Jump to ContentJump to Main Navigation
India’s Reforms: How They Produced Inclusive Growth$

Jagdish Bhagwati and Arvind Panagariya

Print publication date: 2012

Print ISBN-13: 9780199915187

Published to Oxford Scholarship Online: May 2012

DOI: 10.1093/acprof:oso/9780199915187.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 26 February 2017

Economic Reforms and Election Outcomes

Economic Reforms and Election Outcomes

Chapter:
(p.51) Chapter 3 Economic Reforms and Election Outcomes
Source:
India’s Reforms: How They Produced Inclusive Growth
Author(s):

Poonam Gupta

Arvind Panagariya

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199915187.003.0003

Abstract and Keywords

Economists Poonam Gupta and Arvind Panagariya ask whether voters are reacting to the increased rural-urban divide by voting out governments sympathetic to reforms as critics argue or returning the governments that deliver superior economic outcomes as suggested by Jagdish Bhagwati and Panagariya. To inform this debate, the authors offer an in-depth analysis of the 2009 election. They divide the states into high-, medium-, and low-growth categories. They then ask how the candidates of the state-level incumbent parties performed in the 2009 parliamentary elections in the three categories of states. Remarkably, they find that incumbent parties in the high-growth states won 85 percent of the seats they contested, while in the medium- and low-growth states they won 52 and 40 percent of the seats contested.

Keywords:   reforms, 2009 Indian parliamentary election, growth, incumbent party

The National Democratic Alliance (NDA), which introduced wide-ranging economic reforms between 1998 and 2004, lost the 2004 Lok Sabha election.1 Consistent with the conventional wisdom at the time, the Bhartiya Janata Party (BJP), which led the NDA, had taken the view that the electorate welcomed its reforms and therefore fought the election on the now derided “India Shining” platform. Confident of a victory, it even called the elections four months earlier than the due date. Yet not only did it lose the Lok Sabha election, but its close ally Chandra Babu Naidu, who had been seen as the face of the reforms, modernization, and urban growth, also suffered a crushing defeat in the legislative assembly elections in the state of Andhra Pradesh.

Critics of economic reforms immediately sought to capitalize on the defeat of the NDA and its allies by celebrating it as a vote against the reforms. Distinguished author Salman Rushdie expressed this view in an opinion piece in the Washington Post on May 14, 2004, in these words:

The Indian battle for centrality in the debate about the country’s future has always been, to some degree, a battle between the city and the village. It is between, on the one hand, the urbanized, (p.52) industrialized India favored by both the socialist-inclined Jawaharlal Nehru and the free-market architects of the “India Shining,” the new India in which a highly successful capitalist class has transformed the heights of the economy, and, on the other hand, the agricultural, homespun India beloved of Mahatma Gandhi, the immense countryside India where three-quarters of the population still lives and which has not benefited in the slightest from the recent economic boom.

Rushdie continued,

It’s no accident that the ruling alliance lost heavily in Andhra Pradesh and in Tamil Nadu, precisely the states that wooed information technology giants such as Microsoft to set up shop, turning sleepy “second cities” such as Madras, Bangalore and Hyderabad into new-tech boom towns. That’s because while the rich got richer, the fortunes of the poor, such as the farmers of Andhra, declined year by year.

Coming from the proreform angle, Bhagwati and Panagariya (2004) argued that the defeat of the NDA reflected a fundamental change in the attitude of voters: having experienced the benefits of reforms, henceforth they would not tolerate incumbents that did not deliver rapid improvements in their fortunes. With per capita income growth of 1.5 percent per year during the 1950s through the mid-1980s, people saw no perceptible change in their lives. This fact made the electorate pessimistic insofar as their economic fortunes were concerned. Resigned that a perceptible change in the economic conditions was impossible, they repeatedly returned the ruling Indian National Congress party (the Congress) to power. The sole exception was the 1977 election following the Emergency Rule, in which the electorate handed the outgoing prime minister, Indira Gandhi, (p.53) a crushing defeat for trampling the political freedoms that it had come to cherish.

But the liberal reforms that began in an ad hoc fashion in the 1980s and became systematic in 1990s changed this ethos. Incomes began to grow at higher rates on a sustained basis and poverty began to decline. This fundamentally altered people’s aspirations: having experienced change for the better, they wanted more of it and sooner than later. And if an incumbent government would not deliver it, they would look for another one. In the words of Bhagwati and Panagariya (2004), “Perhaps, when little progress takes place all around, the centuries-old Indian fatalism takes over. But when the poor begin improving, then the ‘revolution of rising expectations’ is likely to arise. This is a direct result of the perception of real possibilities.”

The hypothesis here is that the voting behavior has undergone change over time, as voters have become more and more conscious of the potential for better life offered by economic reforms leading to faster growth and improved economic fortunes. Until the 1980s, voters saw little improvement in their economic prospects so that they voted based on other factors such as candidate characteristics, caste considerations, and specific local issues. Occasionally, issues of national importance, such as the imposition of emergency rule, ongoing war or victory in war, and sympathy waves resulting from the assassination of a popular ruler, played a decisive role as well. These special circumstances usually resulted in unusually large voter turnouts as well as more than usual victory margins. Thus the Indian voter was primarily a partisan voter except when there was a wave of nationalism or sympathy. According to the Bhagwati-Panagariya hypothesis, this trend has been broken in recent years as economic growth has become an important factor.

In this context, it is interesting to note that the 2009 national election produced an outcome opposite of the 2004 election: it returned the ruling coalition, the United Progressive Alliance (UPA), to power. (p.54) On the surface, this seemed to undercut the bias against incumbency observed in the 2004 election. But as Panagariya (2009) explained, a more nuanced analysis suggested otherwise. He noted, “The key feature of anti-incumbency in India is that it strikes with potency at the level of the state. The electorate rewards the party of a performing chief minister and punishes that of nonperforming one in not just state assembly elections but national parliamentary ones as well.” Thus, in the 2009 election, the Congress, which led the UPA and was the national incumbent, won just nine out of seventy-two seats in the states of Bihar, Orissa, and Chhattisgarh, which all had performing non-Congress governments. On the other hand, Delhi and Andhra Pradesh had performing Congress chief ministers, and the party respectively bagged seven out of seven and thirty-three out of forty-two seats in those two states. In Rajasthan, where an outgoing BJP chief minister had performed less than spectacularly, the Congress had trounced her party in the state elections held less than six months prior to the national elections. In the national election, the Congress went on to win twenty out of twenty-five seats in that state.

