Jump to ContentJump to Main Navigation
The Political Economy of Hunger: Volume 3: Endemic Hunger$

Jean Drèze and Amartya Sen

Print publication date: 1991

Print ISBN-13: 9780198286370

Published to Oxford Scholarship Online: January 2008

DOI: 10.1093/acprof:oso/9780198286370.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (oxford.universitypressscholarship.com). (c) Copyright Oxford University Press, 2020. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use. date: 26 September 2020

2 Public Policy and Basic Needs Provision: Intervention and Achievement in Sri Lank

2 Public Policy and Basic Needs Provision: Intervention and Achievement in Sri Lank

Chapter:
(p.59) 2 Public Policy and Basic Needs Provision: Intervention and Achievement in Sri Lank*
Source:
The Political Economy of Hunger: Volume 3: Endemic Hunger
Author(s):

Sudhir Anand

S. M. Ravi Kanbur

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780198286370.003.0003

Abstract and Keywords

A substantial part of Sri Lanka's achievements in certain areas of basic needs provision such as health and education standards, has been due to the country's intrinsic and directed public policies. This chapter's econometric analysis, based on time series data, reconfirms that income growth alone would not have achieved that enviable basic needs record — the role of direct intervention has been significant. The expansion of health services has been more effective than food subsidies in mortality decline. There is a need to shift the focus from scrutinizing the effectiveness of intervention to looking at the best patterns and combinations of social welfare expenditure that can achieve the maximum impact on basic needs.

Keywords:   government intervention, health, education, social welfare, intervention-achievement link, time-series analysis

2.1. Introduction

In academic and policy discussions of development strategy, Sri Lanka has become a test case. It is well known that the country has exceptionally high achievements in the areas of health and education. The life expectancy at birth of a Sri Lankan is almost 70 years, which is a figure approaching that found in industrial market economies, and much higher than that typical of developing countries at similar or even considerably higher levels of per capita income. Infant mortality rates in Sri Lanka are below 40 per 1,000 live births, which compares with figures in excess of 100 for most countries at similar levels of per capita income. Literacy rates are 80 per cent or more, compared with the developing countries' average of around 50 per cent.

Assuming that one of the major objectives of development is to enhance the quality of life along the dimensions of health, education, and other basic needs, Sri Lanka appears to have been remarkably successful. Yet the growth of its per capita income has been modest in comparison with other developing countries, and it remains part of the ‘low‐income’ group of developing countries. The remarkable record in achievement is attributed by some to a systematic and sustained policy of government intervention in the areas of health, education, and food over a long period. The counter to this position comes in several forms, which can perhaps best be summarized in the statement that the intervention was, or has become, ‘excessive’ relative to the achievement. It should be clear, therefore, why Sri Lanka is seen as a test case—a verdict on whether or not intervention was excessive in that country will have implications for other countries deciding on levels of intervention. The intensity of debate on Sri Lanka as a test case has been further heightened by the fact that in 1977 the government explicitly introduced changes which were seen as a retreat on intervention. The post‐1977 experience in Sri Lanka is thus also of great importance to the ‘intervention and achievement’ debate.

The object of this chapter is to examine the role of public policy in basic (p.60) needs provision in Sri Lanka. How much does Sri Lanka's enviable record owe to direct intervention by the state? The next section provides an overview of the historical development of public policy and intervention in the areas of health, education, and food, taking in both the pre‐independence and post‐independence periods. It also considers long‐term trends in Sri Lanka's health and education indicators, and recent developments in indicators related to food intake. Section 2.3 moves from the descriptive to the econometric method. We assess the recent literature on linking intervention and achievement in the area of basic needs, which is based on evidence from a cross‐section of countries. After a critique of some aspects of this literature, we move to a direct analysis of intervention and achievement using time‐series evidence for Sri Lanka for the post‐independence period. Section 2.4 concludes the chapter.

2.2. Intervention and achievement: historical overview

Sri Lanka has a long record of government intervention in the field of social welfare. The record includes intervention in health, education, and other social services, as well as in the area of food subsidies. In most areas intervention predates independence in 1948; in some areas it even predates the granting of universal adult franchise in 1931. The origins of intervention in Sri Lanka lie in legislation to regulate the living conditions of Indian immigrant workers on the estates, with the pressure for these regulations coming from the (colonial) Government of India. Tinker (1974) has analysed the role of the Government of India in regulating the process of indentured labour migration to Ceylon and to other areas of the British Empire, including the West Indies, Mauritius, and Malaya. He documents the conflicting concerns of the government in balancing the requirements of employers for cheap labour, its own perception that such migration was a useful way of easing population pressure in India, and its self‐perceived duty to protect its citizens' conditions of passage and work. Revelations of the appalling treatment of indentured labour en route to estates and plantations led to the establishment of inspectorates at ports of origin and destination.

Further requirements, e.g. on working conditions, were often seen as an intrusion into the internal affairs of the colonies concerned, and Tinker gives a fascinating account of how the India Office and the Colonial Office in London mediated these conflicts. However, the demand for labour was such that the employers usually gave way. It would be beyond the scope of this chapter to document in detail this intricate bargaining process in the case of labour migration to Ceylon. But Orde Browne (1943: 20–1) summarizes the process well:

Conditions of service had begun to receive the attention of the Legislature far back in the last century (the Ordinance on Contracts for Hire and Service dates from 1866), but (p.61) the main impetus to improve living conditions came from the gradually increasing requirements of the Government of India. (The Ordinance relating to Estate Labour (Indian) was passed in 1889 and amended in 1890, 1909, 1921, 1927, 1932, and 1941.) Early improvements were chiefly connected with housing and medical attention, but the standard and scope of requirements grew, until employers were called upon to provide hospitals, schools, maternity arrangements, creches, and various other amenities, representing, as a whole, considerable responsibility and expense…

While the employer's responsibility for the welfare of his work‐people may be fully admitted, there is something anomalous about an arrangement whereby the supervision and management of such institutions as a school or a hospital must be undertaken by an estate manager…Consequently, the existing waste of money, effort, and efficiency in maintaining numerous small schools and hospitals could largely be eliminated by grouping these around central institutions, which would admit a higher standard of inspection and supervision by the appropriate Government Officers.

Orde Browne thus hints at the next logical stage in the process, a shift in responsibility from the employers to the Government of Ceylon—a transition in which the employers themselves had no small interest and which had in any case already begun, as we shall see. The main point of interest for us is that the provision of ‘basic needs’ goes back a long way in Sri Lanka and began with immigrant workers on the estates, i.e. among workers who are today the least well off in terms of satisfaction of basic needs. While the process of indentured labour migration may only be a part of the explanation (for instance it cannot explain the different developments in the former colonies during the post‐independence period), it nevertheless provides a backdrop to the consistent intervention of government in basic needs provision in the modern period.

In what follows we will examine in greater detail the historical record of intervention and achievement under the headings of health, education, and food subsidies.

(a) Health

Table 2.1 provides some indicators for the health sector from 1926 onwards. As Alailima (1985) notes, the date 1926 is significant for the fact that the first Health Unit in Sri Lanka was established in this year, providing primary health care, including control of infectious diseases.1 Before this date the network of hospitals was in the main restricted to the estate sector and the urban areas. In 1926 a total of 98 hospitals in Sri Lanka served the entire population, implying a figure of around 50,000 persons per hospital, and 605 persons per bed. There were 285 doctors for the then population of 4.9 million. The formal training of doctors had started in 1870 and that of nurses in 1878. By 1926 there were 437 nurses—one per 11,213 persons.

