Regulation, Ownership, and Costs: A Historical Perspective from Indian Railways

Dan Bogart Associate Professor

Department of Economics University of California, Irvine

dbogart@uci.edu

Latika Chaudhary Assistant Professor

Department of Economics Scripps College

latika.chaudhary@scrippscollege.edu

Draft: July 2010

Abstract

We study the relationship between regulatory structure, operational costs and state ownership in one of the world’s largest railway networks. Between 1874 and 1912, all the former private railway companies in India were taken over by the colonial Government at predetermined dates set by contracts negotiated in the 1850s. Exploiting changes in ownership within the same railway systems, we find the move to state ownership decreased operating costs by 14 percent. We do not find any evidence of biases associated with anticipation effects by private companies. In contrast, we argue the regulatory environment of providing 5 percent guarantees to private companies, weak oversight of private companies and a colonial state with strong fiscal incentives all contributed to improved performance under state ownership. Our findings have implications for leading issues discussed in the literature on ownership and regulation in developing countries.

Keywords: Nationalization, State Owned Enterprises, India, Railways, Institutions JEL codes: D2, D23, H54, L33, N75, O2

We thank several seminar and conference participants for comments in particular Nick Crafts, Bishnupriya Gupta, and Chris Woodruff. We also thank the Center for Global Peace and Conflict Studies for providing valuable grant money. Garrett Neiman, Shivani Pundir, Jennifer Ringoen, Nilopa Shah, and Sanjana Tandon provided able research assistance.

1

1. Introduction

Infrastructure investments are critical to economic growth and development. But, such

investments are also notoriously difficult to administer, contract and regulate especially in

developing economies (Laffont 2005). While rich countries can rely on domestic capital

markets to finance projects owned either by the public or private sector, poor countries

generally have to borrow money on international capital markets or attract foreign capital that

is often not forthcoming without an implicit subsidy or guarantee. If countries opt for the latter

option, they also have to worry about the appropriate form of regulation and ownership. The

improper design of regulatory and ownership structures are believed to be major causes of poor

performance in sectors such as gas, electricity and transportation leading to significant

economic costs on the order of 1 percent of GDP (Chisari, Estache, and Romero 1999). A

strong understanding of the pros and cons of regulatory and ownership policies are thus critical

for designing more effective policies.

Our paper provides a historical perspective by studying the transition from private to

colonial state ownership of Indian railways from 1874 to 1912. In the mid-19th century

prompted by British merchant houses and railway promoters, the British Government in

London encouraged railway development in India. It opted for a system of private British

owned and operated railways. However, the contracts stipulated a 5 percent dividend guarantee

on share capital payable by the colonial Government of India. Such guarantees were common

in other countries and were designed to compensate British investors for the risk involved in

building railways in foreign places (Eichengreen 1995).

On account of conflicts and decades of disappointing performance, the Government of

India began to construct and operate state lines in the 1870s. At the same time, they also began

2

to takeover private companies. Because of a clause in the original concession contract, the

Government could only takeover private companies on either the 25th or 50th anniversary of

their contract. The Government exercised the takeover option in every case and by 1910

formed an extensive ownership stake in the railway sector. But, this process did not eliminate

the private sector. Many companies were allowed to retain operations, but they faced more

stringent Government control and supervision.

Our empirical analysis studies whether this change in ownership and regulatory

structure influenced operating costs. We focus on costs because it is a key performance

measure in the theoretical literature on ownership and regulation. Broadly the literature

emphasizes how different objectives, institutional arrangements, and incentive structures affect

managerial efforts to reduce marginal costs (see Estache and Wren-Lewis 2009 for review).

Thus our analysis speaks to the cost of delivering infrastructure services once fixed network

investments have been made.

Using administrative reports from the Government of India, we constructed a new

historical data set on the major railway systems operating in India. The data include detailed

information on mileage, passenger and goods traffic, fuel prices, wages, costs, accidents, and

ownership status from 1874 to 1912. In terms of costs, the data distinguish between working

expenses and capital expenditures. Working expenses represents the cost of hiring train and

station staff, fuel costs, maintenance expenditures, and administration. Capital expenditure

represents the value of track miles, locomotives, vehicles, and stations.

We begin by estimating a variable cost function where the dependent variable is working

expenses for a railway system in a given year. Our right hand side includes common variables in

a cost function such as output, fixed capital, and input prices along with a dummy for state

3

ownership. The state dummy takes the value 0 under the initial regime where there was private

ownership and operation with high guarantees and weak Government oversight. It takes the

value 1 under the subsequent regime where railway systems had majority Government of India

ownership with Government operation or private operation with lower guarantees and stronger

Government oversight. We also include railway fixed effects, year fixed effects, and railway

specific trends to control for unobservable factors trending up or down that may contaminate the

estimates on state ownership. Our results show that variable costs declined under Government

of India ownership compared to private British ownership. Working expenses were 14 percent

lower on average following a move to state ownership and the accompanying regulatory

changes.

Unlike traditional cross-sectional comparisons, our institutional setting mitigates

endogeneity problems of selection bias and omitted variables. There is no selection problem

because the Government took over all the original private companies. In addition, the

contractual environment ensured the Government could not endogenously time takeovers to

coincide with periods of increasing or decreasing costs. But, private companies could perhaps

foresee organizational changes ahead of their contract deadlines and take actions to increase or

decrease variable costs in anticipation of a state takeover (i.e. anticipation effects). To

investigate such dynamics, we construct a sequence of dummy variables for intervals before and

after state takeovers. But, we find no evidence of a rise in costs either 3 or 5 years before

takeovers. We find no heterogeneous anticipation effects between early switchers where the

probability of takeover was more uncertain relative to post-1900 switchers where there was less

uncertainty about a takeover.

We also find no evidence of the Government lowering maintenance expenditures shortly

4

after the takeover to boost short-run profits while raising long-run costs through a dilapidated

capital stock. If anything, we find the big decline in working expenses occurred 11 to 15 years

after takeover and continued for 20 years. Moreover, the cost reductions did not come at the

expense of quality. We observe no change in safety measured by the frequency of accidental

injuries or deaths following takeovers.

To better understand the mechanisms underlying our findings we replicate the

regressions using subcategories of working expenses as the dependent variable. The

subcategories include working expenses in the traffic, maintenance, locomotive, vehicle, and

administrative departments separately. We find the shift to state ownership decreased working

expenses in traffic and administrative departments, which used more labor. Complementing

these results, we also observe the total number of workers, including Europeans, declined after

takeovers. Thus, it appears the Government of India reduced costs by cutting labor including

their more expensive workers.

Our work contributes to a growing empirical literature on Indian railways. Several works

document the transformative effects of Indian railways on prices, market integration and

agricultural surplus (McAlpin 1974, Hurd 1975, Adams and West 1979, Kerr 2007, Studer

2008, Donaldson 2008). Others have documented the improved performance of the railway

sector in terms of increasing cost efficiency, high productivity, and lower fares (Derbyshire

2007, Hurd 2007). But, as far as we know, no study has empirically investigated the effects of

ownership and regulation on railway performance. In this paper we analyze the effects of such

changes on variable costs. In a companion paper we study the effects on total factor productivity

(TFP) using a production function approach (Bogart and Chaudhary 2010) and find that TFP

was not impacted, indicating that ownership and regulation had a greater impact on the

5

operational margin.

Our paper also contributes to the broader literature on ownership and regulation,

especially in developing countries that lack strong institutional arrangements for regulatory

oversight and enforcement. First, the findings speak to the importance of objectives for public

and private actors. Politicians may use public enterprises to promote political objectives such as

increasing public sector employment rather than maximizing efficiency (Boycko, Shleifer, and

Vishny 1996, Shleifer 1998). Public ownership may relate to socio-political goals of increasing

social welfare in the society, which may or may not be at odds with profit maximization

(Laffont and Tirole 1991, Belsey and Ghatak 2001). In our context, the Government of India’s

non-democratic colonial status diminished any political gains from spending more on labor.

Moreover, the Government had a strong fiscal incentive to manage railways efficiently because

the associated revenues were an important component of the colonial budget. Thus, public

objectives were similar to a private firm.

Second, our findings shed light on the link between regulatory institutions and private

incentives to cut costs. Laffont (2005) develops several theoretical models where incentives for

regulated private firms are weaker in countries with limited regulatory capacity, commitment and

enforcement. Consistent with this view we find that the colonial Government offered ‘low-

powered’ incentives to private companies. Private railway companies were guaranteed a net

return of 5 percent, which undermined incentives to improve cost efficiency. Government

regulators in India were also unable to enforce cost controls on private firms. Guarantees and

weak oversight were linked to colonial institutions underpinning the Government of India.