Against this background, the primary aim of this chapter is to test the validity of the contrasting hypotheses offered by Salman Rushdie, on the one hand, and Bhagwati and Panagariya, on the other. More broadly, we assess the role of economic performance in determining the election outcomes in India. While taking a brief look at some of the earlier elections as well, we mainly focus on the latest parliamentary elections (2009). There are two main reasons for this choice. First, shifts in the voter attitudes resulting from reforms are likely to be reflected in the most recent election, which has taken place against the backdrop of miracle-level growth of 8 to 9 percent nationally. Second, the data for the 2009 election, especially those relating to the candidate characteristics, are more readily available.

As a concomitant to the analysis of the implications of growth, we consider in detail the role the personal characteristics of the (p.55) candidates such as wealth, education, and pending criminal charges play in determining the election outcomes. We also consider the role incumbency plays at the level of the candidate as well as the party. That is to say, we analyze whether outgoing members have a higher frequency of election than other candidates and whether the candidates of incumbent parties at the center and in the states do better than those of the opposition parties. In the present chapter, we study these issues only in terms of unconditional probabilities, leaving the task of more rigorous analysis using an econometric model in a companion article (Gupta and Panagariya 2011).

Our main result may be summarized as follows. First, we find no evidence of a rural-urban split in the election outcomes predicted by the Rushdie hypothesis. If the candidates of the state ruling party win in the rural areas, they also win in the urban areas. Likewise, if they lose in the urban areas they also lose in the rural areas. Second, in the 2009 parliamentary election, on average, candidates of the incumbent party in a state had a better chance of scoring a victory if that state exhibited higher growth than those in a state exhibiting lower growth. Compared with the candidates in the lowest-growth states, the advantage to candidates of the ruling party in the state was higher at the top end of the growth distribution than in the middle. Third, on average, incumbency was helpful in winning the 2009 election. That is to say, on average, an incumbent candidate and the candidates of the ruling parties at the center and states had better chances of victory than other candidates.2 Fourth, on average, better-educated and wealthier candidates had a better chance of victory than their less well educated and less wealthy counterparts. Finally, a key source of the victory for the UPA in the 2009 election was its superior performance in the states with nonperforming governments headed by parties outside of the UPA.

The chapter is organized as follows. In the next section, we summarize the literature on elections, especially as it relates to India. In the (p.56) third section, we describe some salient features of the last three Lok Sabha elections held in 1999, 2004, and 2009. In the fourth and fifth sections, we describe the relevant characteristics of the constituencies and candidates, respectively. In the sixth section, we present our main result on the role of growth in determining the election outcome. In the seventh section, we explain the sources of the Congress gains in the 2009 election. Our main point here is that the UPA made large gains in the states that had nonperforming governments headed by non-UPA parties. In the final section we discuss our conclusions.

The Relevant Literature

Only a small body of the literature on elections focuses on the subject central to our chapter: the role of economic performance in determining the electoral outcomes. The relevant studies include Fair (1978), Lewis-Beck (1988), Powell and Whitten (1993), Alesina and Rosenthal (1989), and Brender and Drazen (2008). Among the early studies, few find a statistically significant effect of economic performance on the election outcomes, the major exception being Fair (1978), which finds a positive and significant effect of growth on the prospects of victory in the United States. A more recent and important exception is Brender and Drazen (2008), which offers an analysis of the role of growth in elections using comprehensive cross-country data and distinguishes between developed and developing countries. These authors find that while faster growth helps incumbents win elections in the developing countries, it has no such effect in the developed countries. Our main result conforms most closely to the one found by these authors though we arrive at it through a comparison of states within a single developing country, India.

A different strand of the literature on elections, developed exclusively in the context of the western democracies and employing the (p.57) principal-agent framework, focuses on the impact the desire to win elections has on the behavior of the incumbent politician. The central question here is how political incumbents manipulate the tax and expenditure policies, manage their legislative votes, and exchange political favors to maximize the chance of electoral victory.3 Given that our objective is to study the determinants of electoral outcomes rather than how the objective of electoral victory conditions political behavior, this literature is at best indirectly relevant to our work.

A final body of the relevant literature on elections examines whether incumbency by itself is an asset or liability in elections. This literature is narrowly focused on the identification of the incumbency advantage. It begins by recognizing that a higher unconditional probability of victory of an incumbent over non-incumbent may be the result of selection bias and therefore need not represent incumbency advantage per se. Conversely, a lower unconditional probability of victory of the incumbent may not represent incumbency disadvantage. Incumbents may win more frequently simply because they happen to be better candidates or have more resources to spend on campaigns. Alternatively, if incumbents lose more frequently than non-incumbents, this may be simply because they fail to keep a number of inconsistent promises made in the prior election or because they prove themselves to be inept during their terms. Therefore, the observed frequencies of losses and wins by incumbents are by themselves insufficient to infer the effect of incumbency. The most compelling approach to identifying the impact of incumbency is regression discontinuity, which tries to identify incumbents and non-incumbents who are otherwise identical in all respects and compares their probabilities of victory in election.4

In the specific Indian context, the literature on the incumbency advantage or disadvantage is relatively new. Linden (2004) uses the regression discontinuity approach and finds that prior to 1991, incumbents had enjoyed an advantage over non-incumbents. But (p.58) beginning in 1991, this relationship reversed with incumbents suffering a disadvantage. For the elections from 1991 to 1999, he estimates that on average incumbents were fourteen percentage points less likely to be elected than similar non-incumbents.5 He reaches this conclusion by comparing the probabilities of victory of candidates in an election that had barely won (incumbents) to those of the candidates who barely lost (non-incumbents) the prior election. The underlying assumption is that the candidates that just win and those that just lose an election are identical in all respects and any advantage or disadvantage to a victorious candidate (incumbent) in the following election must result from incumbency.