Table 2.1  Selected statistics for the health sector in Sri Lanka, 1926–1984

Year

Total no. of hospitalsa

Persons per hospital

Persons per bed

No. of doctors

Persons per doctor

Total no. of nurses

Persons per nurse

1926

98

50,000

605

285

17,193

437

11,213

1927

98

51,020

599

308

16,234

499

10,020

1928

107

47,664

592

321

15,888

545

9,358

1929

108

48,148

575

337

15,302

569

9,139

1930

112

46,429

549

341

15,249

605

8,595

1931

112

47,321

562

341

15,543

612

8,660

1932

113

47,788

566

339

15,929

611

8,838

1933

112

48,214

572

338

15,976

613

8,809

1934

111

50,451

600

333

16,817

613

9,135

1935

112

50,000

471

339

16,519

618

9,061

1936

112

50,000

478

342

16,374

632

8,861

1937

114

50,000

473

367

15,531

688

7,884

1938

115

50,435

564

366

15,847

723

8,022

1939

120

49,167

482

403

14,640

730

8,083

1940

126

47,619

490

404

14,851

744

8,065

1941

129

48,062

507

426

14,554

791

7,838

1942

132

46,970

507

455

13,626

848

7,311

1943

134

47,015

515

450

14,000

895

7,039

1944

141

45,390

514

459

13,943

n/a

n/a

1945

153

43,137

425

514

12,938

769

8,583

1946

189

35,979

408

559

12,261

n/a

n/a

1947

223

31,390

385

n/a

n/a

n/a

n/a

1948

246

29,268

381

689

10,334

948

7,595

1949

256

28,906

376

643

11,594

1,124

6,584

1950

263

29,278

385

674

11,392

1,165

6,609

1951

266

29,699

388

752

10,529

1,762

4,484

1952

268

30,224

388

768

10,486

1,729

4,685

1953

268

30,970

388

773

10,724

1,938

4,283

1954

270

31,555

368

814

10,447

2,105

4,038

1955

274

31,835

358

952

9,163

2,210

3,936

1956

278

32,118

350

984

9,074

2,304

3,863

1957

279

32,849

347

947

9,678

2,587

3,556

1958

282

33,290

341

1,128

8,323

2,767

3,397

1959

283

34,010

345

1,172

8,212

3,129

3,068

1960

289

34,242

332

1,173

8,436

3,232

3,063

1961

291

34,941

340

1,236

8,226

3,547

2,848

1962

292

35,763

318

1,345

7,764

3,270

3,180

1963

295

36,088

329

1,436

7,413

4,420

2,398

1964

294

37,085

329

1,454

7,498

3,435

3,086

1965

296

37,716

330

1,494

7,472

3,642

2,993

1966

297

38,515

342

1,512

7,565

3,499

3,258

1967

298

39,271

328

1,598

7,323

3,999

2,926

1968

302

39,708

332

1,613

7,434

4,382

2,738

1969

310

39,522

332

1,841

6,655

4,734

2,577

1970

455

27,503

320

1,932

6,477

5,542

2,256

1971

450

27,617

320

1,983

8,840

5,003

2,518

1972

457

28,490

325

2,038

6,388

4,955

2,603

1973

456

29,055

329

2,089

6,342

6,234

2,101

1974

451

29,067

332

2,127

6,295

5,288

2,496

1975

458

29,506

332

2,138

6,356

5,653

2,388

1976

460

29,848

334

2,248

6,108

5,640

2,429

1977

469

29,723

340

2,168

6,429

6,266

2,234

1978

484

29,326

352

2,258

6,286

6,169

2,286

1979

480

29,961

342

2,263

6,394

6,673

2,173

1980

480

30,704

340

2,051

7,186

6,227

2,361

1981

488

30,713

340

2,233

6,712

6,805

2,204

1982

479

31,710

350

2,036

7,460

6,931

2,193

1983

483

31,917

350

1,951

7,902

7,114

2,165

1984

484

32,229

n/a

2,822

5,528

7,216

2,162

Notes: aIncludes maternity homes.

n/a denotes not available.

Source: Rasaputra (1986: Statistical Appendix, Table 11).

While we have figures for nominal government expenditure on health from (p.62) (p.63) 1926 onwards (Rasaputra 1986: Statistical Appendix, Table 11), the problem is that we do not have an adequate deflator to express these in real terms before 1951. Intertemporal comparisons and trend calculations are therefore not possible between 1926 and 1950 in so far as real expenditure is concerned. However, an indication that this must have risen is provided by the large increase in the number of hospitals (from 98 to 263), doctors (from 285 to 674), and nurses (from 437 to 1,165). Sri Lanka's population was also rising over this period, of course, but the population per hospital bed, for example, fell from 605 to 385, while the population per nurse fell from 11,213 to 6,609.

A major feature of the 1926–50 period was the campaign against malaria. In terms of expenditure this was reflected in a substantial increase in nominal per capita expenditure on health after the mid‐1930s—from Rs. 3.49 in 1935 to Rs. 8.34 in 1950 (Rasaputra 1986: Statistical Appendix, Table 11). While these nominal figures do not allow for price changes, they do reflect institutional evidence of the concerted effort to combat malaria after the 1935 epidemic, when deaths from malaria were responsible for 23 per cent of all deaths. The system of health units was also expanded, and in the later 1940s DDT was used. The results were dramatic. The mortality rate from malaria fell from 187.3 per (p.64) 100,000 of the population in 1946 to 66.1 in 1947, and then to 32.8 in 1949 and 20.6 in 1953.2 The morbidity rate similarly fell from 41,200 per 100,000 of the population in 1946 to 19,600 in 1947,9,900 in 1949, and 5,800 in 1951 (Rasaputra 1986: Table 22, p. 62). By 1960 malaria had become insignificant as a cause of morbidity and mortality. The campaign against tuberculosis also showed success. From a level of 62 per 100,000 in 1940, the mortality rate from this disease fell to 57 in 1948; by 1960 it had reached a figure of 16 (Rasaputra 1986: 63). These specific campaigns had a major effect on overall mortality rates. As Table 2.2 shows, the crude death rate fell from 22.9 per 1,000 in 1934 to 21.8 in 1936 (during the malaria epidemic of 1935 it reached a high of 36.6). By 1946–50 it averaged 14.6, and in the late 1950s it had fallen to below 10.

Table 2.2  Population, crude birth and death rates, infant mortality rate, and life expectancy, Sri Lanka, 1900–1984

Year

Population (mid‐year) (m.)

Crude birth rate (per 1,000)

Crude death rate (per 1,000)

Rate of natural increase (%)

Infant mortality rate (per 1,000 live births)

Life expectancy at birth

Male

Female

1900

3.9

38.6

28.7

1.0

178

n/a

n/a

1901

4.0

37.5

27.6

1.0

170

n/a

n/a

1902

4.1

39.1

27.5

1.2

173

n/a

n/a

1903

4.1

40.0

25.9

1.4

164

n/a

n/a

1904

4.2

38.6

24.9

1.4

174

n/a

n/a

1905

4.3

38.6

27.7

1.1

176

n/a

n/a

1906

4.4

36.5

35.1

0.1

198

n/a

n/a

1907

4.5

33.6

30.7

0.3

186

n/a

n/a

1908

4.5

40.8

30.1

1.1

183

n/a

n/a

1909

4.5

37.5

31.0

0.7

202

n/a

n/a

1910

4.6

39.0

27.3

1.2

176

n/a

n/a

1911

4.7

38.0

34.8

0.3

218

n/a

n/a

1912

4.8

33.3

32.4

0.1

215

n/a

n/a

1913

4.8

38.6

28.4

1.0

189

n/a

n/a

1914

4.8

38.2

32.2

0.6

213

n/a

n/a

1915

4.9

37.0

25.2

1.2

171

n/a

n/a

1916

5.0

39.0

26.8

1.2

184

n/a

n/a

1917

5.0

40.1

24.7

1.5

174

n/a

n/a

1918

5.1

39.2

31.9

0.7

188

n/a

n/a

1919

5.2

36.0

37.6

−0.2

223

n/a

n/a

1920

5.2

36.5

29.6

0.7

182

n/a

n/a

1921

5.3

40.7

31.1

1.0

192

n/a

n/a

1922

5.4

39.1

27.6

1.2

188

n/a

n/a

1923

5.4

38.7

30.3

0.8

212

n/a

n/a

1924

5.4

37.5

25.8

1.2

186

n/a

n/a

1925

5.5

39.9

24.3

1.6

172

n/a

n/a

1926

4.9

42.0

25.3

1.8

174

n/a

n/a

1927

5.0

41.0

22.6

1.8

160

n/a

n/a

1928

5.1

41.9

26.0

1.6

177

n/a

n/a

1929

5.2

38.3

26.1

1.2

187

n/a

n/a

1930

5.2

39.0

25.4

1.4

175

n/a

n/a

1931

5.3

37.4

22.1

1.5

158

n/a

n/a

1932

5.4

37.0

20.5

1.7

162

n/a

n/a

1933

5.4

38.6

21.2

1.7

157

n/a

n/a

1934

5.6

37.2

22.9

1.4

173

n/a

n/a

1935

5.6

34.4

36.6

−0.2

263

n/a

n/a

1936

5.6

34.1

21.8

1.2

166

n/a

n/a

1937

5.7

37.8

21.7

1.6

158

n/a

n/a

1938

5.8

35.9

20.2

1.6

161

n/a

n/a

1939

5.9

36.0

20.9

1.5

166

n/a

n/a

1940

6.0

35.8

19.9

1.6

149

n/a

n/a

1941

6.2

36.5

18.3

1.8

129

n/a

n/a

1942

6.2

36.7

18.5

1.8

120

n/a

n/a

1943

6.3

40.6

21.3

1.9

132

n/a

n/a

1944

6.4

37.1

21.0

1.6

135

n/a

n/a

1945

6.6

36.7

21.4

1.5

139

47.2

42.5

1946

6.8

38.4

19.6

1.9

141

43.8

41.5

1947

7.0

39.4

13.8

2.6

101

52.7

51.0

1948

7.2

40.6

12.7

2.8

92

54.9

53.0

1949

7.4

39.0

12.6

2.6

87

56.1

54.8

1950

7.7

39.6

12.6

2.7

82

56.4

54.8

1951

7.9

39.8

12.9

2.7

82

56.1

54.0

1952

8.1

38.8

12.0

2.7

78

57.6

55.5

1953

8.3

38.7

10.1

2.9

71

58.8

57.5

1954

8.5

35.7

10.4

2.5

72

60.3

59.4

1955

8.7

37.3

11.0

2.6

71

58.1

57.1

1956

8.9

36.3

9.8

2.7

67

59.9

58.7

1957

9.2

36.4

10.1

2.6

68

59.1

57.9

1958

9.4

35.8

9.7

2.6

64

59.8

58.8

1959

9.6

37.0

9.1

2.8

58

60.9

60.1

1960

9.9

36.6

8.6

2.8

57

61.9

61.4

1961

10.1

35.8

8.0

2.8

52

63.0

62.4

1962

10.4

35.5

8.5

2.8

53

61.9

61.4

1963

10.6

34.1

8.5

2.6

56

62.8

63.0

1964

10.6

33.2

8.7

2.5

55

63.0

63.6

1965

10.9

33.1

8.2

2.5

53

63.7

65.0

1966

11.4

32.3

8.3

2.4

54

63.6

65.0

1967

11.7

31.6

7.5

2.4

48

64.8

66.9

1968

12.0

32.0

7.8

2.4

50

64.0

66.8

1969

12.2

30.4

8.0

2.2

53

n/a

n/a

1970

12.5

29.4

7.5

2.2

47

n/a

n/a

1971

12.6

32.7

8.2

2.4

45

64.0

66.9

1972

12.9

30.0

8.1

2.2

46

n/a

n/a

1973

13.1

28.0

8.7

1.9

46

n/a

n/a

1974

13.2

27.5

9.0

1.9

51

n/a

n/a

1975

13.5

27.8

8.5

1.9

45

n/a

n/a

1976

13.7

27.8

7.8

2.0

44

n/a

n/a

1977

14.0

27.9

7.4

2.1

42

n/a

n/a

1978

14.1

28.5

6.6

2.2

37

67.1

71.2

1979

14.5

28.7

6.5

2.2

38

67.2

71.2

1980

14.7

27.6

6.1

2.2

34

67.0

71.2

1981

15.0

28.0

6.0

2.2

29.5

n/a

n/a

1982

15.2

26.8

6.1

2.1

n/a

n/a

n/a

1983

15.4

26.2

6.1

2.0

n/a

n/a

n/a

1984

15.6

24.8

6.5

1.8

n/a

n/a

n/a

Source: Rasaputra (1986: Statistical Appendix, Table 10).