Third, the Indian context suggests how private agency problems may sometimes tilt the

efficiency balance in favor of state over private ownership (Sappington and Stiglitz 1997).

6

Resident agents managed private railways in India, while the board of directors were situated in

London and interacted with agents via letters and telegrams. It is likely the Board or individual

shareholders were unable to adequately monitor agents and agents may have shirked or abused

their authority. Today modern communications are obviously better, but the difficulty of

monitoring foreign managers is still relevant especially in developing countries.

Finally, our paper relates to the empirical literature on ownership and firm performance.

Many recent studies often focusing on manufacturing have analyzed the effects of privatization

over the last thirty years (Megginson and Netter 2001, Djankov and Murrell 2002, and Estrin et

al. 2009.) finding strong support in favor of private ownership. We use similar econometric

methods and analyze the transition from private regulated ownership to state ownership in

Indian railways. In contrast to this literature, we find positive effects from Government of India

ownership that can be attributed to a weak regulatory environment under private ownership

coupled with a strong fiscal incentive to improve efficiency under state ownership. The Indian

case suggests the nature of regulation has important consequences for the performance of

private and state owned enterprises.

2. Private and State Ownership in the Indian Railway Context 2.1 Regulatory Structure for Private Railways

Ten private companies incorporated in Britain constructed and managed the early lines.

By 1869 there were two mergers, leaving eight major railway companies. They included the (1)

East Indian, (2) Great Indian Peninsula, (3) Eastern Bengal, (4) Bombay, Baroda and Central

India, (5) Sind, Punjab and Delhi, (6) Madras, (7) South Indian, and (8) Oudh and Rohilkhand

(see figure 1 showing the network by company in 1909). Private railways were organized as joint

stock companies through concession contracts entered into with the Secretary of State and

7

administered by the Government of India. The Secretary of State, housed in London, was a

British Cabinet member. The Government of India was headed by the Viceroy who was

appointed by the Secretary of State. In the early years the Government of India was forced to

implement the Secretary’s regulations, but in later years the Government became an agent and

advocate for state owned railways.

The concession contracts had the following general terms. Capital was raised in Britain

primarily through shares. The Government provided a 5 percent guarantee on the share capital at

a fixed exchange rate of 1s. 10d to the rupee. The net receipts were paid into the treasury and

rebated to the company. If the net receipts yielded less than the guaranteed return to

shareholders, the Government compensated the company. If they exceeded the guaranteed return,

the Government was entitled to receive half of all surplus profits and the company received the

other half. After the interest payments were paid off, the company would receive all profits.

The Government retained substantial control over private railways. It had a veto over the

placement and gauge of the lines and could influence operations through a government director

who sat on the company board. But in practice, government appointed engineers lacked

experience in construction and management (Sanyal 1930, p. 63, Kerr 2007). Railway companies

could also appeal to the Secretary when they opposed a policy change. In the 1860s, for example,

the government pressed for a merger among the private railways in southern India. The Madras

railway company refused and was successful in defeating the merger (Sanyal 1930). Business

histories of the early companies, like the East Indian Railway, show that private companies made

the final call on a wide range of financial and operational decisions (Huddleston 1906).

The Government’s right to purchase private railways after 25 or 50 years was another key

feature of the concession contract. The contracts stipulated that the Government had to announce

8

the repurchase within six months after the 25th year or the 50th year of the original contract. The

purchase price was to be based on the mean market value of the company’s stock in the

preceding three years (Bell 1894, p. 66-72). The fixed timing of the repurchase date is a critical

institutional detail for our analysis.

2.2 The Transition to Nationalized Railways in India

By the late 1870s it was still unclear whether the Government would exercise the

takeover option. Several Government owned meter gauge lines were constructed in the early

1870s signaling a move to state ownership, but the war in Afghanistan and famines slowed

construction (Bell 1894). In 1879, the Secretary took the first step by purchasing four-fifths of

the shares in the East Indian. A reconstituted company owned the remaining one-fifth of the

shares and was allowed to manage operations.

Over the 1880s, 1890s, and 1900s, the Government purchased all of the original private

railway companies. Several were bought when the 25th year of the original contract arrived.1 The

Government exercised the purchase option when the 50th year arrived for the Great Indian

Peninsula in 1900, the Bombay, Baroda and Central India in 1906, and the Madras in 1908.

There was a delay for the last three railways because they renegotiated their contract with the

Secretary of State in 1869. The Secretary cleared company debts and voided the Government’s

right to repurchase at the 25th year of the contract. In return companies had to share half of all

surplus net profits with the Government from that point forward. The same renegotiation was

offered to the East Indian but it refused. The Government opposed the renegotiation because it

forgave interest debts just as private companies were beginning to earn profits above 5 percent,

1 The right of purchase according to the original contract could first be exercised at the 25th anniversary of the contract, which was 1879 for East Indian (taken over 1880), 1883 for Eastern Bengal (taken over 1884), 1885 for Sind, Punjab and Delhi (taken over 1886), 1887 for Oudh and Rohillhand (taken over 1889), and 1889 for South Indian (taken over 1890, since this takeover was finalized in December 1890, we code 1891 as the takeover year).

9

but they had no authority to veto the agreement because it came from the Secretary.

After takeovers the Government chose to operate the railway in cases such as the Eastern

Bengal, the Sind, Punjab and Delhi, and the Oudh and Rohilkhand railways, but the precise

reasons for managing operations were different in each case. The colonial public works

department managed these lines and made personnel decisions. Surpluses were paid into the

treasury and capital was provided through annual appropriations from the Government budget.

In all other takeovers, the Government entered into agreements with directors of the former

railway companies. The newly formed companies generally held less than 20 percent of the

capital. The profits were guaranteed, at lower rates of 3 to 4 percent and at smaller capital values.

Surplus profits were shared with the Government in proportion to their respective capital shares.

Government control and supervision increased following the shift to state ownership. A

clause in the new contract gave the Government the right to overturn any company decision and

impose managerial or operational changes. The Government used their new authority to change

policy in a number of areas. Most visibly, the Government used takeovers to merge neighboring

lines to their newly acquired railway system. For example, the Government of India reorganized

the Eastern Bengal system in 1884 shortly after taking over the Eastern Bengal Railway

Company. When Sind, Punjab and Delhi was taken over in 1886, it was merged with the state-

owned Indus Valley and Punjab Northern lines. Great Indian Peninsula Railways was taken over

in 1900 and merged with the Indian Midland Railways (a Government owned and privately

operated system) in 1901. In 1908 when the contract of the Madras Railway Company came due,

the government reorganized and merged the railway lines in South India. The Government had

attempted to implement the same merger between the Madras and South Indian in the late 1860s.

Recall that the private companies refused, and hence the reorganization had to wait 40 years until

10

the Government came to own the railway.

The public-private partnership model became increasingly common after the 1880s.

Some government owned and operated railways were transferred to private operation, and new

companies were set up on similar terms. However, public opinion turned against private

operation and beginning in the 1920s the Government gradually took over all railway operations.

The move to state ownership between 1874 and 1912 was the key first step in India’s move to

complete nationalization of its railways.

2.3 Labor and Management

Before 1913 most private railways in India were managed by a board of directors in

London. The directors included the heads of British companies with interests in India, retired

members of the British military and other members of the British financial elite. The London

boards largely dealt with the finances and purchases of capital goods. To manage the railways,

the Board appointed a British agent in India. The agent was in charge of procurements including

materials, coal and labor in India, but the London boards made final decisions on materials

purchased outside India. The labor force under the agent’s control was considerable. According

to some estimates railways employed a third of the industrial labor force before Independence

(Kerr 2007, p. 86).

The labor force was generally assigned to one of five departments. The maintenance

department dealt with stations and the permanent way. The workers included foreman, masons,

carpenters, and laborers. The locomotive department handled engine operation and repair. The

engine drivers were among the most important and highest paid in this group. The carriage and

wagon department handled operation and repairs for the remainder of the rolling capital stock.

They included mechanics and some unskilled workers. Station and train staff dealing with

11

passenger and shippers generally worked in the traffic department. And, finally an administrative

department included the accountants and clerks.

The labor force was divided into three racial groups: Europeans, Anglo-Indians, and

Indians. The European labor force was primarily British. Anglo Indians were a mix of British

and Indian descent. The Indians were the most numerous representing 95 to 97 percent of the

total labor force. Europeans were around 1 to 2 percent and Anglo Indians were 2 to 3 percent.

There were no unions among the non-European workers, although the Anglo-Indians and some

caste groups apparently controlled access to certain tasks (Kerr 2007).

