percent of the variations observed in all the transfers. These findings agreed with the Indonesian analysis [22].

 

 

Model 1

Model 2

(Constant)

511411.60

(581411.70)

386504.07

(0.008)

POP

0.397

(6.965)

0.454

(0.000)

AREA

0.394

(999.639)

0.349

(0.000)

MREV

-0.131

(0.020)

-

 

FPI

-

0.400

(0.007)

WPI

-

0.412

(0.000)

Adjacent R²

0.334

1.000

Observations

106

106

Standard errors in parentheses; all coefficients are significant at the 1 percent level.


Table 9: OLS Regressions. Dependent Variables Model 1: LABSF3; Standardized Model.

The comparison of the results of Models 1 and 2 seemed very intuitive, especially regarding the coefficients of MREV (which was a proxy for poverty) and the new proxy denoted by WPI. The impact of WPI on LABSF3S was greater by 28 percentage points—in absolute terms—compared to MREV, regardless of the signs. In Table 10, the PCMFG is regressed on its main determinants. The results revealed that by construction as expenditure needs increased by 1 percent—captured by PCMEXP—this significantly increased the districts’ mean fiscal gap by 26 percentage points. In addition, an increase in the revenue capacity by 1 percent—captured by PCMREV—implied a decrease in the mean fiscal gap, by 17 percentage points.

Variable

Model 2

(Constant)

19.80

(3.605)

PCMREV

-0.166

(0.025)

PCMEXP

0.259

(0.035)

Adjacent R²

0.371

Observations

106

Standard errors in parentheses; all coefficients are significant at the 1 percent level.


Table 10: OLS Regression. Dependent Variable: PCMFG Unstandardized Coefficients.

The two independent variables explained about 37 percentage points of the total variations observed in the PCMFG. These results were intuitive, since the only way for districts to minimize their fiscal gaps was to simultaneously implement policies geared towards enhancing OSR generation, while imposing hard budget constraints.

Table 11 shows the double log OLS regression results, where the coefficients can be translated as coefficients of elasticity. The signs of the district log indicated that per capita revenues (LPCMREV) and per capita mean expenditures (LPCMEXP) were as expected and highly significant. The model’s explanatory power was 63 percentage points.

Variable

Model 3

(Constant)

5.990

(0.154)

LPCMREV

–0.204

(0.016)

LPCMEXP

0.142

(0.024)

Adjacent R²

0.628

Observations

106

Standard errors in parentheses; all coefficients are significant at the 1 percent level


Table 11: OLS Regression. Dependent Variable: LABSF3 Unstandardized Coefficients.

The implication was that enhancing the capacities of districts for revenue generation and imposing policies that reduce expenditure, as proposed, would have an equalizing effect. On the one hand, a 1 percent increase in revenue would lower transfers from the central government by 20 percentage points, holding districts’ spending levels constant. On the other, when districts increased their expenditure by 1 percent, their transfer levels would also increase by 14.2 percentage points.

The study corrected the anomalies of the grant transfer system practiced in Rwanda. First, the study explained why the suggested approach was better suited to Rwandan needs during that period (2003), instead of the generic approaches applied in other developed and developing countries. Several notable lessons should be remembered while designing equalization transfer formulae for countries in the initial stages of implementing decentralization:

1. The study corrected the anomalies of the grant transfer system practiced in Rwanda. First, the study explained why the suggested approach was better suited to Rwandan needs during that period (2003), instead of the generic approaches applied in other developed and developing countries. Several notable lessons should be remembered while designing equalization transfer formulae for countries in the initial stages of implementing decentralization:

2. It is hard to estimate revenue capacity with limited data sources, especially for countries in the initial stages of implementing fiscal decentralization; apply mean of actual own revenues at both levels of government.

3. Analogous to the above two points, capture expenditure needs through mean of actual expenditures. This is useful until sectoral data become available.

4. Apply the derived mean revenues and mean expenditures to reflect revenue capacity and expenditure needs, respectively, to establish a mean fiscal gap, which is an important parameter to ensure equalization.

5. Worry about how the various parameters and indices are derived and combined, and enter the model. Determine the effects based on tests using actual data, about whether these parameters should enter the model multiplicatively, additively, divisibly, or in a subtractive way.

6. Use the poverty measure from household budget surveys, calculated by most countries while formulating their poverty reduction policies. This is the most plausible proxy, since it captures the country’s poverty incidence for a given subnational government, which the formula intends to minimize.