Thus despite certain shortcomings in the data, primarily the lack of a series on real health expenditures, the link between intervention and achievement can be documented with some confidence for the two decades before independence. From 1951 onwards, we can use the GDP deflator to calculate year‐to‐year movements in real health expenditure, and indeed real values of other categories of social expenditure. Table 2.3, based on Alailima (1985: Table 8), shows real per capita expenditure from 1950/1 to 1982 on different categories of social services—health, education, food, and other. For the same period, Table 2.4 expresses these categories of expenditure, as well as total social expenditure, as a percentage of GNP. These figures provide a picture of the evolution of the level and pattern of social expenditure in Sri Lanka in the 1950s, 1960s, and 1970s. We turn now to a detailed examination of the record on health over these three decades.

Table 2.3  Real GDP per capita and real public expenditure per capita on social services, Sri Lanka, 1950/1–1982 (Rs. at 1959 prices)

Year

GDP per capita

Education expenditure per capita

Health expenditure per capita

Food subsidy expenditure per capita

Other social welfare expenditure per capita

1950/1

617.59

14.60

8.80

n/a

2.05

1951/2

629.63

16.47

10.51

31.38

2.30

1952/3

619.40

16.96

10.78

15.73

2.54

1953/4

623.65

15.96

10.84

1.45

2.34

1954/5

648.85

16.51

10.67

4.23

2.32

1955/6

634.94

17.84

11.40

9.50

2.60

1956/7

622.72

19.88

11.87

11.76

2.70

1957/8

619.15

22.09

12.90

12.22

3.26

1958/9

617.71

24.84

14.72

15.55

3.55

1959/60

641.41

24.14

14.28

20.25

3.81

1960/1

646.34

29.64

15.86

25.57

4.17

1961/2

649.81

30.31

14.56

23.56

4.57

1962/3

655.75

31.44

14.77

23.99

4.31

1963/4

697.83

32.27

13.82

34.28

4.17

1964/5

694.04

33.38

13.98

25.01

4.04

1965/6

688.95

32.49

14.45

24.44

4.04

1966/7

705.56

30.53

14.38

16.13

3.65

1967/8

744.75

30.13

14.74

22.23

3.41

1968/9

767.87

32.26

15.99

23.15

3.16

1969/70

786.16

35.53

16.10

21.96

3.10

1970/1

780.63

32.42

16.16

34.19

3.08

1971/2

786.43

33.67

15.62

31.42

3.79

1973

802.60

31.03

13.80

34.25

2.35

1974

825.15

24.44

11.63

35.72

2.03

1975

883.11

25.89

12.63

44.18

3.33

1976

842.48

28.66

14.28

31.05

7.29

1977

860.07

25.17

13.45

39.51

4.47

1978

924.96

25.25

15.42

54.46

3.51

1979

956.14

29.78

16.54

62.29

0.62

1980

997.82

32.29

15.84

34.38

2.92

1981

1,034.60

30.93

14.08

24.31

2.77

1982

1,073.03

33.97

15.54

20.72

5.02

Note: Before 1973 the financial year covered the period from 1 Oct. to 30 Sept. With effect from 1973 the financial year was changed to coincide with the calendar year. All the estimates in this table refer to a period of 12 months. (This has been accomplished by multiplying the expenditure figures for the 15‐month period from 1 Oct. 1971 to 31 Dec. 1972 by the factor 12/15.)

Sources: The GDP per capita figures are calculated from Rasaputra (1986: Statistical Appendix, Table 1), which gives GDP at 1959 factor costs, and the mid‐year population estimates in our Table 2.2. The social expenditure figures are taken from Alailima (1985: Table 8).

Table 2.4  Social expenditure as a percentage of GNP, Sri Lanka, 1950/1–1982

Year

Education

Health

Food subsidies

Other social welfare

Total social expenditure

1950/1

2.5

1.5

n/a

0.3

n/a

1951/2

3.0

1.9

5.3

0.4

10.6

1952/3

3.1

2.0

2.8

0.5

8.4

1953/4

2.9

1.9

0.3

0.4

5.5

1954/5

2.7

1.8

0.8

0.4

5.7

1955/6

3.2

2.0

1.5

0.5

7.2

1956/7

3.5

2.1

2.1

0.5

8.2

1957/8

3.8

2.2

2.2

0.6

8.8

1958/9

4.1

2.4

2.6

0.6

9.7

1959/60

3.8

2.2

3.1

0.6

9.7

1960/1

4.7

2.5

3.9

0.7

11.8

1961/2

4.7

2.2

3.5

0.7

11.1

1962/3

4.7

2.2

5.1

0.6

11.0

1963/4

4.8

2.0

5.1

0.6

12.5

1964/5

4.9

2.1

3.6

0.6

11.2

1965/6

4.7

2.1

3.6

0.6

11.0

1966/7

4.6

2.2

2.4

0.6

9.8

1967/8

4.1

2.0

3.0

0.5

9.6

1968/9

4.3

2.1

3.1

0.4

9.9

1969/70

4.6

2.1

2.8

0.4

9.9

1970/1

4.3

2.1

4.5

0.4

11.3

1971/2a

4.4

2.6

4.1

0.5

11.6

1973

3.5

1.5

3.8

0.3

9.1

1974

2.8

1.3

4.0

0.2

9.3

1975

2.8

1.4

4.8

0.4

9.4

1976

3.1

1.6

3.4

0.8

8.9

1977

2.7

1.4

4.1

0.5

8.7

1978

2.7

1.5

5.3

0.3

9.8

1979

2.7

1.5

5.7

0.1

10.0

1980

2.9

1.4

3.1

0.3

7.7

1981

2.7

1.2

2.1

0.2

6.2

1982

2.9

1.3

1.8

0.4

6.4

Notes: GNP for 1951–6 obtained from National Accounts of the Department of Census and Statistics. GNP for 1956 onwards obtained from Central Bank Annual Reports.

Expenditure figures obtained from Treasury Estimates.

(a) Estimated from 15 months' expenditure and GNP figures.

Source: Alailima (1985: Table 7).

After 1950 real social expenditure per capita on health continued to increase from a figure of Rs. 8.80 in 1950/1 to Rs. 15.86 in 1960/1 (except for very small declines between 1953/4 and 1954/5, and between 1958/9 and 1959/60). After that a decline set in, and real expenditure did not overtake the 1960/1 level until the turn of the decade, when it reached just over Rs. 16. There was again a sharp decline till the trough of 1974 (Rs. 11.63) and then a rise to the all‐time peak of Rs. 16.54 in 1979. However, while the 1979 figure is the highest per capita real expenditure on health in the three decades between 1950/1 and 1982, the figures in Table 2.4 indicate that expenditure on health has not risen as fast as GNP. From the mid‐1950s to the end of the 1960s, health expenditure was equal to or exceeded 2 per cent of GNP. After this it hovered around 1.5 per cent in the mid‐1970s and fell to 1.3 per cent in 1982.

It is particularly interesting to note that in the ‘post‐liberalization’ years after 1977, real health expenditure per capita did not fall. If anything, it increased relative to the immediate pre‐liberalization years. Average real health expenditure per capita for the five years 1973–7 was Rs. 13.16 while the average for the next five years 1978–82 was Rs. 15.48—some 17 per cent higher. Similarly, as a fraction of GNP, health expenditure averaged 1.44 per cent during 1973–7, and fell only to 1.38 per cent during 1978–82. Whatever the implications of the post‐1977 period of adjustment for other items of social expenditure, it does seem as though health expenditure was protected in real terms.

(p.65) We have already noted the dramatic improvements in health‐related indicators during the quarter‐century before 1950. These improvements were consolidated during the next three decades. Between 1952 and 1981 the death rate declined at an average annual rate of 1.60 per cent (see Table 2.5). In fact, the bulk of the improvement appears to have come in the early part of this period, which suggests a link with the rapid increases in health expenditure in the 1950s. If we consider just the subperiod 1952–60, the rate of decline is 3.12 per cent per annum (not shown in Table 2.5), compared with 1.60 per cent per annum for the full period. Similarly, the infant mortality rate (IMR) declined at an average annual rate of 3.57 per cent in the earlier period (not shown in Table 2.5), compared with a rate of decline over the whole period of 2.50 per cent per annum.

Table 2.5  Average annual rate of change in infant mortality rate, death rate, and birth rate, Sri Lanka, 1952–1981

Indicator

Annual rate of change 1960–78(%)

Annual rate of change 1952–81 (%)

Infant mortality rate (IMR)

−1.72

−2.50

Death rate (DR)

−0.55

−1.60

Birth rate (BR)

−1.58

−1.28

Note: For each time period, the annual rates of change reported in this table have been estimated by means of a semi‐logarithmic regression of the variable in question on time. The data on IMR, DR, and BR are from Table 2.2.