Europeans were the highest paid workers followed by the Anglo-Indians and then

Indians. Differences in pay were partly due to task sorting by race. Europeans tended to be

engineers, engine drivers, station masters, and accountants. Anglo Indians held some skilled

positions like engine drivers and mid-level occupations like guards (Kerr 2007, p. 83). Indians

usually performed lower skill tasks. Nonetheless, Europeans sometimes earned more for the

same task. For example, on the East Indian, European station-masters earned 150 to 300 rupees a

month while Indians earned 20 to 200 rupees (Mukherjee 1994, p. 110). The turnover for

European workers was also greater. Many moved back to Britain or other parts of the empire

after working for Indian railways.

Although skilled Indian labor was perhaps unavailable in the 1850s and 1860s, beginning

in the 1870s there were increased calls to hire more Anglo Indians and Indians in the mid to

higher paying occupations. Europeans were an expensive source of labor in India and some felt

non-Europeans could do their tasks with equal efficiency. However, the view that Indians could

replace Europeans without a substantive loss in safety or productivity was controversial (Sanyal

1930). Private companies with foreign boards and the Government of India may have diverged in

12

their approach to labor with the Government being perhaps more sympathetic to hiring Indian

workers. Our analysis evaluates the effects of ownership on both costs and labor practices.

3. Data

We created a new data set of Indian Railway systems from 1874 to 1912 for the

econometric analysis using Administration Reports on the Railways in India (Director General

of Indian Railways), The Statistical Abstracts of British India (Board of Trade), and History of

Indian Railways (Government of India 1947). We used data from the Administration Reports

published annually from 1882 in conjunction with the Report to the Secretary of State for India

in Council on Railways in India for the pre-1882 years. The latter report is less detailed, but we

were able to obtain information on variables that are essential to our regressions. We thus begin

the analysis in 1874 and end in 1912 shortly before World War 1; a thirty-eight year period

when all the former private companies were transferred to Government ownership.

Our data are extracted primarily from the tables titled “General Results of Working of

the Principal Indian Railways” reported annually before 1900 and for each half year after 1900.

The tables include capital outlay, passenger and goods earnings, gross earnings, working

expenses, fuel prices, train miles, passenger miles and ton miles (i.e. the number of tons carried

one mile).2 Working expenses measure the operational costs of railways and include the wage

bill for train staff and station staff, spending on fuel, spending on maintenance to the track,

plant, and equipment. Construction costs and the purchase of new capital goods such as

locomotives are included in the capital outlay, not working expenses. Working expenses are

disaggregated by individual departments including maintenance to track and buildings,

locomotive, carriage and wagon, traffic, and general or miscellaneous. Our analysis studies

2 Passenger miles are unreported for state owned railways from 1874 to 1879. We construct this variable by multiplying the number of passengers transported and the average trip length in 1880 for each state owned railway.

13

both the total and the detailed working expenses.3

We also extracted information on labor such as the number of Europeans, Anglo Indians

and Indians employed plus the number of injuries and deaths from train accidents. This data are

reported by railway system much like working expenses. Injuries and deaths by passengers and

employees are reported separately. We excluded injuries and deaths that were the fault of the

passenger or employee, such as suicides or terrorist actions.

To get an exogenous measure of labor costs, we used information on average monthly

wages for agricultural workers reported in the Price and Wage Reports between 1874 and

1912.4 For each railway system, we averaged the wages across the relevant regions (i.e.

provinces) traversed by the railway lines. In a few instances of missing information, we

interpolated wages using the nearest year available. Although the price and wage series are not

without problems (Datta 1915), these reports are the only available information on long run

wages for British India and have been heavily used by economic historians.

In our regression analysis, we would ideally like to follow an individual railway line for

the entire time period even after it merges with another line. We are, however, unable to code

individual lines because of two reporting problems. First, if there is a merger between two

lines, we only have data on the new merged line. Second, there is a significant reporting

change in 1900: the pre-1900 reports provide data on the “principal standard and metre gauge

lines” but after 1900 they report data on the “principal railway system” aggregating the main

company or state line with any other secondary lines worked by the same company or state.

Although the pre-1900 reports occasionally include secondary lines with primary lines, data on

3 Maintenance and locomotive expenses are reported uniformly for the entire period, but other categories are reported differently between the pre- and post-1882 reports. To construct consistent categories, we combined the general and miscellaneous categories into a single group and plotted the annual share of each category to ensure we were capturing the same categories over time. 4 As a robustness check, we also use wages for skilled workers (common mason, carpenter or blacksmith) and the average of skilled and unskilled wages. Our main results for state ownership are unchanged.

14

secondary lines are reported separately for several cases. On average, the secondary lines

account for less than 10 percent of the total mileage of the system.

To illustrate the problem, consider the case of the East Indian Railways, a private

company and the first to be taken over by the state in 1879. For the 1880s and 1890s, the East

Indian managed the operations (i.e. working) of three private company lines (Tarakessur from

1885, Delhi-Umballa-Kalka from 1891 and South Behar from 1899). Data on the latter are

reported separately before 1900, but beginning in 1900 East Indian is only reported as one

system including the three company lines.

We address the pre and post 1900 difference by creating a consistent series of the

principal railway systems from 1874 to 1912. The systems account for 90 percent of the total

mileage in India. We only focus on the standard and meter gauge lines ignoring the smaller

narrow gauge lines for which data are not consistently reported. Since data on the secondary

lines are reported separately in other tables of the pre-1900 reports, we merge their information

to the primary system managing their operation.5 We refer to the panel of principal railways as

the ‘solo’ panel because it drops secondary lines in the years before they merge. Merged lines

may have differed in unobservable ways, such as their terrain or composition of freight.

Therefore, as a robustness check, we also use a ‘joint’ panel which sums operational data, like

working expenses, output, capital, on merged lines in the years before they merge. Wages and

fuel prices for the joint system are defined by the line with the largest mileage prior to merger.

We exclude the three-major Native State owned lines from the analysis because of the

5 Fuel costs are only reported for the principal railway lines and we assume they are equal to the costs faced by the system. Other variables such as working expenses, gross earnings, unit mileage, etc. are reported separately for the main line and other secondary lines.

15

complicated relationship between Native States and the Government of India.6

We code ownership in two ways. First, the dominant railway lines within a system are

coded as either private or state owned based on whether shareholders or the Government of

India owned a majority of shares.7 Second, we also code the fraction of miles of each type

within the system such as the fraction of state-owned miles. For most railways the fraction of

state owned miles is close to 0 or 1 since the dominant railway represented a majority of the

system mileage on average.

It is important to note that our ownership variable also encompasses regulatory

structures. Privately owned railways received 5 percent guarantees and were subject to weak

government oversight. State owned and privately operated railways had lower guarantees of 3

to 4 percent and stronger government oversight. State owned and state operated railways had a

different governance structure dictated by the Department of Public Works. We also code

whether railways were private or Government operated to capture any differences in the two.

Figure 2 plots the fraction of total miles under different ownership and operational

structures for our analysis period from 1874 to 1912. In the early 1870s, privately owned

railway lines comprised over 90 percent of the system but by 1912 they accounted for less than

10 percent. As private mileage decreased, there was a parallel increase in state owned miles

that were either operated by the state (25 percent) or operated by private companies (65 percent

by 1912). Our analysis exploits this change to public ownership. Table 1 reports the means for

the main variables used in the analysis.

4. Empirical Strategy

6 This refers to the Nizam’s Guaranteed State Railway, Jodhpore-Bikaner, Udaipur-Chittoor, Bhavnagar-Gondal-Junagarh-Porbander. The only exception is Mysore state railways, which begins the period as a Native State railway but is merged with the Southern Mahratta Railway Company in 1887. 7 In a few cases, provincial governments and district boards owned public lines. We also code them as state owned.

16

We test for the effects of government ownership using the following specification:

(1)

where is the natural log of working expenses for railway system i in year t, is the natural

log of ton-miles, passenger-miles, rail miles, locomotives, and vehicles for system i in year t,

is the log price of average labor and fuel costs for system i in year t, are year fixed effects,

are railway-system fixed effects, and is the error term.

The functional form resembles the Cobb-Douglas variable cost function, with working

expenses as the dependent variable and output, fixed capital, and input prices as the independent

variables. Increases in labor and fuel costs should contribute to greater working expenses

because they imply higher input prices. Increases in ton-miles, passenger-miles, rail miles,

locomotives, and vehicles should also contribute to higher working expenses because they imply

greater output or a larger fixed capital stock to maintain.

