Development Discussion Papers

Transcription

1 Development Discussion Papers Why Governments Devolve: A Study Using Data from Indian States Sanjay Mitra and Shashi Kant Verma Development Discussion Paper No. 586 June 1997 Copyright 1997 Sanjay Mitra, Shashi Kant Verma, and President and Fellows of Harvard College Harvard Institute for International Development HARVARD UNIVERSITY

2 WHY GOVERNMENTS DEVOLVE: A STUDY USING DATA FROM INDIAN STATES SANJAY MITRA, SHASHI KANT VERMA Recent studies have focused upon the role of devolution of financial and administrative authority to local governments. It is generally believed that devolution leads to better implementation of public projects and programs, and to improvements in the quality of governance and democratic participation. This paper seeks to establish a simple methodology to explore the determinants of devolution and thereafter to broadly examine its impact on government policies. Using data from the states of India, we show that the extent of devolution is determined by the extent of caste or communal diversity in a state, the stability and the policy orientation of the ruling political party at the state level, and, most importantly, the prevailing attitude of its political and bureaucratic elite towards the delegation of financial and administrative authority. We show that greater devolution helps in containing those expenditures of the state government that are purely administrative in nature. A higher degree of devolution also enhances the extent of electoral participation, stimulates productive economic behavior measured in terms of greater agricultural output and increases the demand for social services. We conclude that devolution can be largely explained by political and institutional factors. Economic factors play a relatively smaller role. We suggest that devolution represents the degree of internal openness of the political and economic system., analogous to the concept of external openness used in development literature. Sanjay Mitra belongs to the Indian Administrative Service. He has worked for state and local governments as well as the federal ministry of Rural Development of the Government of India. He has degrees from Delhi University and a Master in Public Administration from the John F. Kennedy School of Government, Harvard University. Shashi Kant Verma is a doctoral student at the John F. Kennedy School of Government, Harvard University. Prior to coming to Harvard he used to work for a private firm in India. He received a Bachelor of Technology degree from the Indian Institue of Technology, Kharagpur and a Master in Public Policy degree from the Kennedy School at Harvard.

3 CONTENTS INTRODUCTION 1 BRIEF REVIEW OF EXISTING LITERATURE 1 SCOPE OF THE PAPER 1 THE INDIAN CONTEXT 2 PART I: MEASURING DEVOLUTION - COMPOSITE DEVOLUTION INDEX 3 PART II: THE DETERMINANTS OF DEVOLUTION 5 ETHNIC/LINGUISTIC DIVERSITY 7 WILLINGNESS TO DEVOLVE ( WTD): 7 POLITICAL STABILITY AT THE STATE LEVEL 8 POLICY ORIENTATION OF THE STATE GOVERNMENT 9 THE MODEL 10 PART III: THE IMPACT OF DEVOLUTION 13 IMPACT ON AGRICULTURAL GROWTH 14 DEVOLUTION AND ELECTORAL PARTICIPATION 14 PART IV: INTERPRETING THE RESULTS 14 CONCLUSIONS 15 LIST OF TABLES TABLE 1: RATIO OF RURAL DEVELOPMENT FLOWS TO TOTAL STATE EXPENDITURE.(PERCENTAGE) 17 TABLE 2A: FINANCIAL AUTONOMY(α) 17 TABLE 2B: FUNCTIONAL AUTONOMY (β) 18 TABLE 2C: COMPOSITE DEVOLUTION INDEX 1, (CDI 1) 18 TABLE 3: ETHNO- LINGUISTIC FRACTIONALIZATION INDEX 19 TABLE 4: DIVERSITY INDEX (DI) 20 TABLE 5: WILLINGNESS TO DEVOLVE (WTD) 20 TABLE 6: CHANGES IN ADMINISTRATION AND POLITICAL PARTIES 21 TABLE 7: TIME SPENT UNDER NON-CONGRESS GOVERNMENTS. 22 TABLE 8: EFFECTS OF DEVOLUTION 23 TABLE 9: QUALITY INDEX 24 TABLE 10: GROWTH IN AGRICULTURAL OUTPUT ( RS./HECTARE) 24 TABLE 11: ELECTORAL PARTICIPATION RATES 25