To suggest further the link between intervention and achievement, we note that these declines in mortality rates were underpinned by a steady improvement in the indicator of persons per doctor, which stood at 10,486 in 1952 and at 6,712 in 1981 (see Table 2.1). While the ratio has fluctuated somewhat from year to year, this long‐run fall of 35 per cent over three decades must indicate a significant improvement in the provision of medical services to the population, especially when added to the fact that the number of persons per nurse fell from 4,685 to 2,204 over this period (an improvement of more than 50 per cent). A Ministry of Plan Implementation report (1985: 21) gives some indication of the quality of treatment when it notes that deaths in government hospitals fell from 229.1 per 100,000 population in 1965 to 174.1 in 1983. Of course, such figures have to be treated with caution as indicators of quality, since the deaths could instead have been occurring outside government hospitals if the system was contracting; but we know that the system was not contracting. The same report gives figures for hospital morbidity: the cases discharged from government hospitals per 100,000 population increased from 14,773.9 in 1965 to 15,471.9 in 1983. Again, this has to be seen in the context of an expanding service, where it can be interpreted as reaching out to more people, rather than as increased morbidity per se.

We end, however, with a note of caution. While the achievements in health at the all‐island level have been remarkable, the picture is by no means uniform. Although the crude death rate and the infant mortality rate have come down for Sri Lanka as a whole, the rates are much higher for the estates than for the rest of the island. Thus, in 1970, the estate sector mortality rate was 11.9 per 1,000 while the all‐island rate was 7.5 (Gunatilleke 1985: Table 4). In 1980, the all‐island death rate had fallen to 6.1 while the estate sector rate was still 11.4. Similarly, in the estate sector, infant mortality rates persisted at above 100 per 1,000 live births throughout the 1950s and 1960s, and were about 85 in the 1970s; for Sri Lanka as a whole they were already below 50 in the 1970s, and down to 38 by 1979 (see Table 2.2 and Gunatilleke 1985: Table 4). An interesting cross‐district regression analysis of infant mortality rates in 1953, 1963, and 1971 is reported in Fernando (1985: Table 9). Fernando (p.66) (p.67) (p.68) (p.69) (p.70) (1985: 83) also reports the results of another study relating to 1971 which shows that 83 per cent of the interdistrict variation in IMR is accounted for by the following variables: the proportion of the district population that is Indian Tamil, the proportion of employed females, and the proportion of females aged 15–19 with more than five years of education. The last of these highlights another factor which may be important in determining the course of IMR—education—and it is to this factor that we now turn.

(b) Education

As with health, government intervention in education in Sri Lanka predates independence. Education became the responsibility of central government with Education Ordinance No. 1 in 1920. This was the culmination of a process which began with the disclosure in the Census Report of 1901 that only 218,479 children out of a total of 867,103 between the ages of 5 and 14 years were actually receiving formal education (Alailima 1985: 6). The period after 1920 saw a steady increase of government responsibility in the field of education, and a corresponding increase in expenditure. In 1926, the expenditure on education was only 0.5 per cent of GNP; thereafter, the ratio rose to 1.5 per cent in 1946 following the adoption of free education (calculated from Rasaputra 1986: Statistical Appendix, Tables 1 and 7).

Selected educational indicators for Sri Lanka are provided in Table 2.6. Between 1926 and 1950 the number of pupils in the country rose from 10.1 per cent of the population to 17.7 per cent (calculated from Tables 2.2 and 2.6). Of course, these figures have to be interpreted with care—there are other reasons why this number may increase than an expansion of education to the previously uncovered population. However, the institutional details of the development of education in Sri Lanka corroborate these figures. As Alailima (1985: 8) notes: (p.71)

In 1931 universal adult suffrage was granted and under the new Constitution there was provision for an elected Minister of Education. For the first time a man who was from the people and knew their problems was put in charge of the education of their children. The electorate was transformed from a restricted literate and property owning minority and the Minister and his Executive Committee of elected representatives had to be responsive to their needs. This change had an immediate effect on the sphere of education. After the General Elections of 1931 and 1936 (the first conducted under the new Constitution) the state assumed much greater responsibility for the provision of education. Enrollment in government schools increased from 216,067 (39% of total enrollment) in 1931 to 378, 861 (44%) in 1945 and the number of government schools almost doubled from 1,341 to 2,391 over this period … Due to the inability of some private schools to pay their teachers during the depression, the state also took on the direct payment of these teachers.

Table 2.6  Number of schools, teachers, and pupils, Sri Lanka, 1926–1984

Year

Total no. of schools

Total no. of teachers

Total no. of pupils

Pupil–teacher ratio

1926

4,523

16,606

494,004

29.7

1927

4,512

17,787

515,221

29.0

1928

4,741

19,162

532,894

27.8

1929

4,941

18,571

562,550

30.3

1930

5,219

17,934

578,999

32.3

1931

5,304

18,242

593,437

32.5

1932

5,183

17,947

613,210

34.2

1933

5,145

18,131

631,122

34.8

1934

5,327

18,516

653,509

35.3

1935

5,351

19,243

717,287

37.3

1936

5,749

20,019

726,502

36.3

1937

6,029

20,553

783,905

38.1

1938

6,151

20,628

802,853

38.9

1939

6,100

21,570

828,090

38.4

1940

n/a

n/a

n/a

n/a

1941

n/a

n/a

n/a

n/a

1942

5,746

22,163

606,051

27.3

1943

5,568

22,698

611,529

26.9

1944

5,686

24,308

833,670

34.3

1945

5,726

25,281

867,309

33.9

1946

5,945

27,693

944,508

34.1

1947

6,097

28,977

1,036,134

35.8

1948

6,409

33,668

1,192,423

35.4

1949

6,447

35,084

1,260,667

35.9

1950

6,487

39,256

1,366,742

34.8

1951

6,708

42,558

1,454,773

34.2

1952

6,636

45,508

1,502,107

33.0

1953

6,731

47,426

1,578,349

33.3

1954

6,894

49,283

1,625,742

33.0

1955

6,755

48,342

1,637,008

33.8

1956

6,844

50,186

1,693,879

33.7

1957

7,119

55,410

1,833,074

33.0

1958

7,406

59,679

1,962,243

32.8

1959

7,586

66,113

2,098,941

31.7

1960

7,860

69,658

2,192,379

31.4

1961

8,434

69,859

2,140,698

30.6

1962

8,765

76,353

2,267,564

29.6

1963

9,327

81,109

2,482,613

30.6

1964

9,434

95,137

2,540,913

26.7

1965

9,550

91,981

2,556,191

27.7

1966

9,560

90,515

2,565,891

28.3

1967

9,585

93,673

2,588,502

27.6

1968

9,801

92,982

2,633,637

28.3

1969

9,955

95,117

2,670,099

28.0

1970

9,931

96,426

2,716,187

28.1

1971

9,502

93,539

2,717,719

29.0

1972

9,417

95,281

2,265,241

26.9

1973

8,952

102,649

2,698,854

26.3

1974

9,645

102,656

2,622,424

25.1

1975

9,629

104,043

2,543,641

24.6

1976

9,683

110,563

2,571,984

23.2

1977

9,701

117,735

2,566,381

21.7

1978

9,726

n/a

3,083,725

25.0

1979

9,626

142,207

3,208,191

22.5

1980

9,794

141,185

3,389,776

24.0

1981

9,789

135,869

3,451,358

25.6

1982

9,901

133,802

3,484,661

26.0

1983

9,947

134,299

3,553,027

26.5

1984

9,914

140,190

3,625,897

25.9

Source: Rasaputra (1986: Statistical Appendix, Table 7).

While we have figures for nominal expenditures on education before 1950 (Rasaputra 1986: Statistical Appendix, Table 7), we do not—as noted in the previous section—have an appropriate index to account for price changes during this period. After 1950 the GDP deflator can be used to convert nominal figures to real magnitudes, and these are shown in our Table 2.3. We turn to a discussion of the post‐1950 period.

An examination of the movements of real per capita expenditure on education in the post‐independence period shows that there was a doubling of such expenditure in the decade of the 1950s, followed by a gradual increase in the 1960s until a peak was reached in 1969/70 (Table 2.3). Thus, real per capita expenditure on education rose from Rs. 14.60 in 1950/1 to Rs. 29.64 in 1960/1, and reached Rs. 35.53 in 1969/70. There was a sharp decline of over 20 per cent between 1973 and 1974, and real per capita expenditure did not recover its 1971/2 value until 1982. The five‐year average for 1973–7 was Rs. 27.04 while that for 1978–82 was Rs. 30.44. As in the case of health, real educational expenditure per capita seems to have been protected in the post‐1977 reform period.

In the thirty years from 1952 to 1981, the number of pupils more than doubled (Table 2.6). In light of this increase, the decrease in the pupil–teacher ratio from an average of 33.3 in the 1950s to an average of 25.6 during 1980–4 can perhaps be seen as an indicator of improved quality of education. The problem with this interpretation, as with interpretations of other educational indicators, is that such indicators should really be viewed as ‘inputs’ rather than ‘outputs’. The problem lies in specifying an appropriate ‘output’ of the educational system beyond such obvious indicators as literacy rates. In fact, as Table 2.7 shows, literacy rates have improved dramatically in Sri Lanka since the turn of the century. They increased from 26.4 per cent in 1901 to 39.9 per cent in 1921, and by 1953—the start of our three‐decade modern period—the literacy rate was already 65.4 per cent. By 1981, it was as high as 86.5 per cent.

Table 2.7  Literacy rate for population over 10 years of age, Sri Lanka, 1901–1981 (%)

Year

Female

Male

All

1901

8.5

42.0

26.4

1911

12.5

47.2

31.0

1921

21.2

56.4

39.9

1946

43.8

70.1

57.8

1953

53.6

75.9

65.4

1963

63.2

79.3

71.6

1971

70.9

85.6

78.5

1981

82.4

90.5

86.5

Source: Alailima (1985: Table 4).