While Linden (2004) studies incumbency disadvantage at the level of the candidate, a number of descriptive analytic studies following the 2004 election have focused on the disadvantage arising from association with an incumbent party. In an early analysis immediately following the 2004 election, Panagariya (2004) pointed out that the election outcome exhibited a strong anti-incumbency pattern that punished the main ruling party in the state: candidates of the party in power in a state faced a higher risk of defeat than the candidates of the main opposition party in the constituencies located in that state. For example, the Telugu Desham Party (TDP), which ruled in Andhra Pradesh, overwhelmingly lost to the Congress in that state. In three other states—Rajasthan, Madhya Pradesh, and Chhattisgarh—the BJP and its allies had defeated the ruling Congress governments in the state assembly elections held in December 2003, and they also overwhelmingly defeated the latter’s candidates in these states in the Lok Sabha election of April–May 2004.

In a more comprehensive analysis, Yadav (2004) also notes that on average the state ruling parties performed poorly in the 2004 national elections in the constituencies located in their own states but with one major exception: candidates of parties that had defeated the party in power in a just-held state election also did well in the (p.59) national election. Yadav characterizes the one to two-year period between the state and national elections as the “honeymoon” period during which the candidates of the incumbent party in the state enjoy a positive advantage. Our definition of the incumbent party in a state takes into account this difference between the “entrenched” and “recent” incumbent, focusing principally on the former.

Ravishankar (2009) carries out a quantitative analysis of the prospects of victory for the incumbent candidates of the main party in power relative to the incumbent candidates of the main opposition party using the national and state election data from 1977 to 2005. Because her analysis is strictly restricted to incumbent candidates, it does not compare incumbent and non-incumbent candidates. She finds that setting aside the parties in their honeymoon period, incumbent candidates of the main party in power in both national and state elections face higher probability of loss in their re-election bids than the incumbent candidates of the main opposition party. Ravishankar (2009) also finds a cross effect flowing from party incumbency at the national level to state elections and vice versa. Once again, setting aside the parties in their honeymoon period, incumbent candidates of the main party in power at the center face a higher probability of defeat than the incumbent candidates of the main opposition party at the center. Symmetrically, incumbent candidates of a party in power in a state face a higher probability of defeat in the national election than the incumbent candidates of the main opposition party within that state.

A key shortcoming of Ravishankar (2009) is that it excludes non-incumbent candidates. If the incumbency effect is related to the party in power, there is no reason why it should not apply to non-incumbent candidates contesting the election on the incumbent party’s ticket. Our data set, though confined to the 2009 national elections, includes all candidates and therefore allows for more complete test of the incumbency effect at the level of the party in the companion paper mentioned above.

(p.60) Salient Features of the Past Three Elections

Table 3.1 reports the broad results of the elections held in 1999, 2004, and 2009. It shows that the national parties numbering six or more have won only a little more than two-thirds of the seats in each of the three elections.6 Unsurprisingly, the party winning the largest number of seats has fallen well short of the majority so that each government has been based on a multi-party coalition. Because the party with the second most seats ends up in the opposition, state parties, which together account for approximately 30 percent of the seats have come to acquire great salience.

Led by the Bhartiya Janata Party (BJP), the National Democratic Alliance (NDA) had ruled from 1999 to 2004. Counting on its popularity at the time, it called for an early election. But, the BJP suffered major losses shrinking its tally from 182 to 138 seats, whereas the Indian National Congress (INC) or simply the Congress improved its tally from 114 to 145 seats.7 Though the latter still fell well short of the 272 seats necessary to form a government, remarkably, it was successful in cobbling together a majority coalition that came to be known as the United Progressive Alliance (UPA). The UPA government successfully served its entire term until 2009. At one level, it could be argued that neither the decline in the seats held by the BJP from 182 to 138 nor the rise in the seats held by the Congress from 114 to 145 represented a major shift away from the incumbent towards the opposition. Yet, given the expectations of a clear mandate in favor of a very popular prime minister, the media uniformly described the outcome as a decisive vote against the incumbents.

Table 3.1 Broad Results of the National Elections in 1999, 2004, and 2009

Party

1999

2004

2009

National Parties

369

364

376

Indian National Congress

114

145

206

Bharatiya Janata Party

182

138

116

Bahujan Samaj Party

14

19

21

Nationalist Congress Party

9

9

Communist Party of India

4

10

4

Communist Party of India (Marxist)

33

43

16

Rashtriya Janata Dal

24

4

State Parties

158

159

146

Samajwadi Party

36

23

Janata Dal (U)

8

20

All India Trinamool Congress

2

19

Dravida Munnetra Kazhagam

16

18

Biju Janata Dal

11

14

Shiv Sena

12

11

All India Anna Dravida Munnetra Kazhagam

0

9

Telugu Desam Party

5

6

Janata Dal (S)

4

3

Other (Unrecognized) Parties

10

15

12

Independent candidates

6

5

9

Total

543

543

543

The 2009 national election was different from the 2004 election in one fundamental sense: it returned the main ruling party, the Congress, to power with a larger number of seats as well as with a larger victory margin. Beating even the most optimistic predictions, the (p.61) (p.62) Congress increased its tally yet again from 145 to an impressive 206 seats. The Marxist Communist Party suffered the worst losses shrinking from 43 to 16 seats. The BJP also declined from 138 to 116 seats. Thus between 1999 and 2009, the Congress and the BJP had more or less exchanged their positions.

Among the national parties, the Marxist Communist Party of India, the Communist party of India, and the Rashtriya Janada Dal (RJD) suffered the largest losses besides the BJP. RJD even lost its status as a national party after the elections. Among the state parties, Samajwadi Party suffered the largest losses. Those making major gains other than the Congress were the Congress ally in West Bengal, All India Trinamool Congress, and opposition parties JD (U) and AIADMK. The other myriad state parties and unrecognized parties broadly maintained their positions.

One immediate reaction to the results in the press was that incumbency had helped rather than hurt in this election, though some observers did question this conclusion. Our own view is that the picture is more nuanced with incumbency at the state government level in conjunction with economic performance playing a decisive role. We return to the study of this idea, closely related to the Bhagwati and Panagariya (2004) hypothesis discussed in the introduction, in a later section.

(p.63) Characteristics of the Constituencies

Table 3.2 shows the salient features of the constituencies, taking all constituencies in the country into account. The averages we report in the table are virtually unchanged when we limit the sample to the nineteen major states on which our subsequent analysis will be based.