4 WHY GOVERNMENTS DEVOLVE: A STUDY USING DATA FROM INDIAN STATES SANJAY MITRA AND SHASHI KANT VERMA B-304, 79 JFK Street, Cambridge MA 02138, USA vermash@ksg.harvard.edu Introduction Devolution and decentralization are important themes in current development thinking and practice. Initial resurgence of interest in this field was due to the empirical findings regarding the higher efficiency and better targeting of programs and projects with decentralized decision-making. In recent years, a widely held belief regarding the causal relationship between decentralization and governance has driven further investigations into various aspects of decentralization and devolution [Cohen and Peterson, 1996]. 1 Brief review of existing literature Development agencies, such as the World Bank [Cheema and Rondinelli, 1984; Silberman, 1992] and the UNDP [Klugman, 1994a], have examined various aspects of decentralization and devolution in considerable detail. They concluded that the decentralization of public decisionmaking has an important, positive impact on development initiatives of all kinds, in addition to fostering a greater degree of public participation. However, in most cases the inadequacy of suitable data renders the extent of the impact indeterminate. Scope of the Paper We aim to deal principally with devolution rather than decentralization. Devolution implies cases where administrative and financial responsibilities are assigned by law to distinct, corporate entities with substantial discretionary authority. At best, decentralization is essentially an administrative mechanism wherein the quantum and nature of local level decision-making is Sanjay Mitra is a graduate of the John F. Kennedy School of Government, Harvard University. Shashi Kant Verma is a doctoral student at the Kennedy School of Government, Harvard University. The authors gratefully acknowledge detailed discussions with Professors John Donahue, Merilee Grindle and Richard Zeckhauser at the Kennedy School. Thanks are also due to Anuradha Mitra for useful discussions. All correspondence may please be addressed to Shashi Kant Verma 1 For clear definitions of decentralization and devolution also see Cheema et al,

5 strongly subject to the current attitudes and perceptions of the governing elite at the national or provincial levels. The absence of corporate entities with adequate legal status, in conjunction with the inevitable political constraints, often renders any decentralization measures unsustainable in the long-run [Klugman, 1994b]. Hence the emphasis on devolution. In Part I of the paper, we propose to develop a set of simple indices designed to measure the extent of devolution in the developing countries, using readily available, standardized data. In Part II, we examine the factors that determine the extent of devolution. Part III will look at the effects of devolution on government expenditures for selected items, electoral awareness and participation in and the quality of certain development programs. Finally, in Part IV, we aim to analyze some issues specific to India arising out of our analysis. Due to the paucity of comprehensive and comparable developing country data on local government 2, we used data from the states of India, many of which are larger than most developing countries of the world. India also has the added advantage of having recently gone through a constitutional exercise aimed at establishing elected local governments at the village, sub-district, district and municipal levels. 3 Use of data from India may affect some of the general conclusions of our model. However, since the entire exercise involves standard definitions, methodologies and readily accessible data [Bennett, 1994], familiar to students of local government, we feel that our model will be useful in conducting similar analyses for other developing countries. Before commencing our analysis, we provide a brief overview of the status of devolution in India for readers not fully conversant with the details. The Indian Context India is a federal country, with strong elected governments at the state (provincial) level. There are well defined constitutional arrangements for power and resource sharing between the Center (federal government) and the States. Of late, Center-State relations have become matters of serious concern and there is increasing pressure to further review the existing arrangements. The 2 3 Indonesia is a notable exception. See Qureshi and Shah, This refers to the 73rd and 74th Amendments to the Indian Constitution in More than 2 million elected functionaries have since assumed office. ( Source-Ministry of Rural Areas and Employment, Government of India, New Delhi) 2

6 73rd and 74th Constitutional Amendments, enacted in 1992, seek to encourage local government by creating viable units 4 of self governance at the state, sub-district (town) and the district levels. The essential logic for creation of units of self governance below the state level is four fold. First, it allows greater participation by the people in the process of governance. This is particularly important in India, where a member of parliament often represents a million electors and a member of the state legislative assembly up to a quarter million. Second, it is expected to lead to greater efficiency in the production and distribution of public goods. Third, it is expected to provide a good training ground for public officials destined for higher responsibilities and last but not the least, it addresses the general desire for a more accessible and responsive government. The arrangements for power sharing between the center, states and local bodies (LBs) has a number of potential stress points. While the states are very keen on greater devolution of powers from the center, they are noticeably less so regarding the devolution of powers from the states to the LBs [Adisheshiah, 1986]. Resource raising and sharing is another problem area. The center retains exclusive rights to a number of important tax sources, including income, corporate and excise taxes, while most of the spending functions are at the two lower levels. This leads to a complicated system of revenue transfers from the center to states and local bodies, which has often been used to wield undue authority. Unlike developed countries such as the USA where the federal government is alleged to exercise indirect control over local governments through regulations and mandates, the center in India wields its authority mainly by restricting the amount or use of funds for the local bodies. Thus, the extent and manner of devolution from the states to LBs is important in analyzing the economic and political development of India. 5 Part I: Measuring Devolution - Composite Devolution Index The first step is to construct a Composite Devolution Index (CDI) for the major Indian states, using standard definitions and techniques well-known in the literature [McCormick and Wolman, 1994; Bennett, 1994a; Bennett, 1994b; Groenewegen, 1994; and Bird and Wallich, 1994; also World Development Report, 1988; Lamberte, 1993; and Shah, 1991 for country specific decentralization studies]. 4 These are called panchayats in the rural areas and municipalities in the urban areas. 3