The railway system fixed effects control for any time-invariant unobserved

heterogeneity at the railway-system level. Geography and railway gauges are two important

sources of heterogeneity. Constructing and operating railway lines in mountainous terrain prone

to land slides is perhaps more expensive than operating railways over a flat terrain. Railway

gauges are also relevant because metre gauge railways (3 feet 3/8 inches) in India had lower

costs than standard gauge railways (5 feet 6 inches). We also include year fixed effects in the

specifications to control for time-varying shocks common to all railways. Our main variable of

interest is : a dummy taking the value 1 in years when the railway system is state

owned and 0 when it is privately owned. As the specification includes railway system and year

fixed effects, we identify the effects of ownership and regulatory changes from variation within

17

the same railway system over time. If variable costs were lowered as a result then the coefficient

on would be negative and statistically significant.

To assess the effects of quality we also examine the link between accidents and

ownership. In our main specifications, we replace the variable for working expenses in (1) with a

variable for the number of deaths or the number of injuries and include controls for total mileage

along with the standard fixed effects and railway specific trends. Thus, we estimate whether

government ownership affected the safety of railways systems after controlling for their output

and other unobservable characteristics.

Railway system and year fixed effects control for a number of unobservable factors, but

there could still be railway specific time varying heterogeneity correlated with working expenses

and a switch to state ownership. If the Government of India happened to purchase private lines

when their costs were trending up or down, the coefficient on state ownership would be biased.

We therefore include individual railway time trends in the regressions, which control for

unobservable trends at the railway system level. By including such trends, we identify the effects

of state ownership by comparing deviations in working expenses from the railway trend before

and after the change to state ownership within a railway system.

Our institutional setting and estimation approach has a number of advantages in terms of

identification. First, the Government by 1908 had purchased all private railways established

before 1874. There is no selection problem of the Government taking over the best or worst

performing private railways. Second, the Government could not time the switch to state

ownership because the year in which the Government purchased private railways was determined

far in advance of the purchase date. As discussed in section 2, the initial contracts stipulated the

Government’s right to purchase the railway at 25 or 50 years from the first year of the contract.

18

Five railways were purchased on the 25th year of their initial contract. Three were purchased on

the 50th year of their contract because of the renegotiation with the Secretary in 1869. The ‘deal

of 1869’ was a unique event. There were no further renegotiations of the purchase date, in part

because of the Government’s protests to the Secretary and Parliament in 1870 (Sanyal 1930, p.

70). Thus for our purposes, the timing of Government takeovers was independent of operating

performance in the 1880s, 1890s, and 1900s.8

The predetermination of the purchase date, however, raises the potential for a different

type of identification problem: anticipation effects by private companies. Given the company

knew in advance the Government was likely to exercise the takeover option, it may have

responded by adjusting operational expenses in the years before the takeover. One possibility is

that company boards, agents, and shareholders exerted less effort in managing or monitoring

because they expected unfavorable treatment by arbitrators and the Government when their

assets were transferred. A related possibility is that private management chose to forgo long-term

opportunities that yielded profits after government takeovers. In either of these cases, working

expenses would begin to rise before the takeover, continue to rise until the year of the takeover,

and then decline.

An alternative anticipation response involves private managers and shareholders exerting

greater effort because they wanted to increase share prices. The average market value of railway

shares over the three years prior to takeover determined the purchase price. Some private

companies were also awarded an operational contract after the takeover allowing them to

continue working and profiting from the railway. Managers may have exerted greater effort prior

8 There is a concern that coefficient estimates for ton-miles, passenger-miles, track miles, locomotives, and vehicles are biased because these variables may be endogenous. However, the bias should be restricted to these coefficients only. If the timing of the transition to state ownership is exogenous after adding railway fixed effects, year fixed effects, and year specific trends, then our estimate for government ownership should be unbiased.

19

to the takeover to convince the Government of their capability in operating the railway.

We use two approaches to address anticipation effects. First, we drop and

plot residuals from our regression for private railways ten-years before and after they were

purchased by the Government and compare them with residuals for the non-switching railways.

If there are strong anticipation effects, then arguably they should be visible in the residuals

graph. Second, we explore anticipation effects in our regressions by interacting state ownership

with 5-year dummies for before the state takeover. This allows us to test whether costs increased

or decreased prior to takeovers. We use the same framework to examine the long-run response

by interacting state ownership with dummies for 0 to 5 years after the state takeover, 6 to 10

years, 11 to 15 years, 16 to 20 years and up to 20 years after takeover.

5. Results 5.1 Baseline Results

Table 2 presents our first set of results. Specifications 1 to 6 focus on the dummy variable

for state ownership, while specifications 7 to 8 use the fraction of state owned mileage within a

railway system. Given our long panel and potential concerns of serial correlation, we cluster the

standard errors at the railway system in all the regressions. Specification 1 is our parsimonious

OLS specification using the solo panel where we control for input costs, output, capital, and year

fixed effects. This regression exploits cross-sectional variation in ownership and finds that

working expenses are 19 percent lower for state owned railways as compared to private railways.

Specification 2 adds railway fixed effects to control for all time invariant unobservables

at the railway level. The effect of state ownership is still negative and statistically significant.

Specification 3 further addresses concerns of omitted variables and unobservable heterogeneity

by including railway trends. This is our preferred specification and we find large statistically

20

significant effects of state ownership—working expenses decline by 14 percent relative to trend

following a switch to state ownership. Some of our systems have longer mileage as compared to

others and we address the heterogeneity in mileage in specification 4 where we report results

from weighted regressions with total mileage serving as the weights. Specifications 5 and 6 focus

on the joint panel, which includes secondary lines in a railway system before they merge to the

main system and address concerns surrounding mergers at the time of takeovers. Our findings

are robust to these additional tests. State ownership is, thus, strongly correlated with lower

operating costs for Indian railways.

Specifications 7 and 8 use the fraction of state owned mileage within a railway system

instead of the dummy variable. A 100 percentage point increase in state owned lines reduces

operating expenses by 14 to 15 percent relative to trend, similar to the coefficient on the dummy

for state ownership. The fraction of state owned miles captures both state and native state lines

within a railway system, but our results are robust to controlling for native state mileage

separately.9

5.2 Anticipation Effects and Dynamics

Given our context, we do worry about potential anticipation effects by private companies

that could bias the coefficients on state ownership if companies drove up or down the costs in the

years before takeover. We test for anticipation effects by first plotting the residuals of working

expenses from regressions controlling for input costs, year fixed effects, railroad fixed effects

and railroad trends against years since takeover in figure 3.10 We plot residuals separately for the

railways companies taken over by the state and those that did not switch ownership in our period

9 The fraction variable only refers to a few Native State lines constructed and managed by private companies or the Government of India, and that are reported along with the data for a primary line. For example, GIPR manages the Amraoti and Khamgaon lines (14 miles total) owned by the Berar State and GIPR data always includes these lines. 10 Here we use the solo panel because anticipation effects by private companies could be mis-measured in the joint panel that artificially joins lines before they actually merged.

21

of study. For the non-switchers we use 1896, the mean year of takeovers, as the switch year.

Companies that were taken over by the state experienced significant declines in working

expenses in the years following takeover with the biggest declines in years 5 to 10 after which

costs began to increase. With regard to anticipation effects, there is no strong evidence in favor

of a ‘hump’ in the immediate years leading up to the takeover.

Similar to the residuals, the regressions in table 3 confirm that anticipation effects did not

lead to a statistically significant hump or decline in costs just before state takeovers of private

railway systems. We explore different intervals in the years leading to takeover. In specifications

1 and 2, we include dummies for year -5 to -1, year -5 to -2 and year -1 before takeover. The

omitted group in these specifications is year -6 and below before the takeover in year 0. The

coefficients on these dummies are generally small and statistically insignificant. Given the price

at purchase was based on the average stock price 3 years before takeover, perhaps a more

appropriate time interval is year -3 to -1 before takeover. But, we find no significant effects for

either the 3 years before takeover or year -6 to -4 before takeover (specification 3).

In specification 4, we exploit the uncertainty in the probability of a takeover across the

early takeovers and the late takeovers. In principle, anticipation effects should be stronger for

companies that participated in the 1869 deal, which delayed their takeover to the 50th anniversary

of their original contract. The three companies, GIPR, BBCI and South Indian, were taken over

after 1900 and given the fate of all the private companies before them, the probability of a

takeover was less uncertain for them. To assess whether these companies behaved differently in

the years before takeover, we interact the dummies for year -3 to -1 and -6 to -4 before takeover

with a dummy for the companies in the 1869 deal. Although the coefficients are positive on the

interaction terms, they are not statistically different from zero. We also dropped theses

22

companies as an additional robustness check and the coefficient on state ownership was

unchanged.