Of course, since the maximum literacy rate is 100 per cent it is inappropriate to compare percentage changes over time—it is easier to get large percentage (p.72) (p.73) increases when the absolute level is low. Rather, we can ask what percentage of the shortfall between the base value of the literacy rate and the upper bound of 100 per cent is made up in any period (Sen 1981: 292). Using this measure for the period 1921–53, the shortfall of 60.1 per cent in 1921 was reduced to a shortfall of 34.6 per cent in 1953, i.e. a proportionate decline of 42.4 per cent over the 32‐year period, or 1.7 per cent a year. Between 1953 and 1981 the shortfall was reduced further to 13.5 per cent, which was a decrease of 61.0 per cent over the 28‐year period, or 3.3 per cent a year. Viewed in this way, the improvements in the literacy rate are seen as being much faster in the post‐independence period, although of course the movement had gathered some momentum by the time independence came.

In the case of mortality rates in the post‐independence period, we saw that declines were much faster in the 1950s than in the 1960s and 1970s. Is the same (p.74) true of the improvement in the literacy rate? We know that in 1963 the average literacy rate stood at 71.6 per cent. Thus, in the 10‐year period 1953–63 the literacy shortfall was reduced from 34.6 per cent to 28.4 per cent—a proportionate decline of 17.9 per cent, or 2.0 per cent a year. In the 18‐year period 1963–81 the shortfall decreased from 28.4 per cent to 13.5 per cent, a proportionate decline of 52.5 per cent, or 4.0 per cent a year. It does seem, then, as if the momentum towards greater literacy was not only maintained but intensified in the later post‐independence period. In contrast to the behaviour of health indicators, the improvement in literacy is much faster during the 1960s and 1970s.

As with health indicators, the overall satisfactory level of all‐island literacy rates masks important sectoral differences. From Rasaputra (1986: Table 18, p. 54) it is clear that literacy rates in the estate sector are much lower than those in other parts of the island. Moreover, most of the improvement in literacy rates seems to have come about in the non‐estate sectors, with the estate sector in fact registering a slight worsening in literacy.

(c) Food subsidies

Food subsidies in Sri Lanka were first introduced as the food ration scheme in 1942, a wartime relief measure. The scheme guaranteed the supply of basic food items at low prices. Rice was the most important component of this scheme and in what follows we will concentrate on rice. After a description of the rice ration schemes in the post‐war period, we will discuss food subsidy expenditure. Having established the nature of the intervention, we will then proceed to a consideration of the achievements, in so far as the data permit us to do so.

A brief history of rice subsidies is provided in Tables 2.8 and 2.9. Before 1954 the rationed quantity varied between adults, children, and infants. In 1954 all individuals became entitled to two measures of rationed rice per week at a low price (a measure is equal to two pounds avoirdupois). In June 1959 a price differential was introduced between the two measures allowed under the ration—the first measure was priced at Rs. 0.25 while the second was priced at Rs. 0.45. In April 1960 both measures were priced at Rs. 0.25. This price remained constant until December 1966.

Table 2.8  Rice ration distribution, Sri Lanka, 1950–1966

Ration quantity (measures per week)

Price (Rs. per measure)

Adults

Children

Infants

Manual workers

Income tax payers

Dec. 1950–

1.25

1.00

0.75

1.25

1.25

0.25

Sept. 1952–

1.00

0.75

0.50

1.25

1.00

0.25

July 1953–

1.25

1.00

0.75

2.00

1.25

0.70

Oct. 1953–

1.25

1.00

0.75

2.00

1.25

0.55

Nov. 1954–

2.00

2.00

2.00

2.00

2.00

0.55

May 1955–

2.00

2.00

2.00

2.00

2.00

0.50

May 1956–

2.00

2.00

2.00

2.00

2.00

0.40

June 1958–

2.00

2.00

2.00

2.00

2.00

0.35

June 1959–

2.00

2.00

2.00

2.00

2.00

0.25 (1st measure)

0.45 (2nd measure)

Apr. 1960–

2.00

2.00

2.00

2.00

2.00

0.25 (1st measure)

Dec. 1966

0.25 (2nd measure)

Note: A measure is equal to two pounds avoirdupois.

Source: Rasaputra (1986: Appendix A, Table A‐6).

Table 2.9  Rice ration distribution, Sri Lanka, 1966–1979

Basic ration

Additional ration

Quantity (measures per week)

Price (Rs. per measure)

Quantity (measures per week)

Price (Rs. per measure)

Non‐income tax payers

Income tax payers

19 Dec. 1966–

1.00

Free

Free

26 Sept. 1970–

1.00

Free

Free

1.00

0.75

10 Nov. 1971–

1.00

Free

Free

1.00

1.00

4 Dec. 1972–

1.00

Free

1.00

1.00

1.00

19 Feb. 1973–

1.00

Free

1.60

1.00

1.60

12 Mar. 1973–

1.00

Free

1.40

1.00

1.40

1 Oct. 1973–

0.50

Free

2.00

29 Oct. 1973–

0.50

Free

2.00

0.50

2.00

11 Nov. 1973–

0.50

Free

2.00

10 Dec. 1973–

0.50

Free

2.00

0.50

2.00

1 Feb. 1974–

0.50

Free

2.00

18 Feb. 1974–

0.50

Free

2.00

0.50a

2.00

18 Mar. 1974–

0.50

Free

2.00

1.00

2.00

15 Apr. 1974–

0.50

Free

2.30

1.00

2.30

29 Apr. 1974–

0.50

Free

2.30

0.50

2.30

6 May 1974–

0.50

Free

2.30

1.00b

2.30

15 July 1974–

0.50

Free

1.50

1.00b

2.50

5 Aug. 1974–

0.50

Free

2.20

1.00b

2.20

6 Nov. 1975–

0.50

Free

2.00

1.00b

2.00

4 Apr. 1977–

0.50

Free

2.00

1.50

2.00

May 1977–

0.50

Free

2.00

2.00

2.00

Feb. 1978c

1.00

Free

1.50

2.00

May 1978c

0.50

Free

1.50

2.00

–Sept. 1979

Notes: A measure is equal to two pounds avoirdupois.

(a) Colombo and suburbs only.

(b) In 21 deficit districts. Additional ration was half a measure for the rest of the country.

(c) Restricted to households with income less than Rs. 300 per month; some adjustment made for households of size greater than five.

Source: Rasaputra (1986: Appendix A, Table A‐7).

In December 1966 the scheme was changed again. The rationed quantity was halved to only one measure, but this measure was provided free. The rest of an individual's or a household's consumption could be made up in the open market. In 1970 there was a shift back to the June 1959 pattern of differential pricing—the first measure was still free but the second measure now cost Rs. 0.75. In December 1972 a very important distinction of principle was introduced, that between income tax payers and non‐income tax payers. The argument for ‘targeting’ was therefore accepted in principle. While non‐income tax payers still received their basic ration of one measure free, the income tax payers had to pay Rs. 1.00 for this measure. The additional ration of (p.75) one measure cost both groups of people Rs. 1.00.3 This basic structure was maintained, with changes in ration quantities and prices, until after the major reforms of 1977—and it is to these that we now turn.

As discussed in Anand and Sen (1984) (see also Jayawardena et al. 1987 and Kelegama 1990) the newly elected government of July 1977 began to introduce a programme of liberalization and adjustment. A devaluation of currency took place and this immediately increased the cost of imported food, and hence the cost of the food subsidy, measured in local currency terms. Over the next two years the subsidy was modified fundamentally by a series of policy changes. In April 1977, before the changes, the basic ration was half a measure per week, which was free to non‐income tax payers and cost Rs. 2.00 per measure to income tax payers. The additional ration was 1½ measures a week at Rs. 2.00 per measure for everyone. In February 1978, the ration was restricted to households with an income of less than Rs. 300 per month. Some adjustments were made to allow for household size: for households with more than five members, each additional member increased the income ceiling by Rs. 60, subject to a maximum of Rs. 750 per month. These new rules are estimated to have restricted the recipients of rationed rice to 7.6 million persons, around half the population (Ministry of Plan Implementation 1982).

(p.76)

In September 1979, the government introduced a food stamp scheme to replace the rationing system which had been in operation in Sri Lanka since 1942. Those families receiving income of Rs. 300 or less per month (excluding the income support of Rs. 50 per month given to the unemployed) were eligible for food stamps, which could be used to purchase a specified basket of goods. Families in receipt of an income in excess of Rs. 300 but less than Rs. 750 per month were also eligible for food stamps, the number of people eligible depending on the income and size of the family and the value of stamps (p.77) received depending on the age‐composition of the family (Ministry of Plan Implementation 1982).

In the final phase of the post‐1977 reforms beginning in 1980, price subsidies on rice, flour, and sugar were removed and their prices raised to reflect costs. The most striking feature of the new food stamp scheme was that the value of stamps received was not indexed to inflation. A total of Rs. 1,800 million, fixed in nominal terms, was allocated from the annual budget to meet the cost of food (and kerosene) stamps. It has been estimated that by 1984 the real value of this expenditure had been eroded by inflation to such an extent that a nominal expenditure of Rs. 3,200 million (nearly 7 per cent of the government budget) would have been required in that year to maintain the real value.