7 We define α as a measure of financial autonomy in a given year, where α = Total state expenditure on assignments and grants to local bodies in a given year Total state expenditure in the same year This is a line item in the state budgets, and thus directly measured. Further, we define a measure of functional autonomy in a given year, β = Quantum of untied funds at the disposal of local bodies in a given year Total funds available with local bodies in the same year As a first approximation, the sum of the Rural Development (RD) and the Urban Development (UD) Budgets, can be used as a reliable proxy for the quantum of untied funds available to the Local Bodies. This assumption is inevitable given the extreme difficulty of determining the extent to which other state government funds are untied 6 and entirely at the disposal of the local bodies. The assumption is, however, justified by the sizable proportion that the combined RD and UD flows occupy in the total non-salary expenditures of the states (Table 1). According to existing guidelines of the Ministry of Rural Development [Ministry of Rural Areas and Employment, 1994], approximately 60% of the total RD flows are exclusively available to the local bodies as untied funds. Thus, it is possible to re-define functional autonomy as β = Total local body assignment (RD flows) + UD flows Total fund availability The two indices are then normalized to yield the Composite Devolution Index (CDI 1) after applying equal weights to both. See Table 2c Much has been written about the extent and manner of devolution. For a general introduction, see Varughese, 1981 Tied and untied funds have their usual meanings. Tied funds refer to funds given to LBs for a particular program and are subject to stringent conditions. Untied funds do not have such conditions attached to them and often represent the proceeds of tax sharing. Own resources, in general, constitute a small proportion of the resource base. See Dutta, See Bennett, 1994 for better ways to define and measure functional autonomy, particularly those regarding activity-wise disaggregation. However, this may need much more detailed budgetary data and it is not clear whether or not it would add to the level of insight. Resource-based indices, showing the percentage of own LB resources to the total resource-base and the total expenditure of the Local 4

8 α measures the size and importance of the Local Body (LB) sector in the states expenditure profile. β measures the extent to which LBs are free to allocate their resources to activities of their choice. A lower value of β would mean that the LBs are mainly being used as agencies of the state/federal governments and not functioning as truly autonomous units of local self governance. We also examine another index, CDI 2, defined as the ratio of total untied flows to the total state expenditure. (See Table 3), where a higher CDI 2 means a higher degree of emphasis on local level functional autonomy. Part II: The Determinants of Devolution What causes governments to devolve powers upon other, administratively smaller governments -- in this case, from the state (provincial) to the local government? To start with, the state government should be willing to divest itself of some of the powers and responsibilities. This willingness to devolve, characterizes the attitudes and perceptions of the existing bureaucratic and political elite, especially in overcoming natural feelings about the loss of patronage [Sab, 1986], lesser opportunities for rent seeking and the growth of alternative spheres of influence 8. There can also be a number of perfectly legitimate apprehensions regarding the availability of resources and the possibilities of power capture by the local elite to the detriment of the disadvantaged groups and indigenous peoples [Satishchandran, 1994] The other significant factor, particularly in democratic polities, is likely to be the extent of ethnic, caste or group diversity 9. The presence of a large number of ethnic or caste based groups who feel inadequately represented by the state legislators, may increase demands for a greater degree of decentralized decision making. An elected government is quite unlikely to go in for of major changes if it is not assured of security of tenure. Very rapid turnovers in the ruling party at the state level is likely to lead to a Bodies, would no doubt be better indicators respectively of the levels of functional and financial autonomy. However, tax data for most developing countries, including India, is rather incomplete. See also, Oommen, Ghosh B., State Institute of Panchayats, Government of West Bengal In the Indian context, ethnic diversity could be supplanted by caste/communal diversity. 5

9 preference for survival- first policies rather than innovations in the nature and style of governance. Finally, the policy orientation of the ruling party at the state level would be another major factor in determining the extent of devolution. In the Indian context, it is a well-recognized fact that, traditionally, opposition Congress governments 10 have been perceived to be more sympathetic towards a greater devolution of financial/administrative powers than the Indian National Congress, given its traditions of centralized decision-making [Mathew, 1994; Das and Mahapatra, 1986]. Thus, as a first approximation, it is suggested that the current extent of devolution, CDI1, is determined by the following factors 11 attitudes and perceptions of the governing elite at the state level, which we call Willingness to Devolve (WTD), the diversity of the population of the concerned state, using the number of caste/community based groups as a measure, which we call Diversity Index (DI), stability of the state administration, defined as the number of changes of the ruling party at the state level (PARTY CHANGE) over the past two decades (since 1977) and policy orientation of the state governments towards devolution, measured either as the proportion of time since 1977 spent under opposition administrations (%OPPOSITION), or as the proportion of state administrations politically opposed to the major ruling party at the 10 That is, state governments ruled by political parties other than the Congress. The Congress or the Indian National Congress has ruled at the federal level almost continuously, for 43 out of 50 years since Independence in 1947, barring short spells in , and since (continuing). It had three Prime Ministers from the same family, Jawahar Lal Nehru, his daughter, Indira Gandhi and her son, Rajiv Gandhi. Together the three were Prime Ministers for 38 years, causing widespread concern over family rule. The strong personalities of the three also contributed to a highly personalized style of politics and an inevitable concentration of power. Also see, Mathew, 1994; Verma, 1981; 11 The choice of 1977 as the relevant cut-off date is due to reasons specific to India, namely the restoration of constitutional governance after a brief period of its suspension from Similarly, the choice of the last variable, %OPPOSITION ADMN is due to the reasons elaborated above and in Das Mahapatra, 1988a; Verma, 1981; and Mathew