In addition to anticipation effects, table 3 also explores what happened to working

expenses five, ten, or fifteen years after the switch to state ownership. There are several

possible short-run and long-run effects implying different trajectories for residual working

expenses. On the firm side, private companies may have had incentives to report higher costs

in their accounts before takeovers than what was actually incurred on account of the

guarantees. If that were the case, we may expect a sharp decline in the years immediately

following takeover and a return to trend because the Government did not face similar

incentives to misreport their costs.

On the Government side, there are two possibilities. First, the colonial Government

may have immediately reduced track and locomotive maintenance upon takeover in order to

boost railway profits. Lower maintenance expenditures should have reduced working expenses

for a period of years, but at some point expenses might increase as the fixed and rolling capital

depreciate, train services are delayed, and accidents increase in frequency. If the effects of

lower maintenance are particularly severe, then working expenses may eventually exceed their

level at the time of the takeover. This suggests an initial decline in working expenses followed

by a sharp rise after a certain point.

Second, the state may have also taken actions to lower costs over the long-run. For

example, the Government might have introduced new technologies and improved coordination

between the former private and government-owned lines. It might have also cut high cost

workers or eliminated administrative redundancies by relying on the existing administrative

set-up for state owned railways. These cases imply a modest decrease in working expenses for

23

a period of years after the takeover, followed by a more rapid decline when the reorganization

yielded larger cost savings.

We examine these dynamics by constructing a sequence of dummy variables to capture

the evolution of working expenses several years after the state takeover of private companies.

The coefficients on the five dummies for years 0 to 5, years 6 to 10, years 11 to 15, years 16 to

20, and years 21 and above capture the evolution of working expenses after government

takeovers. Across the different specifications, working expenses declined in the first 5 years

following a takeover, but the big declines occurred 6 to 15 years after switching to state

ownership.

Table 3 also shows that costs continued to be below trend between years 16 and 20, but

the savings were lower than in previous years. Of additional interest is the coefficient on years

21 and over after takeover, which indicates whether working expenses were significantly less

than, equal to, or greater than their pre-takeover level. Although the magnitude is similar to

years 16 to 20, the coefficient on 21 and above is statistically insignificant suggesting there

were no significant cost reductions for state owned railways after 21 years. These findings are

consistent with a long run decline in costs following Government of India takeovers. Neither

short run cost cuts by the Government, nor misreporting costs by private companies appear to

be significant.

5.3. State Ownership and Safety

Our findings suggest a decline in costs following a transition to state ownership that is not

driven by firms anticipating state takeovers. But, there may be concerns that the Government

achieved these cost declines at the expense of quality. Here we use our data on accidental

injuries and deaths to both passengers and employees as a measure of quality of service to

24

examine the effects of state ownership. One could imagine that safety decreased on government

owned railways because of less maintenance or poor monitoring of workers. On the other hand,

the government might have been more concerned with safety than private companies, reducing

the frequency of accidental deaths or injuries.

We test this hypothesis by running regressions of deaths or injuries per unit of output on a

dummy for state ownership, total mileage, railway fixed effects, year fixed effects, and railway

specific trends (specifications 1,2, 4 and 5 in table 4). Output is the weighted average of

passenger and ton miles. The weights are the respective shares of passenger and goods earnings

to total earnings. We also estimate a negative binomial regression with the count of injuries or

deaths as the left hand side variable (specifications 3 and 6).

As seen in table 4, the coefficient on state ownership is small and statistically

indistinguishable from zero in all the specifications. Thus there is no evidence that safety was

sacrificed in order to lower operating expenses. There are, however, other aspects of quality, like

delays, speed, and comfort. Unfortunately, we do not observe such measures in our data, but

anecdotal evidence suggests that quality did not deteriorate along these dimensions following the

shift to Government ownership.

5.4. Detailed Working Expenses

The evidence, thus far, broadly supports the view that Government ownership of Indian

railways translated into substantial cost reductions that continued for a period of 20 years after

takeovers without any detrimental effects on quality. There is no evidence suggesting the

Government was using shortsighted strategies such as lowering maintenance or locomotive

expenses to achieve modest declines in the short run at the expense of creating long run

inefficiencies. An examination of the detailed working expenses provides further insights into

25

how state ownership achieved such cost reductions.

Table 5 reports findings on the effects of state ownership on the different categories

included under total working expenses such as maintenance, locomotive, carriage and wagon,

traffic and general plus miscellaneous grouped as one category. Traffic and general expenses are

the only two categories that are significantly different from zero. These two experienced

significant declines after state takeovers falling by 13 and 18 percent respectively. The findings

on traffic and general expenses are robust to weighting the observations, using the fraction

variable for state ownership or using the joint panel.

The traffic and general categories together accounted for one-third of total operating

expenses (traffic was 17 percent and general was 16 percent on average) and thus were important

components of operational costs. The above regressions, however, do not show that the effects of

ownership on traffic and general were different from the other categories because the coefficients

on state ownership across the different categories are not statistically different from each other.

But these findings do provide some suggestive evidence of the underlying mechanisms driving

cost reductions. Therefore, we examine the traffic and general categories further.

General and miscellaneous include a variety of expenses such as general administration,

audits, agent’s office, stores and medical departments, telegraph, law charges, taxes, etc.

Anecdotal evidence suggests state ownership reduced these expenses by eliminating

redundancies. Costs associated with accounting, audits and law charges had a significant fixed

cost component. After taking over private companies, the government could reduce such costs by

relying on their existing general and administrative setup for managing state owned lines. To

give one example, the East Indian railway operated a telegraph on one side of its railway and the

government operated a telegraph on the other side. In 1877, during negotiations for the state

26

takeover, the government agreed to repair and maintain the companies’ telegraph and the East

Indian railway administration no longer needed a staff of its own for this purpose (Huddleston

1906, p. 76).

Traffic department expenses largely included wage payments for the station and train

staff. The results thus imply that state ownership perhaps contributed to lower labor costs. This

finding may seem very surprising at first because modern forms of state ownership are often

associated with an excessive and expensive labor force due to the power of public sector unions.

Labor costs were lower following state ownership in part because employment was reduced for

all workers, especially European workers. To examine changes in employment, we ran a

regression of the log number of European, Anglo-Indian, or Indian workers on a dummy for state

ownership, total mileage, railway fixed effects, year fixed effects, and railway specific trends.

The results are reported in table 6. Employment fell significantly for all types of workers

with estimates ranging from a 25 percent decline to a 7 percent decline depending on the

specification. A substantial portion of this decline is driven by European employment. The

coefficients on European workers are negative and statistically significant in all the

specifications. State takeovers reduced European employment by as high as 25 percent

(specification 3, solo panel). While the number of Anglo Indians and Indians also declined, these

findings are not robust to the choice of panel.

The employment regressions complement the evidence from the detailed working

expenses. The Government of India reduced working expenses in part by cutting labor,

especially European workers that earned the highest salaries. These findings also speak to

contemporary debates about paucity of Indians working in higher-level railway positions. The

transition to Government ownership may have reduced the number of European workers in the

27

railway labor force, but there is no evidence of Indian workers substituting for them. The number

of Indians in higher paying positions did not substantively increase until the 1920s when the

political environment in India began to change in favor of Indians (Sanyal 1930).

6. Government of India Operation and Guarantees

The analysis thus far focuses on the bundle of ownership and regulatory changes

following government takeovers. We have not sought to disentangle the effects of changes to

majority Government ownership versus changes to Government operation or the incentive

effects of lowering or eliminating guarantees. In this section we test for the individual effects of

Government operations and guarantees. However, we are cautious in claiming to identify their

effects because changes in operations or guarantees may be endogenous.

In a majority of cases reconstituted private companies retained operations following

government takeovers. But in some cases the Government chose to operate the railways instead

as in the case of the Eastern Bengal, the Sind, Punjab, and Delhi, and the Oudh and Rohilkhand

railways. The effects of private versus government management were hotly debated in the early

twentieth century (Ghose 1927). Some Indian nationalists suggested that Government operations

would foster cost cuts. Alternatively one could argue that retaining private operations enabled

state owned railways to engage in cost cutting behavior.

In table 7 we test these theories by including a dummy variable for state operations or a

dummy for state owned and privately operated lines. Neither state operation nor private operation

is statistically significant. In these regressions, state ownership is the primary driver of reduced

costs (specifications 1 to 4). The interpretation of the operations result is complicated because it

is unclear why some railways became government operated and others not. It is possible that

operational choices were based on the performance of companies prior to takeover or

28

expectations about performance afterwards. Therefore the results provide tentative evidence that

choice of operations had little effect on working expenses.

In this table, we also try to assess the importance of guarantees. Contemporaries generally

regarded the guarantee system as ineffectual in restraining construction and operational costs.