A total of 8,071 candidates contested the 2009 election. Of these, as many as 3,825, or 47.4 percent, were independent, another 30 percent were affiliated with the national or regional parties and the rest belonged to the unrecognized parties. In the 2009 election, in all, 372 parties fielded one or more candidates. Party affiliations in general and with a national or state party in particular played a crucial role in determining the outcome: candidates with a party affiliation accounted for more than 98 percent of the top four candidates and for the majority of the winning candidates. 534 winning candidates out of a maximum possible of 543 had some party affiliation. Only nine winning candidates had contested as independents.

The average number of candidates per constituency was fifteen; the maximum and minimum number of candidates in any constituency were forty-three and three, respectively. Remarkably, as the latter figure indicates, there was not a single constituency with direct contest between two candidates. Countrywide, 59.4 percent of the voters turned up to vote. The maximum turnout was 90.4 percent (in Tamluk constituency in West Bengal) and the minimum 25.6 percent (in Srinagar constituency in Jammu and Kashmir). Constituencies near the higher limit were in West Bengal, followed by the northeastern and southern states (Andhra Pradesh, Kerala, Tamil Nadu). Those near the lower end were in the states of Jammu and Kashmir (J&K), Bihar, Uttar Pradesh, and Rajasthan.

On average, the winning candidates secured about 44 percent of the votes cast and the second candidate from top obtained about (p.64) 34 percent of the total votes. Sushma Swaraj of the BJP won with the highest proportionate majority in Vidisha constituency of Madhya Pradesh, claiming 78.8 percent of the total votes cast. The simple average of the percentage-point victory margins across all constituencies was 9.7, and the margins ranged from .04 to 70 percentage points.

Table 3.2 Description of the Constituencies across All States

Average

Minimum

Maximum

Number of voters

1,319,916

45,981

2,343,012

Number of candidates

15

3

43

Voter turnout (%)

59.4

25.6

90.4

Votes obtained by the top candidate (%)

43.9

21.3

78.8

Votes obtained by the second candidate (%)

34.3

8.7

48.7

Victory margin

9.7

0.04

70.1

Votes obtained by the top four candidates (%)

93.8

62.6

100

The top four candidates summing to 2,170 out of a total of 8,071 candidates accounted for the bulk of the votes polled in most constituencies. In aggregate, these candidates accounted for more than 90 percent of the total votes polled. This can be gleaned from the fact that the density of votes in Figure 3.1 is heavily concentrated in the first four candidates.

                      Economic Reforms and Election Outcomes

Figure 3.1. Percentage Votes Obtained by Candidates at Different Ranks

(p.65) Characteristics of the Candidates

We now turn to a consideration of some key characteristics of the candidates that have implications for the electoral outcomes. These relate to wealth, education, pending criminal cases distinguished by the seriousness of the charges, gender, and incumbency status. We provide the data for all candidates, the top four candidates and the victorious candidates.

Wealth

Table 3.3 provides the distribution of candidates by wealth across five different wealth categories. For each wealth category, column 4 shows the top four candidates as a proportion of the total number of (p.66) (p.67) candidates in the given wealth category. Similarly, column V shows the number of winning candidates as a proportion of the total number of candidates.

Two features of the table stand out. First, candidates from all wealth categories are able to participate in elections and make it to the list of the top four candidates; nearly half of the top four candidates come from the lowest wealth category of 5 million rupees or less. The system does seem to offer an opportunity to run for election without regard to wealth status. Second, the unconditional probability of victory rapidly rises with wealth: while three-fourths of the candidates belong to the lowest two wealth categories, only a little more than a quarter of the elected candidates come from them. Alternatively stated, 56 percent of the winning candidates possess at least INR 10 million in declared wealth (and perhaps much more in reality). The contrast is brought out most sharply by a comparison of unconditional probability of victory of a candidate in the highest wealth category (24.1 percent) to that of the lowest wealth category (.4 percent). It bears cautioning, of course, that no causal relationship between wealth and election outcome can be drawn from these data. Wealth can very well be positively correlated with other attributes defining a good candidate in the eyes of the electorate.

Table 3.3 Distribution of Contesting and Winning Candidates according to Wealth (candidates in all Indian states)

Wealth Category

Wealth (million rupees)

Number of Candidates (% of total)

Number of Top Four Candidates

Number of Candidates Winning

Probability of Being in the Top Four Candidates

Probability of Victory

I

II

III

IV: (II/I) x 100

V: (III/1) x 100

1

0–0.5

3,176 (39%)

274

14

8.6

0.4

2

0.5–5

2,835 (35%)

642

134

22.6

4.7

3

5–9

700 (8.7%)

329

89

47

12.7

4

9–50

896 (11%)

629

194

70.2

21.7

5

50-higher

464 (5.8%)

296

112

63.8

24.1

Total

All wealth categories

8,071

2,170

543

26.9

6.7

Education

Next, we consider the distribution of candidates by education level. Once again, we identify five education levels, the lowest one being no formal education and the highest one a postgraduate or higher or a technical degree. Table 3.4 reports the frequency distribution and unconditional probabilities of being in the top four and the winning candidate.

Table 3.4 Distribution of Contesting and Winning Candidates according to the Level of Education (candidates in all Indian states)

Education Category

Education Level

Number of Candidates

Number of Top Four Candidates

Number of Winners

Probability of Being in Top Four

Probability of Victory

I

II

III

IV: (II/I) x 100

V: (III/1) x 100

0

No formal education

134

5

0

3.7

0

1

Up to Class V

964

106

15

11

1.6

2

Middle or high school

2,665

495

104

18.6

3.9

3

Undergraduate

1,623

603

157

37.2

9.7

4

Postgraduate or higher or technical

1,984

875

260

44.1

13.1

Total

All education levels

7,370

2,084

536

28.3

7.3

Three features of the table are noteworthy. First, contrary to the common impression, most candidates contesting elections have some formal education. Indeed, the vast majority of those contesting (p.68) (p.69) have at least gone through the middle school. Second, while the proportion of those with an undergraduate or higher degree is approximately half among those contesting, it is more than 80 percent among those winning. Four out of every five members in the 2009 Lok Sabha boast of an undergraduate or higher degree. At the other extreme, while 134 candidates with no formal education contested elections, none actually won. Finally, the unconditional probability of getting elected consistently rises with the education level. The biggest jump takes place as we move from high school to a college degree. We remind, however, that as in the case of wealth, this fact need not reflect causation if education is correlated with other factors that make a candidate attractive to the electorate.