10 federal level, i.e., the Congress for most of the period under consideration (%OPPOSITION ADMN). Ethnic/linguistic diversity Here, we further developed the methodology used by Mauro, where Ethno-linguistic Fractionalization is defined as EFI = 1 M 1 ni N 2 where M is the number of distinct groups, N the total population and n i, the population in the i th group. EFI gives the probability that any two persons chosen at random from within a group belong to different ethno-linguistic groups [Mauro, 1995]. The higher the EFI, the higher the extent of ethnic diversity. However, considering the fact that diversity in the Indian context implies using the caste and kinship factor, often more pronounced in its divisive potential, we considered it more appropriate to use another measure of diversity. Thus, we took recourse to the extremely comprehensive People of India Project (POIP) undertaken by the Anthropological Survey of India (ASI) during According to the POIP data, there are 2753 distinct communities in the country, distinguished by identity, occupation, kinship and marriage rules, food habits, ecology etc. [Gadgil, 1993]. If a community occurring in more than one state is counted for each state, there are 4635 communities. Table 4 gives state-wise figures for the number of communities. It also shows the total number of traits represented in the communities in a particular state. Since the POIP study encompasses characteristics that are particularly important in India, it is a good measure of dissimilarities among the communities present in each state. The higher the Diversity Index, the greater the extent of diversity in the concerned state. Note that states like Karnataka and Maharashtra, which include the caste features of both north and south Indian society, have high DIs whereas Haryana and Punjab, which clearly form a part of north Indian society, have fairly low DIs. Willingness to devolve ( WTD): WTD quantifies the attitudes and perceptions of the ruling elite at the state level. It has long been recognized that these are possibly the most important factors behind the readiness of the government of any developing country to part with powers and patronage. Greater the WTD, the 7

11 more positive the attitude of the ruling elite towards a devolution of powers to the local bodies and higher the extent of devolution. We compared actual flows of funds over the last four years from the state to the LBs with the levels of fund flows recommended by the Tenth Finance Commission (TFC) 12. The TFC has, for the first time, recommended targeted flows from the states to the LBs over the period (See Table 5) We measure WTD as the ratio of actual flows and recommended flows of funds over a period of time. CDI1, on the other hand, is derived from actual flows of tied and untied funds from the state to the LBs in a particular year and the total state expenditure in the same year. 14 To elaborate further, a four year expenditure profile should suffice to indicate the policy imperatives of the state government vis-à-vis devolution. High flows in a single year could be the result of many factors, including short term electoral gains. Comparing the actual flows over a four year period with the norms set by the TFC 15, would indicate the extent to which the state elite was prepared to match the needs of LBs. Political stability at the state level There are essentially two measures. One, the total number of changes in administration in a given state (STABILITY). A change in administration means a change of the elected state chief executive, called the Chief Minister in India. Political stability can also be defined as the number of changes in the ruling party at the state level during (PARTY CHANGE). The two measures are different, since it is quite The Tenth Finance Commission, the latest in a series of expert groups with constitutional status, entrusted with the task of developing resource allocation principles between the center and the states. For a discussion on finance commissions in general see Das and Mahapatra, 1988b. TFC data quoted in Oommen, It is also possible to calculate WTD by looking at the number of LB elections held since 1977 and prior to As mentioned earlier, the relevant Constitutional Amendment Act related to LBs was passed in 1993 and the states have since had no choice but to implement the provisions of the Act, which include, the holding of regular elections, establishment of mechanisms for resource transfer and the assignment of specific responsibilities and taxes to the LBs. The higher the number of LB elections prior to 1993, the higher the WTD of the concerned state government, since no government unwilling to devolve would go in for LB elections and create a group of vocal critics at the local level. However, this posed a few problems since quite a few state governments had held elections to different tiers/categories of LBs in different years [Mathew, 1986], rendering meaningful interstate comparisons of the number of LB elections during , difficult. TFC recommendations are finalized after detailed discussions with the states and involve a thorough scrutiny of the receipt and expendture positions of the states. 8