(Bell 1894, Sanyal 1930). In fact, GOI viceroys and officials made some of the most vociferous

complaints. Subsequent historians have also written extensively about the guarantee system

weakening private incentives. According to Thorner (in Kerr, ed. 2001), “Since companies had a

guarantee of 5 per cent interest, they had little incentive to economy. Building went ahead both

carelessly and lavishly. Where works were defective, they were simply rebuilt. Stations were

erected in a grand style, scarcely suited to India’s impoverished economy” (p. 85). The

guarantees may have also worsened traditional agency problems because the foreign board of

directors sitting in London often had limited local knowledge relative to their agents in India.

And, this perhaps also made supervision more lax.

To assess the effects of guarantees, we created a dummy variable for lines that were

receiving guarantees in each of the 5 years before takeover and interacted it with state ownership

(specifications 5 to 8 in table 7). Not all lines were receiving guarantees at the time of takeovers.

It is possible therefore that the decline in working expenses was greater for railways that were

receiving guarantees because their incentives were most affected by the Government takeover,

which either lowered the guarantee to 3 or 3.5 percent or introduced Government operations. The

regressions, however, do not point to any differential effect of railways receiving guarantees. The

coefficient on the interaction between state ownership and railway companies receiving

guarantees is small in magnitude and statistically insignificant. In specifications 7 and 8, we

interact state ownership with the number of years a company received guarantees since 1874.

29

The interaction term is small and statistically insignificant. The coefficients for state ownership,

however, are negative, statistically significant and similar in magnitude to table 2.

Nonetheless, we are cautious in drawing strong conclusions that guarantees did not matter

for working expenses. Similar to operations, endogeneity is a concern because companies not

receiving guarantees such as the East Indian benefitted from their location and the goods traffic

transported on their lines. Moreover, there is substantial anecdotal evidence in the reports and

secondary sources about the ‘extravagant’ spending of private companies because of the 5

percent guarantee.

7. Discussion

We interpret our findings in the context of the literature on ownership and regulation

especially in developing countries, which emphasizes the importance of incentives and

regulatory institutions for performance (Laffont 2005, Estache and Wren-Lewis 2009). Some

researchers argue against state ownership because political actors may have different objectives

such as hiring more labor than private firms (Boycko, Shleifer and Vishnu 1996). In our context,

the political benefits of excessive spending on labor were arguably swamped by the costs of

forgone profits to the treasury.

Figure 4 graphs government railway revenues as a fraction of total government revenues

(gross) from 1874 to 1913. Although railways contributed a small share to public income in the

1870s, by 1913 almost 35 percent of its total tax revenues came from state owned railways

(calculation using data in Statistical Abstract relating to India 1915). The Government’s fiscal

reliance on railways meant it had a greater incentive to operate railways at low cost. Moreover,

the colonial nature of the Government meant it was unencumbered by labor unions demanding

higher wages or more employment. It was also largely free from local constituents demanding

30

specific services not justified on efficiency grounds. In short the Government of India had the

flexibility to make unpopular decisions. Broadly our results suggest that autocratic governments

or those that are fiscally reliant on state owned enterprises are more likely to pursue profit

maximization objectives and hence may have similar efficiency as private companies.

The strong objectives of the state in India were also coupled with low powered incentives

for private firms under the initial regulatory regime. Weak incentives for private railways were

linked to a number of institutional features of India in the 1850s and 60s. Local capital markets

were inadequate to support large infrastructure projects for most of the 19th century. The colonial

Government of India was established in 1857 following the end of East Indian Company rule. It

was unable to borrow large sums for railways on favorable terms in its early years. Partly

because of its fiscal weakness it opted for a system of private guaranteed railways in the 1850s

and 60s. On several occasions it tried to encourage companies to build railways without a 5

guarantees, but was never successful (Bell 1894, Sanyal 1930). Guarantees thus can be viewed as

a necessary evil. They were necessary to attract international capital to build Indian railways, but

once implemented they weakened incentives to reduce operational costs.

Unexpected interventions by the Secretary of State in London were also a problem. The

Secretary was more susceptible to lobbying by private companies and often resolved

disagreements in favor of the private companies. The Secretary could overturn regulatory

decisions of the colonial Government of India because he had ultimate authority on policy

matters. He could also impose decisions without consulting the Government. One example is the

renegotiation of company contracts in 1869 against the wishes of the Government of India. As

this case illustrates, the Government lacked credibility in enforcing its regulations.

Insufficient regulatory capacity was a further problem. In the early years there were

31

frequent complaints by companies that regulators lacked experience. Rightly or wrongly, they

were blamed for slow construction and hampered operations. According to Kerr (2001) strong

regulatory oversight was often lacking. “Technically, the Government of India officers deputed

to watch over the railways had the right not to sanction expenditure incurred without their

approval. The contractors and the railway companies’ men did not want to wait for their

operations. In practice, they went ahead, did things, and later told the government what they had

done” (p. 84). A foreign board of directors in London coupled with local agents in India further

exacerbated traditional agency problems, which may have also undermined efforts at economy in

the era of private ownership.

The institutional environment changed towards the end of the nineteenth century. The

Government of India’s fiscal efficiency increased as evident by its ability to borrow on terms

similar to the British government. The Secretary promised to consult with the Government on all

regulatory matters following the controversy over the 1869 renegotiation of contracts. In the

1870s the Government also established a Director of Railways to oversee regulations and created

a Providence Fund, which shared a portion of state railway profits with state railway employees.

These institutional developments were arguably linked with the strengthening of incentives

following Government takeovers. Guarantees were eliminated or lowered and firm caps on fares

and freight charges were introduced. The Government also exercised greater control and more

effectively mandated operational changes. Perhaps government oversight and regulation could

have prevented some of the early excesses associated with private guaranteed railways but they

were difficult to implement given the initial institutional environment.

8. Conclusion

In this paper, we measure the changes in operating efficiency when Indian railways

32

switched to state ownership. Using a novel data set on the principal Indian railway systems and

by comparing changes within systems, we find a switch from private to state ownership reduced

working expenses by 14 percent controlling for a variety of factors including railway specific

trends. The results are not due to anticipation effects or short run cost reductions at the expense

of long run cost increases or safety. A switch to state ownership reduced costs for a period of 20

years. After taking over private companies, the Government of India achieved cost declines by

reducing employment of all workers and in particular European workers. Moreover, these gains

did not come at the expense of quality.

We attribute our results to a combination of weak regulatory incentives, oversight and

enforcement under private ownership and strong state objectives to minimize costs under state

ownership. In the early years, government regulators lacked technical expertise and the Secretary

of State in London often resolved conflicts between the Government and private companies in

favor of the latter. When the Government took over private companies, they were finally

successful in introducing stronger incentives to improve efficiency that lead to significant cost

declines. Overall the Indian case supports the view that as regulatory capacity, accountability and

fiscal effectiveness improves, incentive structures become stronger and cost efficiency increases.

33

References

Adams, John, and Roberts Craig West. “Money, Prices, and Economic Development in India 1861-1895.” Journal of Economic History 39, no. 1 (1979): 55-68. Bell, Horace. Railway Policy of India: With Map of Indian Railway System. Rivington, Percival 1894. Besley, Timothy and Maitreesh Ghatak. “Government versus Private Ownership of Public Goods.” Quarterly Journal of Economics 116, no. 4 (2001): 1343-1372. Bogart, Dan and Latika Chaudhary. “State Ownership, Productivity and Colonialism: The Case of Indian Railways”, Working Paper, 2010. Boycko, M. A. Shleifer, and R. Vishny. “A Theory of Privatization.” Economic Journal 106, (1996): 309-319. British Board of Trade. The Statistical Abstract of British India. Years 1880-1912. Chisari, O., A. Estache, and C. Romero. “Winners and Losers from Privatization and Regulation of Utilities: Lessons from A General Equilibrium Model of Argentina.” The Work Bank Economic Review 13, (1999): 357-378. Datta, K.L. Report on the Enquiry into the Rise of Prices in India. Superintendent Government Printing, Calcutta,1915. Derbyshire, Ian. “Private and State Enterprise: Financing and Managing the Railwaysof Colonial North India, 1859-1914.” In Kerr, Ian ed. 27 Down: New Departures in Indian Railway Studies. New Delhi, Orient Longman, 2007. Djankov, Simeon and Peter Murrell. “Enterprise Restructuring in Transition: A Quantitative Survey.” Journal of Economic Literature 40, no. 3, (2002): 739-92. Donaldson, Dave. “Railways and the Raj: The Economic Impact of Transportation Infrastructure.” Working Paper, 2008. Eichengreen, Barry. “Financing Infrastructure in Developing Countries: Lessons from the Railway Age.” The World Bank Research Observer 10, no. 1 (1995): 75-91. Estache, Antonio and Liam Wren-Lewis. “Toward a Theory of Regulation for Developing Countries: Following Jean-Jacques Laffont’s Lead.” Journal of Economic Literature 47, no. 3, (2009): 729-770. Estrin, Saul, Jan Hanousek, Evzen Kocenda, and Jan Svejnar. “The Effects of Privatization and Ownership in Transition Economies.” Journal of Economic Literature 47, no. 3, (2009): 699-728.