What seems to have happened, then, is that a major component of savings for the government budget in the post‐1977 period has come not from retargeting but from the post‐1979 erosion of real expenditure on food stamps. The strain on the budget has, of course, been a major theme of discussions on food subsidy policy ever since the first rationing scheme was introduced over four decades ago. Tables 2.3 and 2.4 show the real value of food subsidy expenditure per capita, and food subsidies as a percentage of GNP, respectively. As can be seen, the early period is characterized by a sharp fall in real food subsidy expenditure per capita from a high of Rs. 31.38 in 1951/2 to Rs. 1.45 in 1953/4 (at 1959 prices). The high values were the result of increases in the price of imported rice as a consequence of the Korean War. The low value was the result of a policy decision effectively to end all subsidies on food. The decision led to ‘food riots’, and the government changed. The new government reversed the policy, and real expenditure on food subsidies began a steady increase to Rs. 34.28 in 1963/4, which was more than its 1951/2 value. The rise during the ten‐year period from 1953/4 to 1963/4 was the longest sustained increase or decrease in the post‐independence period. After the peak of 1963/4, real expenditure per capita fluctuated with a three‐ or even two‐year cycle—troughs in 1966/7, 1969/70, 1971/2, and 1976, and peaks in 1968/9, 1970/1, 1975, and 1979. By 1982 real per capita expenditure on food subsidies had fallen to Rs. 20.72, a value comparable with that of the late 1950s (in the middle of the long period of sustained increase).

This brief account shows that government intervention in the area of food subsidy has been extensive. What have been its achievements? Fairly clearly, the record in achievements in the area of food has to be seen in terms of the extent to which food consumption—on average and for the poor—has changed over the years. We can consider real food consumption in aggregate, or the consumption of particular commodities. More directly, we can analyse how the nutritional status of the population—measured in terms of its calorie intake for example—has changed over time. A particularly serious problem is faced if we are interested in measuring intertemporal variations of the nutritional status of the poor in the country. For this we would need data on the joint distribution of food consumption and the variable with respect to which (p.78) poverty is defined. To match the annual figures for food subsidy expenditure, we would require corresponding distributional data for each year during the past three decades. Such data are simply not available for Sri Lanka.

What we have are a small number of surveys, undertaken at different points in time over the previous twenty years. There are the Consumer Finance Surveys (CFSs) conducted by the Central Bank of Ceylon for 1953, 1963, 1973, 1978/79, and 1981/82. Other surveys that have been used in the literature are the 1969/70 and 1980/81 Socio‐Economic Surveys conducted by the Department of Census and Statistics. The major problem that arises in using these surveys as seven observations spanning the post‐independence period is their comparability. As Pyatt (1987: 518) notes, the Socio‐Economic Surveys differ from the Consumer Finance Surveys in a number of respects which make comparisons difficult. Moreover, even if we stick to the Consumer Finance Surveys, for example, there is some question about comparability of the later surveys with the 1953 and 1963 ones, and about the quality of the data in the earlier period (see Anand and Harris 1985).

Given these problems, we will restrict ourselves to a comparison of the results of the 1973, 1978/79, and 1981/82 Consumer Finance Surveys. Anand and Harris (1985: 53–82) have argued for the comparability of these surveys in terms of their income and expenditure concepts, definition of unit of enumeration, continuity of Central Bank staff participation, etc. The 1978/79 and 1981/82 surveys also span the major food policy change in Sri Lanka—the removal of the generalized food subsidy scheme and its replacement by a targeted food stamp scheme. As discussed above, the change in structure was accompanied by a sharp decline in the real value of transfers to the poor accomplished through the scheme. We will, first of all, examine the changes in real food consumption and calorie intake between these two years, and then move on to a comparison with 1973.

Anand and Harris (1985) have constructed food price indices based on detailed food price and quantity data from the 1978/79 and 1981/82 surveys. Using these, they show that real food consumption per capita increased by 2.2 per cent for Sri Lanka as a whole between the survey years. But this aggregate increase hides major sectoral differences. While the urban sector increased its real food consumption per capita by 5.5 per cent, and the rural sector by 3.2 per cent, the estate sector experienced a fall of 8.7 per cent. A possible explanation is that while urban workers were protected from the decrease in food subsidy through cost‐of‐living related wage increases, and the rural sector benefited from higher paddy prices, the estate sector lost out in the changeover from the ration to the food stamp scheme (perhaps because of relatively easy to monitor money incomes from estates' wage registers).

So much for change in average real food consumption per capita. What happened at the lower end of the food consumption distribution? Anand and Harris (1985) calculate the incidence of food poverty for two poverty lines—monthly food expenditure per capita in 1978/79 of Rs. 70 and Rs. 60, (p.79) respectively. (These poverty lines are adjusted to take account of both intersectoral and intertemporal price differences.) Taking first the higher of the two poverty lines, they find that the incidence of poverty fell from 22.7 per cent to 21.9 per cent for Sri Lanka as a whole. Again, this overall small improvement conceals a significant improvement in the urban sector (24.4 per cent to 19.6 per cent), a minor improvement in the rural sector (23.8 per cent to 23.2 per cent), and a major deterioration in the estate sector (8.9 per cent to 13.8 per cent).

The above results are perhaps to be expected given the movements in average real food consumption per capita for Sri Lanka as a whole and for the sectors taken separately. However, results for the lower poverty line indicate some interesting changes at the bottom end of the distribution. With the Rs. 60 poverty line, the incidence of food poverty in Sri Lanka goes up from 12.9 per cent to 13.3 per cent. This increase in all‐island incidence is driven largely by an increase in rural sector incidence from 12.8 per cent to 13.6 per cent; a small positive contribution is also made by the rise in estate sector incidence from 3.6 to 5.8 per cent, but this is more than offset by the fall in urban incidence from 14.3 to 12.4 per cent.

The results for real food consumption are supported by Sahn's (1987) calculation of the percentage of individuals with a calorie intake per adult equivalent below certain levels. (Household calorie intake was derived from the CFSs by converting food quantities into calorie equivalents using food composition factors estimated by the Medical Research Institute of Sri Lanka.) Sahn (1987: Table 5, p. 818) shows that between 1978/79 and 1981/82 the percentage of individuals who belong to households with a daily intake per adult equivalent below 2,200 calories stayed constant at 31.1 per cent. But there was an increase in the percentage of individuals with an intake below 2,000 calories per day from 20.8 to 22.7 per cent. The percentage with an intake below 1,800 calories rose more sharply between 1978/79 and 1981/82, from 12.6 to 15.5 per cent. However, if an even lower cut‐off of 1,600 calories per day is used, the increase (from 7.0 to 10.2 per cent) is even more pronounced—the incidence of undernutrition goes up by 45 per cent.

An indirect method of looking at the distribution of calorie intake is to consider the calorie intake of those who are poor in terms of income or total expenditure. This is the strategy followed by Edirisinghe (1987: Tables 22 and 23, pp. 38–9) in his analysis of the CFS 1978/79 and 1981/82 data. His Table 22 shows that between 1978/79 and 1981/82 mean calorie consumption in the island as a whole—and in the urban and estate sectors—fell, while in the rural sector it rose. Furthermore, his Table 23 shows that the mean calorie consumption of the bottom three all‐island deciles fell between the survey years. A recent paper by Anand and Harris (1987) also estimates changes in nutrition in Sri Lanka between 1978/79 and 1981/82. Although critical of the methodology used—and the cleaning of CFS data—by Edirisinghe (1987), it nevertheless confirms the decline in per capita calorie intake by the lowest 30 per cent of the (p.80) all‐island population.4 At the sectoral level, however, the findings of Anand and Harris (1987) are significantly different: they find an increase in per capita calorie intake in the urban and rural sectors—and in the island as a whole—and a decrease only in the estate sector.5

Between 1979 and 1982 real government expenditure per capita on food subsidies fell from Rs. 62.29 to Rs. 20.72 (Table 2.3). While it would be difficult, given the other forces at play and the nature of the available data, to establish a clear and unambiguous link between this cut and food consumption of the population—the results are at the very least suggestive. The food stamp scheme replaced the earlier ration scheme in September 1979 and, despite leakages, the burden of real cuts in the food subsidy budget is likely to have fallen disproportionately on the poor. The increase in food poverty using the Anand and Harris (1985) low poverty line corroborates this suggestion, as does the increase in the percentage of individuals with calorie intake below a low cut‐off.

In order to investigate further the link between food subsidy expenditure and poverty, it would be instructive to compare 1979 with an earlier period. This is possible using the CFS 1973 data. Although there are no distributions of calorie intake available for that year,6 Anand and Harris (1985) have calculated the change in food poverty between 1973 and 1978/79. Using the poverty line of Rs. 70 (at 1978/79 prices), there was a fall in the incidence of poverty from 27.6 per cent to 22.7 per cent. The same trend is seen with the Rs. 60 (at 1978/79 prices) poverty line—the incidence of poverty fell from 15.0 per cent in 1973 to 12.3 per cent in 1978/79. Given that real food subsidy (p.81) expenditure per capita increased from Rs. 34.25 in 1973 to Rs. 62.29 in 1979 (see Table 2.3), this would tend to confirm the link between intervention and achievement. However, before entertaining such a conclusion, we should note that both estate and urban sector poverty increased during this period, no matter which poverty line is chosen. The causes of poverty are manifold, and without further detailed investigation of the pattern of food subsidy distribution, we cannot so easily draw a firm connection between food subsidy expenditure and poverty. Nevertheless, we would argue that there is a prima‐facie case for the link between intervention and achievement given this description of the historical record in Sri Lanka.

2.3. Intervention and achievement: an econometric analysis

The previous section has provided a historical overview of intervention and achievement in Sri Lanka. The discussion is suggestive of the link between intervention and achievement. It cannot be more than suggestive as we have not established a statistically significant relationship between them. This is where econometric analysis comes in. While such an analysis cannot do justice to the institutional detail of the historical development, it does provide a framework for testing relationships between variables in a stochastic setting.