12 common to have changes of Chief Ministers without having a change in the ruling political party. For example, during the period the state of Bihar had 12 different governments (administrations), but since 9 of them were from the Congress this was associated with only four changes in the ruling party. In contrast Kerala, which had 13 governments in the same period had 12 changes in the ruling party, since every change of government was associated with a change in the ruling party. For the purposes of this exercise we counted phases of direct bureaucratic rule by the federal government, often in supercession of democratically elected state governments, called President s Rule, as a change in the ruling party in the state concerned. This makes sense, since President s Rule has often been used for direct political purposes by the federal government to try and modify state policies 16. As mentioned earlier, the choice of 1977 as the relevant cutoff year is governed by the fact that 1977 marks the restoration of formal democratic process in India. This period also saw the institution of a high-level Committee under the Chairmanship of a senior statesman to examine the local body system. 17 Fixing similar cutoff dates for other countries should not be very difficult. Policy Orientation of the State Government The policy orientation of the state government towards devolution is defined as the proportion of time spent since 1977 under opposition governments (%OPPOSITION), i.e., governments opposed to the predominant federal ruling party, the Congress. (See Table 7.) It is also possible to define the policy orientation as the ratio (percentage) of the number of opposition administrations to the total number of administrations since 1977 (%OPPOSITION ADMN). (See Table 5.) Article 356 of the Indian Constitution permits the temporary supercession of elected state governments by the federal authorities. Though the grounds specified in the Constitution for doing so are very grave, the relevant provision has often been interpreted to suit partisan political ends. Article 356 has been under debate for much of the past decade, especially as it is seen as inhibiting federalism. See Das and Mahapatra, 1986; Varughese and Kurien, 1981; and Chandrashekhar, 1988 for general discussions on this issue. This refers to the Asoka Mehta Committee. Recommendations from this committee form the basis of recent attempts at devolution. See Mathew, 1986c. 9

13 It has long been felt by students [Mathew, 1994; Verma, 1981; Bhambri, 1981] of Indian politics that the Congress party, with its tradition of family-run leadership and centralized power structure, has not been in favor of decentralization 18. On the other hand, almost all opposition governments, irrespective of their ideologies, have attempted to put into effect, varying degrees of devolution. 19 It turns out that the latter, %OPPOSITION ADMN, is a better explanatory variable and is entirely consistent with the general appreciation of state politics in India. The difference in policy emphasis on devolution between Congress and non-congress governments is large [Gani, 1990; Verma, 1981; Mathew, 1994]. It has been generally observed that in most cases, whenever non- Congress parties have come to occupy power in a state, they have tried to expand their support base by creating favorable political constituencies at the district and village levels, mainly to counter the powers of the central government run mostly by the Congress. Each such move towards devolution has proved even harder for intervening Congress administrations to undo. The greater the number of chances that non-congress parties have had to govern a state, the greater the relative degree of emphasis on devolution, notwithstanding the fact that many of these administrations have been quite short lived. (See Tables 6 and 7.) For example, Kerala has seen 13 changes of party. Of these 5 (38%) have been non-congress. This has contributed to a high CDI for Kerala. In contrast, Bihar (%OPPOSITION ADMN = 33%, %OPPOSITION = 47.9%) has a low CDI, even though it has spent a long time under non-congress rule. The Model Consider the model CDI1 20 = f(wtd, DI, PARTY CHANGE, %OPPOSITION ADMN, STABILITY), where the function is linear in parameters as well as variables. In that case we have the following functional form CDI1 = α 1 + α 2 (WTD) + α 3 (DI) + α 4 (PARTY CHANGE) + α 5 (%OPPOSITION) + α 6 (STABILITY) + α 7 (WTD*STABILITY) + α 8 (%OPPOSITION ADMN) 18 Honorable exceptions are the states of Gujarat and Maharashtra. 10

14 Regression results: Dependent variable CDI1 Variable WTD * * * * * * * (7.099) (6.619) ( (5.135) (6.849) (6.763) (7.849) DI * * (1.718) PARTY CHANGE * (2.580) * (2.37) ) * (1.971) * * (1.922) * (2.143) * (3.071) * (3.911) (0.818) * (2.455) STABILITY (0.458) (1.160) WTD*STABILTY (-1.188) %OPPOSITION * (1.850) (1.046) %OPPOSITION ADMN * (2.049) 0.196* (3.269) Constant (1.320) 0.155* (3.47) * (5.760) * (2.597) * (3.922) * (2.721) (0.635) R R 2 (adjusted) F-Statistic From these results it is possible to state, as a first order estimation, that the basic determinants of devolution are, willingness to devolve (WTD), an essentially institutional factor reflecting attitudes and perceptions of the ruling elite at the state level; the extent of caste/kinship group diversity in the state (DI) ; the extent of Non-Congress rule during , as measured by (%OPPOSITION ADMN) and the stability of the political policy of the state government (PARTY CHANGE). However, it turns out that using a slightly different definition of CDI, i.e. CDI 2, where CDI 2 = Total untied flows to LB (as defined in Part I) Total state expenditure Once again, there are exceptions. Bihar has not held LB elections since 1980 and continues to resist despite clear Constitutional directives. CDI values refer to the current, values. 11