34

Ghose, S.C. Lectures on Indian Railway Economics-Part I. Calcutta, Calcutta University Press, 1927. Government of India. Prices and Wages in India. Department of Statistics, India: Calcutta, Superintendent Government Printing, 1896 and 1922. Government of India. History of Indian Railways, Constructed and in Progress Corrected up to 31 March 1955. Ministry of Railways, 1955. ___________. History of Indian Railways, 1947. House of Commons. “Administration Report on the Railways in India” for 1882-1912. Parliamentary Papers. Huddleston, George. History of the East Indian Railway, Volume 1. Calcutta: Thacker, Spink, and Co., 1906. Hurd, John II. “Railways and the Expansion of Markets in India 1861-1921.” Explorations in Economic History 12 (1975): 263-288. ___________.“Railways” in ch. 8 of the Cambridge Economic History of India, vol. 2:1757-1970. Eds. D. Kumar and M. Desai. London: Cambridge University Press, 1983. __________. “A Huge Railway System But No Sustained Economic Development: The Company Perspective, 1884-1939: Some Hypotheses.” In Kerr, Ian ed. 27 Down: New Departures in Indian Railway Studies. New Delhi, Orient Longman, 2007. Kerr, Ian. Engines of Change: The Railroads that Made India. Westport: Praeger, 2007. Kerr, Ian (ed.). Railways in Modern India. New Delhi: Oxford University Press, 2001. Laffont, Jean-Jacques. “Regulation and Development.” New York: Cambridge University Press, 2005. Laffont, Jean-Jacques and Jean Tirole. “Privatization and Incentives.” Journal of Law, Economics, and Organization 7, (1991): 84-105. McAlpin, Michelle Burge. “Railroads, Prices and Peasant Rationality: India, 1860-1900.” Journal of Economic History 34, no. 3 (1974): 662-84. Megginson, William and Jeffrey M. Netter. “From State to Market: A Survey of Empirical Studies on Privatization.” Journal of Economic Literature 39, no. 2, (2001): 321-89. Ramamurti, Ravi. “Testing the Limits of Privatization: Argentine Railroads.” World Development 25 (1997): 1973-93.

35

Sanyal, Nalinksha. Development of Indian Railways. Calcutta: University of Calcutta, 1930. Sappington, David and Joseph Stiglitz. “Privatization, Information, and Incentives.” Journal of Policy Analysis and Management 6, no. 4 (1987): 567-82. Shleifer, Andrei. “State versus Private Ownership.” Journal of Economic Perspectives 12, no. 4 (1998):133-150. Statistical abstract relating to British India from 1903-04 to 1912-13. London: His Majesty's Stationary Office. Thorner, Daniel. “The Pattern of Railway Development in India.” Far Eastern Quarterly, XIV (1955): 201-6. ___________. Investment in Empire: British Railways and Steam Shipping Enterprise in India 1825-1949. Philadelphia: University of Pennsylvania Press, 1977.

36

Figure 1: Map of Indian Railways in 1909

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1874

1876

1878

1880

1882

1884

1886

1888

1890

1892

1894

1896

1898

1900

1902

1904

1906

1908

1910

1912

Figure 2: Indian Railways by Organization

Fraction State

Fraction State-Private

Fraction Private

-0.1

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

-11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Figure 3: Working Expenses Residuals Against Years Since Switch

Non Switchers

Switches to State Ownership

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

1868 1873 1878 1883 1888 1893 1898 1903 1908 1913

Figure 4: Railway Revenues as Percentage of Total Government Revenues

1874-1912 1880 1890 1900 1910

Working Expenses 9,439,038 3,642,254 5,551,550 8,869,492 17,900,000Net Earnings/Capital Outlay 5.1% 4.3% 4.9% 5.5% 5.4%

Avg Fuel Costs / Ton 11 13 9 11 11Avg Agricultural Wages 5.9 4.6 4.8 6.8 7.8

Total miles 1244 639 1,002 1,320 1,836Locomotives 278 170 245 283 473Vehicles 6,227 3,581 5,289 6,273 11,184

Passenger Ton Miles 439,000,000 208,000,000 320,000,000 416,000,000 908,000,000Goods Ton Miles 399,000,000 148,000,000 240,000,000 401,000,000 839,000,000

Injuries 6 4 3 5 5Deaths 2 1 1 1 1

Source: See text for details. Summary statistics are for solo panel only. Working expenses, fuel and labor costs are expressed in real 1900 rupees.

Table 1: Summary Averages by Year

(1) (2) (3) (4) (5) (6) (7) (8)

State Ownership -0.1941*** -0.0852* -0.1405*** -0.1398*** -0.0957*** -0.1081***[0.0484] [0.0404] [0.0252] [0.0327] [0.0222] [0.0235]

Fraction of State Owned Miles -0.1548*** -0.1365***(in logs) [0.0334] [0.0443]

Avg Cost of Fuel 0.0710** 0.0505 0.0727** 0.0301 0.0641** 0.0222 0.0761** 0.0329[0.0295] [0.0344] [0.0264] [0.0273] [0.0271] [0.0240] [0.0283] [0.0299]

Avg Wage 0.2349*** 0.0072 0.0419 0.1257 0.0334 0.1463 0.0413 0.1218[0.0481] [0.1021] [0.0867] [0.0890] [0.0889] [0.0955] [0.0893] [0.0908]

Passenger Miles 0.0948* 0.1206* 0.1859*** 0.1136** 0.2456*** 0.1322** 0.1871*** 0.1114*[0.0503] [0.0613] [0.0589] [0.0518] [0.0551] [0.0564] [0.0583] [0.0524]

Ton Miles 0.3990*** 0.2994*** 0.2532*** 0.2466*** 0.2418*** 0.2252*** 0.2621*** 0.2540***[0.0448] [0.0436] [0.0474] [0.0283] [0.0470] [0.0251] [0.0502] [0.0321]

Total Mileage 0.079 0.3004*** 0.3125*** 0.2423*** 0.2630*** 0.2831*** 0.3121*** 0.2215**[0.0911] [0.0672] [0.0451] [0.0718] [0.0596] [0.0736] [0.0476] [0.0794]

Locomotives 0.3990*** 0.1513 0.1693* 0.1215 0.2258*** 0.1354** 0.1762** 0.1323[0.0997] [0.1131] [0.0809] [0.0858] [0.0708] [0.0589] [0.0789] [0.0894]

Vehicles -0.057 0.0666 0.0657 0.2257** 0.0206 0.1941*** 0.057 0.2209**[0.0757] [0.1016] [0.1117] [0.0880] [0.1009] [0.0630] [0.1072] [0.0918]

Year FE Yes Yes Yes Yes Yes Yes Yes YesRailroad FE No Yes Yes Yes Yes Yes Yes YesRailroad Trend No No Yes Yes Yes Yes Yes YesWeights No No No Yes No Yes No YesPanel Solo Solo Solo Solo Joint Joint Solo Solo

Observations 513 513 513 513 518 518 513 513R-squared 0.982 0.99 0.993 0.991 0.993 0.991 0.993 0.991

Robust standard errors clustered at the railway system level in brackets. *** p<0.01, ** p<0.05, * p<0.1

Table 2: Effect of State Ownership on Log of Working Expenese

(1) (2) (3) (4)

State Ownership*Years -5 to -1 before takeover 0.0013[0.0268]

State Ownership*Years - 5 to -2 before takeover 0.0261[0.0496]

State Ownership*Year - 1 before takeover -0.0023[0.0591]

State Ownership*Year -3 to -1 before takeover 0.0045 -0.0093[0.0300] [0.0313]

State Ownership*Year -6 to - 4 before takeover 0.0212 0.011[0.0278] [0.0311]

Deal 1869*Year -3 to -1 before takeover 0.0574[0.0802]

Deal 1869*Year -6 to - 4 before takeover 0.0406[0.0276]

State Ownership*Years 0-5 after takeover -0.0992** -0.0872* -0.0880* -0.0842*[0.0429] [0.0465] [0.0454] [0.0447]

State Ownership*Years 6-10 after takeover -0.1649*** -0.1511*** -0.1534*** -0.1493***[0.0445] [0.0500] [0.0476] [0.0458]

State Ownership*Years 11-15 after takeover -0.2096*** -0.1946*** -0.1987*** -0.1922***[0.0561] [0.0503] [0.0569] [0.0572]

State Ownership*Years 16-20 after takeover -0.1167* -0.1028* -0.1079* -0.0999[0.0606] [0.0509] [0.0603] [0.0614]

State Ownership*Years 21 and over after takeover -0.101 -0.0842 -0.0915 -0.081[0.0820] [0.0754] [0.0829] [0.0833]

Weights No Yes No YesPanel Solo Solo Solo Solo

Observations 513 513 513 513R-squared 0.993 0.992 0.993 0.992

Table 3: Evolution of Working Expenses

Robust standard errors clustered at the railway system level in brackets. *** p<0.01, ** p<0.05, * p<0.1All regressions include fuel and labor costs, passenger miles, ton miles, locomotives, engines and total mileage, as well as railway system FE, yar FE and railway trends.