Criminal Cases

Perhaps the most interesting characteristic relates to criminal cases pending against the contesting and winning candidates. Table 3.5 documents the relevant data. In constructing the table, we identify five categories based on the number of pending cases against a candidate.

Two features of the table stand out. First, as many as 14 percent of the candidates have criminal cases pending against them. And a much larger percentage of elected members of Lok Sabha—30 percent—have such cases. Even if we exclude the candidates with just one case since the prospects of frivolous cases are high against those in politics, more than eighty members of the current Lok Sabha, accounting for 15 percent of total members, have two or more criminal cases pending against them. Second, somewhat disconcertingly, the within group probability of victory rises steadily with the number of pending criminal cases.

Table 3.5 Distribution of Contesting and Winning Candidates according to Criminal Cases (candidates in all Indian states)

Crime Category

Number of Criminal Cases

Number of Candidates

Number of Top Candidates

Number of Winners

Probability of Being in Top 4 Candidates

Probability of Victory

I

II

III

IV: (II/I) x 100

V: (III/1) x 100

0

0

6,894

1,578

381

22.9

5.5

1

1

627

286

76

45.6

12.1

2

2 to 4

392

212

59

54.1

15.1

3

5 to 9

92

61

16

66.3

17.4

4

〉10

44

30

11

68.2

25

Total

All Crime Categories

8,049

2,167

543

26.9

6.7

We note that closer examination leads to a more nuanced picture from the one emerging from the aggregate data shown in Table 3.5. The criminal charges against the candidates range from the (p.70) (p.71) benign such as participation in rallies declared unlawful to more serious ones such as kidnapping, extortion, and murder. To understand the true picture, we must disaggregate the data further. To economize on space, we relegate this task to an appendix available upon request from the authors. Here we simply note that the probability of a serious crime by a contestant rises with the number of criminal cases registered against him or her. Whereas only 40 percent of the candidates with one case registered against them had been accused of a serious crime, nearly 90 percent of those with ten or more pending cases had one or more serious criminal charges against them.

Detailed examination also reveals a significant degree of concentration of criminal cases against candidates by state. The state that tops the 2009 chart is Bihar, where more than a quarter of the candidates have at least one criminal case and more than 17 percent having at least one serious charge against them. Other states, which exhibit large proportions of candidates with criminal cases, include Jharkhand, Orissa, Uttar Pradesh, Gujarat, West Bengal, and Maharashtra. We have an interesting case in Kerala, where 22 percent of the candidates have criminal cases registered against them but the vast majority of them involve benign charges such as participation in demonstration. Only a third of the cases against the members from Kerala involve serious crimes.

Distribution of Candidates by Gender

Table 3.6 reports the gender distribution of contesting and victorious candidates. The data show the expected pattern. Much fewer women than men contest election. As a result, there are many more male members in the Lok Sabha. One mildly interesting feature is that the unconditional probability of a woman winning the election is higher than that of a man.

Table 3.6 Gender Composition of Contesting and Winning Candidates (Candidates in All Indian States)

Gender

Total Candidates

Top Four Candidates

Number of Winning Candidates

Probability of Being in Top Four

Probability of Victory

I

II

III

IV: (II/I) x 100

V: (III/I) x 100

Women

556

183

58

32.9

10.4

Men

7,515

1,987

485

26.4

6.5

Total

8,071

2,170

543

26.9

6.7

(p.72) Incumbents and Non-incumbents Compared

We may now compare the incumbent and non-incumbent candidates within the populations of all, top four and victorious candidates. As Table 3.7 shows, while non-incumbent candidates far outnumber incumbent ones, virtually all of the latter are among the top four. With fifteen candidates per constituency on average, it should be no surprise that even if half of the incumbents were voted out, the unconditional probability of their victory relative to non-incumbents would be very high. Therefore, losses to a large number of incumbents are quite consistent with the incumbents having a strong showing in a statistical sense. In a similar vein, even as the main parties such as the Congress and the BJP might experience a decline in their tally of seats, the statistical probability of their candidates winning would still remain very high relative to the rest of the main parties taken together.

Table 3.7 Incumbents among All Contestants, Top Four and Winners (Candidates in All Indian States)

Incumbency Status

Total Candidates

Top Four Candidates

Number of Winning Candidates

Probability of Being in Top Four

Probability of Victory

I

II

III

IV: (II/I) x 100

V: (III/I) x 100

Incumbents

387

376

184

97.2

47.5

Non-incumbents

7,684

1,794

359

23.3

4.7

Total

8,071

2,170

543

26.9

6.7

(p.73) The Profile of an “Average” Candidate

Finally, we provide the average of each characteristic across all, top four and winning candidates in Table 3.8. If we could construct a winning candidate with these average characteristics, he would be a wealthy male (with mean assets worth 59 million rupees and median assets worth 12 million rupees) in his mid-fifties with at least an undergraduate degree. He would come from one of the main political parties. There is a 30 percent chance that he would have at least one criminal case against him and a 15 percent chance that he will have two or more criminal cases against him, and a 14 percent chance that the case would involve a serious crime. There is also 34 percent probability that he had served as an MP in the previous parliament.

Table 3.8 Average of the Characteristics across Various Candidates (Candidates in All Indian States)

Characteristics

All Candidates

Top Four Candidates

Winning Candidates

Age

46

51

53

Wealth category

2.1

3

5

(Average wealth in million rupees)

-17

-41

-59

Criminal record (probability in %)

14

27

30

Serious crime (probability)

7.4

13

13.8

Member, national party (probability in %)

20

60

69

Member, state party (probability)

9

19

27

Male (probability in %)

93

92

89

Education (category)

2.6

3.1

3.2

Incumbent (probability in %)

4.8

17.3

34

(p.74) Rural-Urban Divide versus Economic Performance

Having analyzed the role of candidate characteristics, we now turn to a consideration of the key question raised at the beginning of this chapter. Are the critics right in their claim that the reforms have (p.75) exclusively benefited the urban capitalists and done precious little for the masses in the countryside thereby leading the latter to vote out governments sympathetic to growth-oriented reforms? Or is it that they are now more inclined to pay attention to the economic performance and vote out the governments that fail to deliver good economic performance, as the reform advocates argue?