15 gives the intuitively 21 correct (negative ) sign for the number of changes in political ideology, a large number of which should hamper attempts at devolution. Regression results: Dependent variable CDI2 Variable 1 2 WTD 2.964* (2.496) 0.853* (2.421) DI (1.533)) (0.126) PARTY CHANGE * (-2.061) * (-2.143) SDP * (-2.017) Population (0.833) % SDP (-0.341) Constant 2.964* (2.496) 1.870* (3.200) R Though this redefinition does yield the correct signs for PARTY CHANGE, it is our feeling that CDI 2 does not adequately capture the role of the state government in the devolution exercise. This is because it simply measures the total untied flows as a fraction of the total state expenditures, which tends to get biased towards the much larger central flows under the rural development account, without taking into consideration the states own level of commitment (Index α). 21 The positive sign for PARTY CHANGE could well reflect the fair degree of correlation between %OPPOSITION ADMN and PARTY CHANGE. It is also true that in the Indian context, a large number of party changes, as distinct from mere changes in administration, would imply a greater tendency towards non-congress governments. This could well be a partial explanation for the rather counter-intuitive regression result. 22 Per capita state annual domestic product 23 Growth rate of annual per capita state domestic product 12

16 Part III: The Impact of Devolution The results in this section are derived from a preliminary analysis of some of the trends in state finances and development in India. It is by no means comprehensive, though it does provide some interesting explanations. We propose to undertake a more thorough investigation of the effects of devolution in a forthcoming paper. Unlike Part II, where it was possible to hypothesize meaningfully regarding the determinants of devolution, it is far more difficult to isolate and quantify the effects of devolution. For example, the precise linkage between devolution and state-wise poverty trends is not easy to compute. At best, CDI could turn out to be one of several explanatory variables. Other factors such as inflation and rural wages are certain to be highly significant as determinants of poverty. However, a simple exercise, wherein the states are divided into high and low CDI categories and the relationships between CDI and a number of plausible variables are examined, yields surprisingly rich results. Intuitively, greater devolution should lead to a lower emphasis on the state government. With this idea in mind, we decided to look at state-government expenditures on purely administrative matters. 24 Miscellaneous Services in the Indian context means expenditure on the State Secretariat (the Administrative Headquarters). It appears that for low CDI states, an average of 7.47% of total state expenditures is on Miscellaneous Services, as compared to only 0.75% by the high CDI states. It also turns out that the low CDI states registered an average of 17.58% annual compound growth in total non-development 25 expenditure (net of debt-service) during , in contrast to the high CDI states, which showed a 11.31% growth. Similarly, expenditure on Police in low CDI states was 4.7% of total state expenditure and 3.95% in high CDI states Accounting norms for all states in India are uniform and are set by the Comptroller and Auditor General of India. Thus, the data is comparable across states, unlike some countries like the USA where state accounting norms often vary. Non-development expenditure includes Organs of State, Miscellaneous Services, Police, District Administration, Debt Service and Expenditure on the Collection of Taxes. There is interesting anecdotal evidence to back up this statistical finding. A recent evaluation of the Local Government (Panchayati Raj) in West Bengal (CDI1=0.568), points out the relative success of Local Government functionaries in containing communal violence in 1992 [Mukherjee and 13

17 Expenditure on district administration in high CDI states was 0.67% of total state expenditures as against 0.9% in high CDI states, an entirely expected trend since greater devolution of powers to LBs should lead to a lower degree of emphasis on the bureaucratic mechanism at the district level. Do more evolved LBs lead to higher expenditures on social services, given that local governments are likely to be more interested in the expansion of social services [Klugman, 1994c] It does seem to be true. High CDI states tend to spend 4.56% more (as a percentage of total expenditures) than low CDI states. All the above differences between low and high CDI states are also statistically significant at the 5% level. Impact on Agricultural Growth Agricultural output is measured in Rupees/ hectare in Rupees. For this particular correlation, a modified CDI is used, called the Rural CDI. This is derived from the CDI by abstracting the UD flows, which go to urban bodies and Municipalities. Clearly, the difference in the growth of agricultural output between high and low CDI states is statistically significant. Devolution and Electoral Participation Though there appears to be a substantial difference in the post-1977 turnout rates in elections to the state legislatures 27 between the high and the low CDI states, 64.93% versus 60.1%, it is not statistically significant 28. (Table 14). Part IV: Interpreting the Results The foregoing analysis clearly has important policy implications for India. Changes in the ruling party, diversity and the policy orientation of the state governments are all significant variables in the determination of the extent of devolution. However, the structural characteristic 27 Bandyopadhyaya, 1994]. This fact is also borne out by one of the authors ( Mitra) personal experience as a District Officer during that period. Only elections to state legislatures have been considered, since Parliamentary elections are often influenced by extraneous factors not immediately connected to local issues 14