(1) (2) (3) (4) (5) (6)

State Ownership 0.0008 0.0005 -0.2884 0.002 0.001 0.0205[0.0009] [0.0006] [0.5701] [0.0017] [0.0007] [0.4982]

Weights No Yes Neg Bio No Yes Neg BioPanel Solo Solo Solo Solo Solo Solo

Observations 521 521 522 521 521 522

State Ownership 0.0001 -0.0001 -0.4246 0.0016 0.0007 -0.0263[0.0004] [0.0004] [0.4429] [0.0015] [0.0008] [0.3939]

Weights No Yes Neg Bio No Yes Neg BioPanel Joint Joint Joint Joint Joint Joint

Observations 530 530 531 530 530 531

*** p<0.01, ** p<0.05, * p<0.1 Robust standard errors clustered at the railway system level in brackets

InjuriesDeaths

Table 4: Did Quality Decline following State Takeovers?

All regressions include total mileage, year FE, railroad FE and railroad trends. In columns 1, 2, 4 and 5, deaths and injuries are per 100,000 of output.

Maintenance Locomotive Carriage Traffic General & Misc Traffic General &

Misc(1) (2) (3) (4) (5) (6) (7)

State Ownership -0.1474 -0.0685 -0.1319 -0.1292* -0.1842* -0.1686** -0.2101**[0.0911] [0.0648] [0.0864] [0.0640] [0.1000] [0.0598] [0.0959]

(in logs)Avg Cost of Fuel 0.0735 0.0444 -0.0301 -0.026 0.0248 -0.0779*** 0.0002

[0.0928] [0.0407] [0.0669] [0.0291] [0.0425] [0.0231] [0.0478]Avg Wage -0.0378 -0.0677 0.0432 -0.1173 0.051 0.0304 0.1887

[0.1805] [0.1487] [0.1544] [0.1265] [0.1544] [0.1190] [0.1631]Passenger Miles 0.036 0.1018 0.078 0.1353 0.1831* 0.1583** 0.1938**

[0.0730] [0.0621] [0.0997] [0.0779] [0.0961] [0.0642] [0.0842]Ton Miles 0.2396*** 0.2947*** 0.3741*** 0.1362*** 0.2338*** 0.1455*** 0.2053***

[0.0727] [0.0573] [0.0641] [0.0398] [0.0636] [0.0392] [0.0549]Total Mileage 0.3661*** 0.2168** 0.1469 0.1474* 0.4547*** 0.0401 0.2768*

[0.1080] [0.0781] [0.1578] [0.0720] [0.1184] [0.0665] [0.1349]Locomotives 0.1108 0.2276* 0.2067 0.0784 -0.0902 0.0685 -0.1908

[0.1850] [0.1270] [0.2414] [0.0990] [0.1741] [0.0647] [0.1452]Vehicles 0.1476 -0.0169 0.0377 0.2532 0.1465 0.3532** 0.3259**

[0.2263] [0.1437] [0.1931] [0.1508] [0.1683] [0.1273] [0.1478]

Weights No No No No No Yes YesPanel Solo Solo Solo Solo Solo Solo Solo

Observations 511 512 512 512 512 512 512R-squared 0.952 0.984 0.968 0.985 0.97 0.988 0.966

Table 5: Detailed Working Expenses (in logs)

Robust standard errors clustered at the railway system level in brackets. *** p<0.01, ** p<0.05, * p<0.1

(1) (2) (3) (4) (5) (6) (7) (8)

State Ownership -0.247** -0.153*** -0.246*** -0.222*** -0.1726* -0.1784* -0.254** -0.154***[0.084] [0.035] [0.060] [0.057] [0.0975] [0.0901] [0.086] [0.034]

Log Total Mileage 0.850*** 0.769*** 0.606*** 0.616*** 0.8486*** 0.7841*** 0.863*** 0.776***[0.090] [0.073] [0.052] [0.068] [0.1413] [0.1588] [0.093] [0.073]

Weights No Yes No Yes No Yes No YesPanel Solo Solo Solo Solo Solo Solo Solo Solo

Observations 519 519 519 519 519 519 519 519

State Ownership -0.152** -0.073 -0.233*** -0.197*** -0.100 -0.1109 -0.155** -0.071[0.069] [0.048] [0.052] [0.049] [0.0766] [0.0682] [0.071] [0.049]

Log Total Mileage 0.765*** 0.702*** 0.616*** 0.589*** 0.7991*** 0.7326*** 0.774*** 0.708***[0.058] [0.068] [0.055] [0.071] [0.1217] [0.1665] [0.061] [0.067]

Weights No Yes No Yes No Yes No YesPanel Joint Joint Joint Joint Joint Joint Joint Joint

Observations 527 527 527 527 527 527 527 527

*** p<0.01, ** p<0.05, * p<0.1 Robust standard errors clustered at the railway system level in bracketsAll regressions include year FE, railroad FE and railroad trends.

Anglo Indians

Table 6: Employment and State Ownership

Total Employment Europeans Native Indians(in logs)

(1) (2) (3) (4) (5) (6) (7) (8)State Ownership -0.1428*** -0.1469*** -0.1534*** -0.1317*** -0.110** -0.101 -0.137*** -0.139**

[0.0354] [0.0376] [0.0293] [0.0407] [0.037] [0.059] [0.038] [0.052]State Operation 0.0045 0.0285

[0.0475] [0.0381]State Ownership*Private Operation 0.0201 -0.0112

[0.0397] [0.0386]State Ownership*Receiving Guarantee -0.053 -0.062

[0.051] [0.079]State Ownership*Number of Years 0.000 0.000Receiving Guarantee [0.002] [0.002]

(in logs)Avg Cost of Fuel 0.0748** 0.0309 0.0749** 0.0304 0.086*** 0.039 0.076** 0.031

[0.0288] [0.0284] [0.0292] [0.0286] [0.024] [0.026] [0.028] [0.030]Avg Wage 0.041 0.1282 0.0431 0.1267 0.041 0.122 0.041 0.126

[0.0860] [0.0874] [0.0889] [0.0889] [0.087] [0.093] [0.089] [0.092]Passenger Miles 0.1857*** 0.1184** 0.1846*** 0.1154** 0.185*** 0.107** 0.185*** 0.113**

[0.0583] [0.0465] [0.0594] [0.0454] [0.058] [0.050] [0.059] [0.050]Ton Miles 0.2548*** 0.2512*** 0.2516*** 0.2488*** 0.256*** 0.247*** 0.254*** 0.247***

[0.0439] [0.0256] [0.0456] [0.0263] [0.048] [0.029] [0.047] [0.028]Total Mileage 0.3166*** 0.2432*** 0.3158*** 0.2437*** 0.310*** 0.250*** 0.317*** 0.244***

[0.0443] [0.0745] [0.0435] [0.0729] [0.046] [0.070] [0.044] [0.071]Locomotives 0.1687* 0.1098 0.1719* 0.1166 0.176** 0.129 0.170* 0.121

[0.0824] [0.0915] [0.0805] [0.0919] [0.078] [0.087] [0.081] [0.084]Vehicles 0.0622 0.2268** 0.0609 0.2255** 0.062 0.226** 0.062 0.224**

[0.1111] [0.0897] [0.1100] [0.0890] [0.110] [0.086] [0.111] [0.087]

Weights No Yes No Yes No Yes No YesPanel Solo Solo Solo Solo Solo Solo Solo Solo

Observations 513 513 513 513 513 513 513 513

Table 7: State Operations and Guarantees

Robust standard errors clustered at the railway system level in brackets. *** p<0.01, ** p<0.05, * p<0.1