That the case of the reform critics is weak is readily shown. There is now compelling evidence (including in chapters 3 and 4 of this volume) that the reforms that led to accelerated growth on a sustained basis had actually led to a significant decline in poverty in both rural and urban areas by 2004. The decline in poverty has also been across the board extending to the socially disadvantaged, Scheduled Castes, and Scheduled Tribes. Therefore, the very premise on which the critics’ argument is based is empirically invalid.

Equally important, an analysis of 2009 election results reveals that the voting pattern exhibits no urban-rural divide. In Table 3.9, we report the number of urban and rural constituencies and the rates of victory of the state incumbent party in the state as a whole and in urban and rural regions separately. We define a constituency to be urban if less than 60 percent of its population is rural. The table includes the nineteen major states on which our subsequent analysis will focus. The main point to note is that proportions of victories of the state incumbent party in the rural and urban areas closely match. For example, the incumbent party, which did well overall in Tamil Nadu and Andhra Pradesh, did well in both rural and urban constituencies. In Uttar Pradesh, the incumbent party did poorly overall and in rural and urban constituencies taken separately. In Gujarat, the results were moderate for the incumbent party overall as well as separately in rural and urban regions. If the assertion by Rushdie quoted in the introduction that the election was a contest between cities and countryside were right, we would find a divide in the vote across rural-urban boundaries.

Table 3.9. Victory Proportions of State Incumbent Party in Rural and Urban Constituencies

State

Urban Constituencies

Rural Constituencies

Incumbent Party Candidates Elected (as % of total)

Statewide

Urban Constituencies

Rural Constituencies

Andhra Pradesh

5

37

79

80

78

Bihar

2

38

80

100

79

Chhattisgarh

0

11

91

NA

91

Goa

2

0

100

100

NA

Gujarat

11

15

58

55

60

Haryana

1

9

90

100

89

Himachal Pradesh

0

4

75

NA

75

Jharkhand

3

11

67

100

56

Kerala

3

17

29

0

36

Madhya Pradesh

4

25

55

100

48

Maharashtra

18

30

54

72

43

Delhi

7

0

100

100

NA

Orissa

1

20

78

100

76

Punjab

4

9

40

0

57

Rajasthan

4

21

16

0

19

Tamil Nadu

24

15

82

77

89

Uttar Pradesh

8

72

25

13

26

Uttarakhand

0

5

0

NA

0

West Bengal

9

33

28

0

36

Note: “NA” stands for “not applicable.”

(p.76) (p.77) (p.78) The case of the critics is, thus, decisively refuted. To establish the case of the reform advocates that poor economic performance contributes negatively to the election outcome and good performance positively, we measure the economic performance in terms of the gross state somestic product (GSDP). Our approach is to divide the states into three groups of high, medium and low growth and then ask how the candidates of incumbent party in each of these groups of states do in relation to one another.

To make this approach operational, we must first identify the group of states to be included in the sample. We note at the outset that center-administered union territories do not have territory-level governments by definition. Therefore, the concept of incumbent versus non-incumbent territory-level government is redundant for them and they are automatically excluded from the sample. In addition, we exclude Jammu and Kashmir, and the seven northeastern states (including Assam) and Sikkim. These states have a very strong presence of the central government, which a priori undermines the idea of the electorate seeing the state government as being the principal entity responsible for the economic outcomes. Finally, due to multiple turnovers of the government, we are unable to determine the incumbent government in Karnataka. These exclusions limit the sample to 19 states already introduced in Table 3.9. These 19 states include 478 out of the total of 543 constituencies and 90 percent of the eligible voters. Of the included constituencies, 364 are in general categories, 78 are reserved for the Scheduled Caste candidates, and 36 are reserved for the Scheduled Tribe candidates.

We measure economic performance in terms of the average growth in the GSDP from the beginning of the fiscal year 2004–5 to the end of 2008–9.8 This period approximately coincides with the period between the May 2004 and May 2009 general elections and also defines the approximate period of the incumbent rule in the states. We rank the states in the declining order of the average growth rate (p.79) and divide them into three groups: high, medium, and low growth. In Figure 3.2, we depict the deviation of the average growth in the GSDP from that in the GDP for each of the nineteen states, with the states stacked in order of declining growth rates to accomplish this task. The dotted lines in turn divide the nineteen states into a high-growth group with seven states and low- and medium-growth groups of six states each.

                      Economic Reforms and Election Outcomes

Figure 3.2. Difference between the Average Growth Rate of State Domestic Product and the GDP Growth Rate (2004–2008)

We define the incumbent party as the main ruling party (or two main ruling parties when power is shared) in 2007 and in at least two consecutive preceding years. If there was an election for the state legislative assembly in 2008 or 2009 and the party ruling until 2007 lost this election, it was still considered the incumbent party for purposes of the national elections held in April–May 2009. The underlying logic is that the electorate would treat the party that was in government for several years prior to 2009 responsible for the policies and (p.80) performance of the state rather than the party that took over the government in the year just preceding the general election.

Armed with this classification of the states and the definition of the incumbent party, we can ask the following key question: what proportion of the candidates fielded by the state incumbent party in the Lok Sabha constituencies located in that state won the national election? The outcome is depicted in Figure 3.3. Remarkably, on average, incumbent parties in the high-growth states won 85 percent of the seats they contested. In contrast, those in medium- and low-growth states could win on average only approximately 52 and 40 percent of the seats contested, respectively. This strong relationship between growth performance and election outcomes handsomely survives every model we test econometrically in Gupta and Panagariya (2011).