Given the importance of Sri Lanka as a test case, it should not be surprising that much is written about the country in the applied econometric literature. In a series of papers, Isenman (1980, 1987), Sen (1981, 1988), Bhalla and Glewwe (1986), Glewwe and Bhalla (1987), Pyatt (1987), Ravallion (1987), and Bhalla (1988 a, 1988b) have all contributed to a debate on whether Sri Lanka's achievements are exceptional, and the links of these achievements to intervention. A characteristic feature of this literature is that it is based on econometric analysis of a cross‐section of countries, Sri Lanka being one of them. The debate centres around establishing Sri Lanka as an ‘outlier’ in the sample, and around the interpretation of its outlier status.7

Our major concern in this chapter is with examining intervention and achievement in Sri Lanka over time. Accordingly, we wish to investigate the link by means of econometric analysis of time‐series data for Sri Lanka. In doing so we circumvent many of the problems that are peculiar to the cross‐section framework. Section 2.3(a) provides a brief review of the cross‐section evidence, focusing on why a time‐series approach is more appropriate. Section 2.3(b) proceeds to the time‐series analysis.

(p.82) (a) A critique of the cross‐section literature

In our brief excursion into the cross‐section literature, we will adopt the basic notation used by Bhalla and Glewwe (1986: equation (4), p. 39). They posit the following model to explain some measure of living standard, H it, for country i at time t:

H i t = α t + β Y i t + δ E i t + λ i + u i t ' '
(2.1)

where Y it is per capita income; E it is social welfare expenditure; α‎t is a time‐specific but country‐invariant effect assumed to reflect technological advances (e.g. disease eradication techniques); λ‎i is a country‐specific and time‐invariant ‘fixed effect’; δ‎ is the marginal impact of social expenditure on living standards; and u i t ' ' is a random error term.8

If we had data on all the variables of (2.1), then of course we could estimate the equation directly. However, data on E it and λ‎i are typically not available for a cross‐section of countries, and Isenman (1980), Sen (1981), and others usually estimate

H i t = α t + β Y i t + e i t
(2.2)

for a cross‐section of countries at a given point in time. They find that Sri Lanka is an outlier, having much higher values of H than predicted by the estimated relationship. This they attribute to Sri Lanka's record in intervention on basic needs. Comparing (2.1) and (2.2) we see that

e i t = δ E i t + λ i + u i t ' '
(2.3)

so that a large positive residual for a country could be attributed either to a large E it (assuming δ‎ > 0) or to a large λ‎i, or to some combination of the two. This is the crux of the Bhalla and Glewwe (1986) and Bhalla (1988a, 1988 b) criticism of the Isenman–Sen analysis. Of course, a large λ‎i may itself be due to past expenditures on social welfare, but presumably the focus is on the period in question.

In order to control for the effect of λ‎i, Bhalla and Glewwe (1986: equation (5), p. 39) suggest the first‐difference model

Δ H i t = Δ α t + β Δ Y i t + u *
(2.4)

where, for a variable x, Δ‎x t is defined as

Δ x t = x t + 1 x t .

(p.83) A comparison of (2.1) and (2.4) shows that

u * = δ Δ E i t + Δ u i t "
(2.5)

Bhalla and Glewwe (1986: 40) argue that ‘It is the residual of equation [2.4], and not the residual of equation [2.2], that may be useful in assessing country performance over time’. They estimate equation (2.4) for a cross‐section of countries, with t = 1960 and t + 1 = 1978. They argue that for this regression Sri Lanka is no longer an outlier.

Let us return to the basic model in (2.1). We are interested in the sign and magnitude of the coefficient δ‎, and also its magnitude relative to the coefficient δ‎. It is this comparison which allows us to comment on the efficacy or otherwise of the direct and indirect (i.e. income‐growth) route to improving living standards. If we do not have data on E it and estimate (2.2), can we nevertheless infer the sign and magnitude of δ‎ from the residual of the cross‐section regression?

If the regression is as in (2.2), and we denote the estimated value of the residual as e ˆ i t , then

E ( e ˆ i ) = ( Y i Y ) ( β E ( β ˆ ) ) + δ ( E i E ) + ( λ i λ )
(2.6)

where the t‐subscript has been suppressed because the regression is cross‐section, β ˆ is the ordinary least squares (OLS) estimate of β‎, and a bar over a variable indicates its sample mean.9 As can be seen from (2.6) it is more likely that the residual for a country i will be large and positive in expectation if: (1) δ‎ > 0 and E i > E , which is the Isenman–Sen argument; or (2) λ i > λ , which is the Bhalla–Glewwe critique; or (3) ( Y i Y ) ( β E ( β ˆ ) ) > 0 , a possibility which is not entertained to any great extent by either Isenman–Sen or Bhalla–Glewwe. It is easy to show that in the OLS estimate β ˆ of β‎ in (2.2) the bias arising from the omitted variables (E i and λ‎i) is

β E ( β ˆ ) = δ Σ i ( Y i Y ) ( E i E ) Σ i ( Y i Y ) 2 Σ i ( Y i Y ) ( λ i λ ) Σ i ( Y i Y ) 2 .

Assume for the moment that the correlation between λ‎i and Y i is zero; we have little reason to suppose otherwise. Then if δ‎ > 0 and E i and Y i are positively (p.84) correlated,10 we have β E ( β ˆ ) < 0 . Now from (2.6), if Y i < Y (country i's income is less than the sample average) then we will get an upward bias in the residual e ˆ i . Of course, a non‐zero correlation between λ‎i and Y i will also confound the inference that can be drawn from the residual e ˆ i .

What if we estimate the Bhalla–Glewwe first‐difference model (2.4)? What can be inferred from its estimated residual u ^ * ? Analogously to (2.6), we get

E ( u ^ * ) = ( Δ Y i Δ Y ) ( β E ( β ^ ) ) + δ ( Δ E i Δ E )
(2.7)

where β ˆ is now the OLS estimate of β‎ in (2.4) and the bias depends on the correlation between Δ‎E i and Δ‎Y i. As can be seen from (2.7), even though there is no ( λ i λ ) term, an insignificant value of u ^ * does not necessarily imply that δ‎ is zero. A zero value for E ( u ^ * ) is quite consistent with a positive value for δ‎. For example, if there is no bias in β ˆ , i.e. β = E ( β ˆ ) , then Δ E i = Δ E will give a zero value for E ( u ^ * ) even with δ‎ > 0, and this is indeed a line of defence adopted by Sen (1988: 550–2).

Sen argues that during the period under consideration (1960–78), the increment in social welfare expenditure in Sri Lanka was not exceptional relative to the sample; thus, it is not surprising that the increment in H is not exceptional. In order to adjudicate on this issue we would need fuller data on Δ‎E i in the sample, and this is indeed the problem—if we had those data we could estimate the relationship directly and not have to rely on the residual method to give us an indication of the value of δ‎.

Thus, our conclusion is that while the Isenman–Sen method may be open to certain criticisms, the Bhalla–Glewwe alternative cannot resolve the basic question of the relative magnitudes of δ‎ and β‎. Given the lack of cross‐section data on social welfare expenditures it is difficult to see how it could, in fact, be resolved. However, with time‐series data for a particular country, we can obtain estimates of δ‎ and β‎ directly for that country. It so happens that Sri Lanka is indeed a country for which such data are available. In the next section we proceed to utilize these data in an econometric analysis of the relationship between intervention and achievement. This corresponds to the ‘explicit approach’ of Sen (1988: 550):

The common wisdom of the approach is based on the idea that we cannot really measure the impact of a policy of social welfare programs without explicitly incorporating it as a variable in a causal framework and testing its effect.

(b) Time‐series evidence for Sri Lanka

For our time‐series investigation, we retain the Bhalla–Glewwe (1986) specification given in equation (2.1). We focus attention on Sri Lanka for the Bhalla–Glewwe period 1960–78, but also discuss the longest time period for (p.85) which a consistent series is available, namely 1952–81. Relevant measures of social expenditure E it are explicitly included for each year.

Our data for this purpose are drawn almost exclusively from the paper by Alailima (1985), ‘Evolution of Government Policies and Expenditure on Social Welfare in Sri Lanka during the 20th Century’, which is extensively referred to and used by Bhalla–Glewwe (1986) and Bhalla (1988 a, 1988b). Alailima's Table 8, on which our Table 2.3 is based, gives a 32‐year series from 1950/1 to 1982 for real per capita expenditure on social services (separately for education, health, food subsidies, and other social welfare).

Alailima (1985: Table 8) calculates real expenditures by using the GDP deflator, which includes social welfare expenditure as a component. She calculates per capita expenditure by using population estimates from the Department of Census and Statistics. Finally, her Table 3 presents vital statistics data from the same source for the period 1900 to 1981. This series consists of estimates of the crude birth rate, the crude death rate, and the infant mortality rate. The same information is available, but up to 1984, in Rasaputra (1986: Statistical Appendix, Table 10), and this is reproduced in our Table 2.2. Rasaputra's paper (1986: Statistical Appendix, Table 1) also contains a consistent series for real GDP at factor cost from 1950 onwards, using the same GDP deflator (with 1959 = 100) as Alailima does for her series on real social welfare expenditures. It is important to use this comparable income series for GDP at factor cost because, as is well known, the GNP series in Sri Lanka has been revised twice—in 1958 and 1970—and the new series is not consistent with the old one. Real GDP per capita has been calculated by us using the same mid‐year population figures (Table 2.2) as Alailima (1985: Table 8) uses to calculate her per capita social expenditures. These real GDP per capita estimates are shown in our Table 2.3.

We are now ready to estimate the Bhalla–Glewwe specification of the living standards relationship

H i t = λ i + α t + β Y i t + δ E i t + u i t ' '
(2.1)

where

country i = Sri Lanka, fixed in the sample

  • H it = some measure of living standard such as infant mortality rate (IMR), death rate (DR), or birth rate (BR) in year t

  • λ‎i = country‐specific intercept term for Sri Lanka

  • α‎t = technical progress term, specified simply as α‎.t (with α‎ constant)

  • Y it = real GDP per capita in year t

  • E it = real social expenditure per capita in year t (separately for health, education, and food).