18 WTD is of greatest interest. Though a formal examination of the factors underlying WTD is beyond the scope of this paper, it would be reasonable to conjecture that it depends on the absolute size of the state deficit, its growth, the extent of political commitment and the attitudes/perceptions of the state-level bureaucracy ( these attitudes could once again be linked to the stability of the state administration, since it would be natural for the permanent bureaucracy to disregard the wishes of an administration perceived to be transient). The second part of our analysis, though preliminary, shows that there is a significant correlation of the devolution index with selected items of government expenditure. This would have important policy implications for the government. It is likely, as our results show, that devolution is linked to improvements in the expenditure position of the state government, a slower growth in non-developmental expenses, lower absolute levels of expenditure on the state secretariat, district administration and police and higher agricultural growth. The link with higher expenditure on social services is also of interest, though it would help if this could be disaggregated into wage increases for social service personnel, higher establishment charges and the pure program component reaching the target groups. The apparent lack of statistical significance in the voter turnout rates of the low and high CDI states is surprising. It is possible that the general rise in voter turnouts across the states in the post 1977 period is swamping the effect due to devolution. It is also possible that a significant correlation exists between devolution and the extent of voter awareness, as measured by the percentage of invalid votes cast in an election. Conclusions This paper shows that it is possible to explain the degree of devolution largely through structural and institutional characteristics of the political and bureaucratic system in a given state. Though there may be alternative explanations for the extent of devolution, we believe that the institutional factors are of great importance. We anticipate two major criticisms. The first relates to the paucity of data points. We tried to address this by using lagged variables from the panel data given in Table 2c. However, this did not yield results that were significantly different. The data 28 It is felt that the impact of devolution is likely to be more on voter awareness, as measured by the percentage of invalid votes, rather than mere participation rates. This is a mere conjecture 15

19 shows that some states have consistently been more open to sharing power with the LBs. There have been no major year to year shifts in the CDI rankings among the states and there is a clear pattern across the states that is best explained by a combination of political and institutional factors. The second concern would relate to the absence of control variables, such as population, state domestic product, per capita state domestic product and fiscal deficit. Analysis of these variables showed that CDI rankings have little to do with economic factors. 29 We believe that the variable WTD captures the effect of the economic variables listed above. Further analysis of the variable WTD should yield the same results. Even if the empirical evidence for the impact of devolution, or lack thereof, had indicated otherwise, the reasons for greater or lesser devolution in a given state would remain unchanged. Only now, we would have to develop adequate explanations for the inability of the devolution exercise to influence government activities. It would be useful to hypothesize that CDI in some sense represents the degree of internal openness, analogous to the concept of external openness [Sachs, 1995]. A greater degree of devolution is expected to reduce transaction costs, lead to greater efficiency in the demand for and delivery of public services and to generate a greater efficiency in the process of governance. In turn, devolution also appears to have precisely the kind of linkages it is supposed to have with government expenditure, electoral participation, economic behavior and the demand for social services. It would be very surprising if all of this did not add up to give devolution a far greater importance than hitherto assigned, in determining the rates of growth, poverty alleviation and human development as well as the quality of governance in a country. 29 Authors calculation based upon data in IMF Occassional Paper #134. India Economic Performance and Growth, 1995, Tables 3.5 and

20 Table 1: Ratio of Rural Development Flows to Total State Expenditure.(percentage) Year Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal Source: CMIE, 1996, Table 230 Table 2a: Financial Autonomy(α) State Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal Source: CMIE, 1996, Table 247; Authors Calculation 17

21 Table 2b: Functional autonomy (β) State Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal Source: CMIE, 1996, Tables 230, 247; Authors Calculations Table 2c: Composite Devolution Index 1, (CDI 1) State Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal Source: CMIE, 1996, Tables 204, 211, 230; Authors calculations. 18

22 Table 3: Ethno- Linguistic Fractionalization Index Distribution of persons by major languages Total Assamese Bengali Gujrati Hindi Kannada Kashmiri Malayalam Marathi Oriya Punjabi Sindhi Tamil Telugu Urdu Others EFI Andhra Pradesh % Assam Bihar % Gujarat % Haryana % Himachal Pradesh % Karnataka % Kerala % Madhya Pradesh % Maharashtra % Orissa % Punjab % Rajasthan % Tamil Nadu Tripura % Uttar Pradesh % West Bengal % India % Source: Census of India,