                      Economic Reforms and Election Outcomes

Figure 3.3. Proportion of the Candidates of the Incumbent Party in the State Winning the National Election (sorted by the states’ growth rates)

One final issue remains: if the reward for high growth in the elections is a recent phenomenon since the exception growth of 8 to 9 percent itself took root only recently, a pattern similar to Figure 3.3 (p.81) will not obtain in the prior elections. This indeed turns out to be the case. Without reproducing the detailed calculations, we note that in the 2004 election, percentages of candidates of the incumbent parties winning were on average 11, 72, and 39, respectively, in high-, medium-, and low-growth states. The same lack of pattern existed in the 1996 elections with winning candidates of the state incumbent parties being on average 26, 56, and 25 percent of the contesting candidates in high-, medium-, and low-growth states, respectively.9

Explaining the Superior Performance of the UPA10

The 2009 parliamentary election returned the UPA to power with more seats than even the most optimistic predictions. From 145 seats in 2004, the Congress increased its tally to 206 seats. Attributing this gain to the stellar growth performance at the national level, as done by some in the Indian press, is too simplistic. After all, the Congress performance varied widely across the states in the elections. For instance, as previously noted, it won just nine out of seventy-two seats in the states of Bihar, Orissa, and Chhattisgarh. Despite the high growth they experienced, these states voted overwhelmingly against the Congress. Clearly we need a more nuanced analysis to explain the increase in the Congress and UPA tallies.

Our analysis in the previous section offers a part of the explanation: states where the Congress ruled and growth performance was superior, it was able to retain and perhaps even increase its tally. For example, in Haryana and Delhi, two of the fastest-growing states, where the Congress was the incumbent party, it won most of the parliamentary seats, increasing its tally from 15 seats in 2004 to 16 seats in 2009 out of a total of 17 seats. But this factor by itself is not sufficient to explain the rise of the Congress tally from 145 in 2004 to 206 in 2009.

(p.82) The larger part of the gains to the Congress and its allies came from states that had non-Congress governments and did not perform economically well. In these states, the state incumbent parties lost large number of seats, which the Congress and its allies were able to pick up. Table 3.10 provides the full details of the gains made by the Congress. In the states of Madhya Pradesh, Punjab, Rajasthan, and Uttar Pradesh, where the Congress was in the opposition and the growth record of the incumbent state governments turned out to be poor, it made huge gains, picking up forty-two extra seats. The Congress ally Trinamool Congress (AITC) made similar gains in West Bengal, another slow-growing state ruled by a rival party. The AITC added seventeen seats to its tally in the state, making it easier for the Congress to put together a coalition government. The Congress more or less maintained its 2004 position in most of the medium-growth states; Kerala is the major exception. In that state, the Congress added as many as thirteen states, though it also lost three between Jharkhand and Uttarkhand.

Table 3.10 Additions to the Congress Tally in 2009 over 2004 Election

High-Growth States

Medium-Growth States

Low-Growth States

Congress government

1 (Delhi, Haryana)

8 (Andhra Pradesh, Maharashtra)

Non-Congress government

2 (Bihar, Orissa, Chhattisgarh, Gujarat)

10 (Jharkhand, Kerala, Uttarakhand)

42 (Madhya Pradesh, Punjab, Rajasthan, Uttar Pradesh)

(p.83) Our analysis to date does not allow us to explain why the Congress was the major beneficiary of the losses incurred by the incumbent parties in states with non-Congress governments that performed economically poorly. Our analysis predicts the losses to the incumbent party but does explain why the voters then choose the candidates of one or the other party. There is some tentative evidence that the main opposition party is the major beneficiary, but further work is required to establish this hypothesis more definitively.

Concluding Remarks

The issue of how the changes in the economic fortunes induced by policy changes affect election outcomes has been a long-standing one. In India, some analysts have argued that the liberal economic reforms of the 1990s and beyond hurt the vast majority of the poor residing in rural India. As a result, the electorate votes out the governments that try to implement such reforms. An alternative view is that voters reward governments that adopt policies that deliver better economic outcomes as reflected in faster economic growth. In this chapter, we have provided evidence contradicting the first hypothesis and prima facie supporting the second hypothesis. In Gupta and Panagariya (2011), we carry out detailed econometric analysis, which robustly established this hypothesis.

In addition to this key result, we have also noted the role the candidate-specific characteristics might play in the election outcomes. Once again, our results in this chapter are tentative, but we give them further rigor in the companion article mentioned in the previous paragraph. We offer data showing that educated and wealthy candidates win disproportionately. Moreover, incumbents have had much higher rates of victory than non-incumbents. These results, of course, do not establish causation, since wealth and education may be (p.84) correlated with other unobserved characteristics that the voters prefer. We conclude with a word of caution. Growth itself may be correlated with several attributes that the voters value. For example, superior growth performance may be positively correlated with good governance including law and order. It may also be associated with reduced levels of poverty. Therefore, there remains scope for differences of opinion on whether the voters rewarded growth in the 2009 elections or other variables with which it might be correlated. From a policy perspective, however, we do not see this as a major issue. Even if the voters value these other variables over growth per se, growth would serve as a reasonable target variable for the state politicians to win the elections.

Notes

References

Bibliography references:

Alesina, A., and H. Rosenthal (1989). “Moderating Elections.” NBER Working Papers 3072.

Alford, J. R., and J. R. Hibbing (1981). “Increased Incumbency Advantage in the House.” Journal of Politics 43, 1042–61.

(p.86) Baron, David (1989). “Service-Induced Campaign Contributions, Incumbent Shirking, and Reelection Opportunities.” In P. C. Ordeshook, ed., Models of Strategic Choice in Politics, 93–120. Ann Arbor: University of Michigan Press.

Besley, T., and A. Case. (1995). “Does Electoral Accountability Affect Economic Policy Choices? Evidence from Gubernatorial Term Limits.” Quarterly Journal of Economics 110(3), 769–98.

Bhagwati, J., and A. Panagariya (2004). “Great Expectations.” Wall Street Journal, May 24.

Brender, A., and A. Drazen (2008). “How Do Budget Deficits and Economic Performance Affect Reelection Prospects?” American Economic Review 98(5), 2203–20.

Collie, M. P. (1981). “Incumbency, Electoral Safety, and Turnover in the House of Representatives, 1952–1976.” American Political Science Review 75, 119–31.

Erikson, Robert S. (1971). “The Advantage of Incumbency in Congressional Elections.” Polity 3, 395–405.

Fair, Ray C. (1978). “The Effect of Economic Events on Votes for President.” Review of Economics and Statistics 60(2), 159–73.

Garand, James C., and Donald A. Gross (1984). “Change in the Vote Margins for Congressional Candidates: A Specification of the Historical Trends.” American Political Science Review 78, 17–30.

Gelman, A., and G. King (1990). “Estimating Incumbency Advantage without Bias.” American Journal of Political Science 34, 1142–64.