In conclusion, we suggested a pragmatic model that was successfully implemented to address previous intergovernmental inequities. Apart from other structural problems, the findings helped to consolidate the political gains from implementing fiscal decentralization in Rwanda. Furthermore, the study advanced several policy recommendations, which were all adopted.

Ludovick Leon Shirima, Department of Business, Higher Colleges of Technology, Abu Dhabi Men’s College, Abu Dhabi, UAE, Email: lleon@linuxmail.org

[1] MINALOC. (2001). National decentralization policy. Kigali, Rwanda: Government of Rwanda.

[2] MINALOC. (2000). Fiscal and financial policy. Kigali, Rwanda: Government of Rwanda.

[3] Bird, R. M. & Smart, M. (2002). Intergovernmental fiscal transfers: International lessons for developing countries. World Development 30(6), 899–912.

[4] Financial and Accounting Manual (2003).

[5] Boex, J. (2002). Concepts in the design of intergovernmental fiscal transfers. Presentation for UNCDF, Local Bodies Fiscal Commission and Georgia State University.

[6] Searle, R. J. (1994). Intergovernmental fiscal transfer: Institutions and data. Paper presented at the Qingdao Conference on Intergovernmental Fiscal Transfer in China, sponsored by the IMF and the Ministry of Finance, July 25–29.

[7] Bahl, R. W., & Linn, J. (2002). Urban public finance in developing countries. New York: Oxford University Press.

[8] Olson, M. (1969). The principle of “fiscal equivalence”: The division of responsibilities among different levels of government. American Economic Review 59(2), 479–87.

[9] Ehdaie, Jaber (1994). Fiscal decentralization and the size of government. Policy Research Working Paper Series No. 1387. Washington DC: World Bank.

[10] Martinez-Vazquez, J. (2001). An evaluation of Indonesia’s DAU equalization grant. Jakarta, Indonesia: IRIS/USAID.

[11] McLure, Jr. C. E. (1994). The tax assignment problem: Ends, means, and constraints. Australian Tax Forum 11, 153–83.

[12] Davis, A. & Lucker, R. (1982). The rich-state-poor-state problem in a federal system. National Tax Journal 35(3), 337–63.

[13] Rafuse, R. W. (1991). Revenue raising powers, practices and policy coordination in the federal system of the United States. Processed.

[14] Shah, A. (1994a). “The Reform of Intergovernmental Fiscal Relations in Developing and Emerging Market Economies.” Policy and Research Series No. 23. Washington, DC: World Bank.

[15] Shah, A. (1994b). “Intergovernmental Fiscal Relations in Indonesia: Issues and Reform Options.” Discussion Paper No. 239. Washington, DC: World Bank.

[16] Lewis, B. D. (2002). Revenue – Sharing and grant-making in Indonesia: The first two years of fiscal decentralization. In Smoke, P. (Ed.), Intergovernmental transfers in Asia (forthcoming). Manila: Asia Development Bank.

[17] Agarwala, Ramgopal. (1993). Reforming intergovernmental fiscal relations in China. World Bank Discussion Paper No. 173. Washington, DC: World Bank.

[18] Broadway, R. W., & Hobson, P. A. R. (1993). Intergovernmental fiscal relations in Canada. Canadian Tax Paper no. 96. Canada: Canadian Tax Foundation.

[19] Rye, R. C., & Searle, B. (1996). The fiscal transfer system in Australia. In Ahmad, E. (Ed.), Intergovernmental fiscal transfer (forthcoming). Washington, DC: International Monetary Fund.

[20] Shah, A. (1991). “Perspectives on the Design of Intergovernmental Fiscal Relations.” Policy Research Working Paper Series No. 726. Washington, DC: The World Bank.

[21] Shirima, L. (2009). War aggravating poverty: A case for Angola. Hamburg, Germany: Lambert Publishers.

[22] Government of Rwanda. (2002). Law number 33/2002 of 2002. Kigali Rwanda: Government of Rwanda.

[23] Lewis, B. D. (2001). The new Indonesia equalization transfer. Bulletin of Indonesia Economic Studies 37(3), 325–43.

[24] Brodjonegoro, B., & Martinez-Vazquez, J. (2002). An analysis of Indonesia’s transfer system: Recent performance and future prospects. Conference on Can Decentralization Help Rebuild Indonesia? Atlanta, Georgia: Andrew Young School of Policy Studies, George State University.