23 Table 4: Diversity Index (DI) State Number of Communities Number of Traits Diversity Index (DI) Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal Authors calculations. Table 5: Willingness to devolve (WTD) Total Flows TFC norms % shortfall WTD (Normalized 30 ) (Actual) (Recommended) (- denotes excess flows) Andhra Pradesh % Assam % Bihar % Gujarat % Haryana % Himachal Pradesh % Karnataka % Kerala % Madhya Pradesh % Maharashtra % Orissa % Punjab % Rajasthan % Tamil Nadu % Tripura % Uttar Pradesh % West Bengal % Source: Oommen and Dutta, 1995, p.54 and Authors calculations. 30 {(%shortfall)-95.6*(maximum shortfall)}/95.6% 20

24 Table 6: Changes in Administration and Political Parties State # Changes in # Changes in Chief Ministers since 1977 Administration Party PARTY CHANGE Andhra Pradesh 13 5 Chenna Reddy 78, Anjaiah 80, BV Reddy 82, KVB Reddy 82, NT Rama Rao (NTR) 83, NB Rao 84, NTR 84, NTR 85, Chenna Reddy 89, J Reddy 90, Vijay Bhaskar Reddy 92, NTR 94, CB Naidu 95 Assam Chandra 78, Hazarika 79, President s Rule (PR) 79, Taimur 80, PR 81, Gogla 82, PR 82, Saikia 83, Mahanta 85, PR, Saikia 89, Bhumidhar Barman 96, Mahanta 96 Bihar 12 4 PR 77, Karpoori Thakur 77, RS Das 79, PR 80, Jagannath Misra 80, CS Singh 83, B Dubey 85, BJ Azad 88, SN Sinha 89, Jagannath Misra 89, Laloo Yadav 90, Laloo Yadav 95 Gujarat 10 5 Babubhai Patel 77, PR 80, MS Solanki 80, MS Solanki 85, AS Choudhary 85, Chiman Patel 90, Suresh Patel 93, PR, Keshubhai Patel 96, SS Vaghela 97 Haryana 11 6 Devi Lal 77, Bhajan Lal 79, Bhajan Lal 80, Bansi Lal 86, Devi Lal 87, OP Chautala 89, BD Gupta 90, OP Chautala 90, Hukum Singh 90, Bhajan 91, Bansi Lal 96 Himachal Pradesh 7 4 Shanta Ram 77, Ram Lal 80, Virbhadra Singh 83, Virbhadra Singh 85, Shanta Ram 90, PR 92, Virbhadra Singh 94 Karnataka 10 4 Gundu Rao 80, RK Hegde 83, SR Bommai 85, PR 89, V Patil 89, S Bangarappa 90, V Patil 91, V Moily 94, HD Deve Gowda 95, JH Patel 96 Kerala K Karunakaran 77, AK Anthony 77, Vasudevan Nair 78, PKM Koya 79, PR 79, EK Nayanar 80, PR 81, K Karunakaran 81, PR 82, K Karunakaran 82, EK Nayanar 87, K Karunakaran 91, EK Nayanar 95 Madhya Pradesh 11 4 K Joshi 77, VK Saklecha 78, S Patwa 80, Arjun Singh 80, ML Vora 85, Arjun Singh 88, ML Vora 89, SC Shukla 89, S Patwa 90, PR 92, Digvijay Singh 93 Maharashtra 12 4 Vasantdada Patil 77, S Pawar 78, AR Antulay 80, B Bhosle 82, Vasantdada Patil 83, S Nilangekar 85, SB Chavan 86, S Pawar 88, S Pawar 90, Sudhakar Rao Naik 91, S Pawar 93, M Joshi 95 Orissa 6 3 N Routray 77, Janaki Pattnaik 80, Janaki Pattnaik 85, H Biswal 89, Biju Pattnaik 90, Janaki Pattnaik 96 Punjab 9 6 PS Badal 77, PR 80, Darbara Singh 80, PR 83, SS Barnala 85, PR 87, Beant Singh 92, HS Brar 95, Rajinder Kaur 96 Rajasthan 11 6 PR 77, BS Shekhawat 77, PR 80, J Pahadia 80, SC Mathur 81, HL Deopura 85, HD Joshi, SC Mathur 88, BS Shekhawat 90, PR 92, BS Shekhawat 94 Tamil Nadu 10 8 PR 77, MG Ramachandran 77, PR 80, MG Ramachandran 80, Janaki Ramachandran 88, PR 88, MK Karunanidhi 89, PR 91, J Jayalalitha 91, MK Karunanidhi 96 Tripura 4 2 RR Gupta 77, N Chakrabarty 78, S Majumdar 88, D Deb 93 Uttar Pradesh 15 9 RN Yadav 77, B Dass 79, VP Singh 80, S Misra 82, ND Tiwari 84, VB Singh 85, ND Tiwari 88, MS Yadav 89, PR 89, Kalyan Singh 89, PR 92, Mayawati 93, PR 94, MS Yadav 95, PR 95, PR 96 West Bengal 1 0 Jyoti Basu 77 Source: Lahiri, Roy and Butler,