Gupta, Poonam (2011). “For Performers, Incumbency Helps.” Financial Express, April 19.

Gupta, P., and A. Panagariya (2011). “Growth and Election Outcomes in a Developing Country.” Columbia Program on Indian Economic Policies, Working Paper no. 2011-8.

Jacobson, Gary C. (1987). “The Marginals Never Vanished: Incumbency and Competition in Elections to the U.S. House of Representatives, 1952–1982.” American Journal of Political Science 31, 126–41.

Lee, David S. (2001). “The Electoral Advantage to Incumbency and Voters’ Valuation of Politicians’ Experience: A Regression Discontinuity Analysis of Elections to the US House.” NBER Working Paper no. W8441.

Levitt, S. D. (1994). “Using Repeat Challengers to Estimate the Effect of Campaign Spending on Election Outcomes in the U.S. House.” Journal of Political Economy 102(4), 777–98.

Levitt, S. D., and J. M. Poterba (1999). “Congressional Distributive Politics and State Economic Performance.” Public Choice 99(1–2), 185–216.

Lewis-Beck, Michael S. (1988). Economics and Elections. Ann Arbor: University of Michigan Press.

Linden, L. (2004). “Are Incumbents Really Advantaged? The Preference for Non-incumbents in Indian National Elections.” Unpublished ms., Columbia University.

(p.87) Panagariya, A. (2004). “Vote against Reforms?” Economic and Political Weekly 39(21), May 22, 2079–81.

Panagariya, A. (2009). “Is Anti-incumbency Really Passé?” Economic Times, May 28.

Payne, J. L. (1980). “The Personal Electoral Advantage of House Incumbents.” American Politics Quarterly 8, 375–98.

Powell, G. B., Jr., and G. D. Whitten (1993). “A Cross-National Analysis of Economic Voting: Taking Account of the Political Context.” American Journal of Political Science, 37(2), 391–414.

Ravishankar, N. (2009). “The Cost of Ruling: Anti-incumbency in Elections.” Economic and Political Weekly 44(10), 92–98.

Rogoff, K., and A. Sibert (1988). “Elections and Macroeconomic Policy Cycles.” Review of Economic Studies 55(1), 1–16.

Rushdie, Salman (2004). “India’s New Era.” Washington Post, May 14, http://www.washingtonpost.com/wp-dyn/articles/A25770-2004May13.html

Snyder, J. M., Jr. (1990). “Campaign Contributions as Investments: The U.S. House of Representatives, 1980–1986.” Journal of Political Economy 98, 1195–227.

Uppal, Y. (2009). “The Disadvantaged Incumbents: Estimating Incumbency Effects in Indian State Legislatures.” Public Choice 138(1), 9–27.

Virmani, Arvind (2004). “Economic Growth, Governance and Voting Behavior: An Application to Indian Elections.” ICRIER Working Paper no. 138.

Yadav, Y. (2004). “The Elusive Mandate of 2004.” Economic and Political Weekly 39(51), December 18, 5383–95. (p.88)

Notes:

(*) The authors are at National Institute of Public Finance and Policy, New Delhi, and Columbia University, New York. They can be reached at pgupta@nipfp.org.in and ap2231@columbia.edu, respectively. Work on this chapter has been supported by Columbia University’s Program on Indian Economic Policies, funded by a generous grant from the John Templeton Foundation. The opinions expressed in the chapter are those of the authors and do not necessarily reflect the views of the John Templeton Foundation. The authors are indebted to the participants of Columbia-NCAER Conference on Trade, Poverty, Inequality and Democracy on March 31–April 1, 2011, for many helpful comments and to Ganesh Manjhi for his excellent research assistance.

(1) . Lok Sabha, translated as “Assembly of the People,” is the lower house of the Indian Parliament. For purposes of elections to Lok Sabha, the country is divided into 543 constituencies, principally on the basis of population, with each constituency electing one member. Elections to the upper house, called Rajya Sabha and translated as “state Assembly,” are indirect with the vast majority of its members elected by the state legislative assemblies.

(2) . This incumbency effect could be due to a variety of reasons, such as the incumbent candidates and parties having more resources to spend on election campaigns, having better name recognition, or even being more charismatic. Our results here do not separate the pure incumbency effect on which a great deal of the political science literature focuses.

(p.85) (3) . For example, Rogoff and Sibert (1988) and Alesina and Rosenthal (1989) analyze the use of fiscal and monetary policy actions, and Besley and Case (1995) analyze the use of tax-expenditure choices by incumbents to gain electoral support. Levitt and Poterba (1999) study the effect of congressional representation on state economic growth. Levitt (1994), Baron (1989), and Snyder (1990) examine the response of politicians to campaign contributions. Lee (2001) provides additional references.

(4) . An excellent example of this analysis is Lee 2001. A vast body of political science literature is devoted to the analysis of the incumbency effect in election outcomes. For example, see Erikson 1971, Collie 1981, Garand and Gross 1984, Jacobson 1987, Payne 1980, Alford and Hibbing 1981, and Gelman and King 1990.

(5) . Uppal (2009) also finds that incumbency has hurt the candidates in recent Indian elections.

(6) . India has more than one thousand registered political parties. These are divided into national parties, state parties, and unrecognized parties. Any registered party that lacks the status of state or national party is an unrecognized party. The Election Commission confers the status of state party on any party that meets certain thresholds in terms of votes received and seats won in an election. A state party acquires monopoly on the use of its party symbol in the state. A party qualifying as state party in four states gets the national status and then has the monopoly over the use of its election symbol over the entire country. It is not unusual for parties to lose the national status if they lose the qualifications for it.

(7) . Virmani (2004) offers an analysis of the voter behavior in the 2004 election.

(8) . Data in India typically relate to its fiscal year, which begins on April 1 and ends on March 31. Therefore, fiscal year 2004–5 refers to the period from April 1, 2004, to March 31, 2005.

(9) . We do not do this calculation for the 1998 and 1999 elections because they took place as a result of the fall of unstable coalition governments, which fell well before their full five-year terms ended. Because the state elections in the majority of states had taken place in 1996 or before, the identification of the incumbent parties in the states for these years is not a clean affair.

(10) . This section draws heavily on Gupta (2011).