In their cross‐section analysis, Bhalla–Glewwe (1986) and Bhalla (1988 a) consider six indicators of living standard—life expectancy, primary school (p.86) enrolment, adult literacy rate, infant mortality rate, death rate, and total fertility rate. Given the time‐series data available to us, we are obliged to restrict attention to the infant mortality rate (IMR), the death rate (DR), and the birth rate (BR). Since the H it variables are bounded below by zero, we use them in logarithmic form—as In H—so that the dependent variable in equation (2.1) can be negative (to minus infinity) for negative realizations of the right‐hand side. Since the H it variables are also bounded above by 1,000 (IMR, DR, and BR are all measured per 1,000 population), we can allow unbounded variation upwards (to plus infinity) of the dependent variable by subtracting In (1,000 − H) from In H, i.e. by using H it in the logistic form In [H/(1,000 − H)]. We have not done this here because the sample values of H it occur in a region much closer to zero than to 1,000; hence the further transformation is unlikely to affect significantly the estimates of the coefficients of the independent variables (excepting, of course, the intercept term). In any case, this has been confirmed by doing the regressions in logistic form (not reported here).

Table 2.10 presents the results of our time‐series regressions. The right‐hand side independent variables (except t) have been entered in both linear (non‐log) and logarithmic form; t is always entered in linear form. The ‘A’ and ‘B’ equations in Table 2.10 refer respectively to these forms. First we report results for the Bhalla–Glewwe period 1960–78, for which the authors claim that Sri Lanka is not an outlier. This is followed by results for the full three‐decade period 1952–81 for which a consistent time series was available. We have entered the real per capita health and food subsidy expenditures separately—as the variables HEXP and FEXP. There are two interrelated reasons for doing this. First, the impact of health and food subsidy expenditures may be expected to be different from one another. Secondly, the food subsidy accounts for a relatively small proportion of total food consumption, so that variations in it will not reflect corresponding variations in the total food consumed by the population. By contrast, the coverage of health is more nearly universal, so that variations in health expenditure will more closely track health provision for the population. Current and capital expenditures on health have been aggregated in the variable HEXP. A similar procedure is followed for real per capita education expenditure, EEXP, which is also introduced into the regressions separately to allow for possible differential effects.

Table 2.10 Time‐series estimates of living standard regression equations, Sri Lanka

Dependent variable

Equation number

Time period

Intercept

HEXP

FEXP

EEXP

PCY

t (year)

F‐statistic

SEE

Mean of dependent variable

R 2

Log of likelihood function

Ln IMR

(1A)

1960–78

44.54

−0.0381

−0.002900

0.00599

0.000412

−0.02060

17.84

0.0471

3.89

0.873

34.72

(2.14)

(−3.21)

(−1.33)

(1.33)

(0.52)

(−1.90)

1952–81

24.27

−0.0322

0.001970

−0.00117

−0.000982

−0.00975

90.16

0.0572

3.94

0.949

46.63

(2.30)

(−2.79)

(1.63)

(−0.28)

(−3.01)

(−1.77)

(1B)

1960–78

44.99

−0.5620

−0.054400

0.20200

0.274000

−0.02130

17.20

0.0478

3.89

0.869

34.42

(2.46)

(−3.36)

(−0.91)

(1.62)

(0.47)

(−1.91)

1952–81

25.97

−0.4300

0.018300

−0.04260

−0.764000

−0.00802

75.35

0.0622

3.94

0.940

44.08

(2.18)

(−2.42)

(0.82)

(−0.35)

(−2.22)

(−1.10)

Ln DR

(2A)

1960–78

11.43

−0.0460

0.001380

0.00795

−0.000229

−0.00446

3.52

0.0565

2.09

0.575

31.24

(0.46)

(−3.23)

(0.53)

(1.47)

(−0.24)

(−0.34)

1952–81

8.53

−0.0482

0.002420

0.00063

−0.000787

−0.00266

40.84

0.0582

2.13

0.895

46.11

(0.79)

(−4.10)

(1.96)

(0.15)

(−2.37)

(−0.47)

(2B)

1960–78

17.32

−0.6630

0.072000

0.25400

−0.095700

−0.00708

4.30

0.0532

2.09

0.623

32.38

(0.85)

(−3.56)

(1.08)

(1.83)

(−0.15)

(−0.57)

1952–81

9.61

−0.6450

0.030100

−0.00733

−0.613000

−0.00092

36.03

0.0615

2.13

0.882

44.44

(0.82)

(−3.68)

(1.37)

(−0.06)

(−1.80)

(−0.13)

Ln BR

(3A)

1960–78

19.59

0.0232

0.002720

−0.00520

−0.000733

−0.00805

61.51

0.0224

3.44

0.959

48.80

(1.98)

(4.10)

(2.62)

(−2.43)

(−1.95)

(−1.56)

1952–81

39.13

0.0191

0.000487

−0.00176

0.000258

−0.01830

96.58

0.0281

3.47

0.953

67.96

(7.54)

(3.37)

(0.82)

(−0.88)

(1.61)

(−6.77)

(3B)

1960–78

24.20

0.3290

0.068300

−0.15000

−0.451000

−0.00933

58.46

0.0230

3.44

0.957

48.34

(2.75)

(4.09)

(2.37)

(−2.50)

(−1.60)

(−1.74)

1952–81

37.88

0.2810

0.016900

−0.07630

0.200000

−0.01840

100.90

0.0275

3.47

0.955

68.59

(7.21)

(3.58)

(1.72)

(−1.43)

(1.32)

(−5.75)

Notes: The coefficients of the ‘A’ equations refer to the independent variables entered in linear (non‐log) form; those for the ‘B’ equations refer to the independent variables (except for year t) entered in logarithmic form. Year t is entered in linear form in both the ‘A’ and the ‘B’ equations. Thus, the estimated equation (1A) for the time period 1960–78 is:

Ln IMR = 44.54 − 0.0381 HEXP − 0.002900 FEXP + 0.00599 EEXP + 0.000412 PCY − 0.02060 t.

The estimated equation (1B) for the time period 1960–78 is:

Ln IMR = 44.99 − 0.5620 ln HEXP − 0.054400 ln FEXP + 0.20200 ln EEXP + 0.274000 ln PCY − 0.02130 t.

t‐statistics are shown in parentheses below the coefficient estimates.

For the period 1960–78, the results (equations (1A) and (1B) in Table 2.10) show that health expenditure HEXP has a very significant negative effect on IMR, but that FEXP, EEXP, and PCY (real GDP per capita) are insignificant.11 What is important about the results is that direct intervention, as reflected in government health expenditure, has a statistically significant beneficial impact on IMR. Note that our procedure based on time‐series data (p.87) (p.88) has allowed a direct test of this relationship, and is not open to the problems of the cross‐section approach (mentioned in Section 2.3(a)).

(p.89) For the full period 1952–81, the results show that HEXP remains significant in reducing IMR: government intervention continues to matter. For this longer period, in contrast, the coefficient on the income term, PCY, becomes significant.

These results shed some light on the role of direct intervention versus an indirect, income‐growth strategy in reducing the infant mortality rate. According to our results for the Bhalla–Glewwe period 1960–78, income growth did not matter at all. This is a rather striking finding because not only is it the case that direct intervention has worked—a claim which is at the heart of the Bhalla–Glewwe versus Isenman–Sen controversy—but the estimates suggest that reliance on an income‐growth strategy would not have worked during 1960–78.

For the full period 1952–81, the estimates do show a significant income effect, but this effect is small. A rupee of government health expenditure diverted to income in the hands of the population would have led to an immediate rise in the infant mortality rate. Comparing the coefficient (−0.0322) on HEXP with that (−0.000982) on PCY in equation (1A), the former is larger than the latter by a factor of 33 in absolute terms. This implies that to redress the effect on IMR of a Rs. 1 decrease in health expenditure would require, ceteris paribus, a Rs. 33 increase in equivalent income—a manifestly adverse trade‐ off for the income‐growth strategy.

For both the periods 1960–78 and 1952–81, the results for IMR appear to be well determined and robust with respect to the functional form (various diagnostic tests not reported here support this conclusion). Thus in Table 2.10 the picture for equation (1B) turns out to be very similar to that for equation (1A). When the independent variables are entered in logarithmic instead of linear form, hardly any difference is made to the significance of the coefficients, though obviously their magnitude changes.12

In the next set of equations the dependent variable is the logarithm of the death rate, In DR. For the period 1960–78, the results in equation (2A) are again very striking with real health expenditure per capita being highly significant in reducing the death rate. No other variable is significant. For 1960–78 the results in equation (2B) show the same pattern as in (2A) and, as before, the functional form does not seem to make any qualitative difference to the findings.

For the full period 1952–81, the estimates in equations (2A) and (2B) do display a degree of sensitivity to the functional form chosen. In equation (2B) the income variable In PCY is not significant while in equation (2A) the income variable PCY is indeed significant (but again with very small coefficient). Whereas in equation (2A) the food subsidy variable FEXP is significant (and with positive coefficient), in equation (2B) In FEXP is not significant. The education expenditure variable EEXP is insignificant in both functional forms. The only robust inference for the 1952–81 In DR regressions seems to be that health expenditure HEXP has a very significant beneficial impact in reducing the death rate.

The final demographic indicator considered is the birth rate (BR), for which the regression results are shown in equations (3A) and (3B). At the outset, it should be emphasized that this is arguable as a living standard indicator, but we include it here only to correspond to Bhalla and Glewwe's total fertility rate.13 Over the period 1960–78, the birth rate turns out to be positively related to both health and food expenditure by government, and negatively related to educational expenditure—all significant at the 5 per cent level. In the non‐log form, the coefficient on income is negative and (almost) significant; in the log form the income coefficient is insignificant but still negative.

The coefficients on educational expenditure and on average (across‐the‐board) income are not difficult to rationalize, and may be considered to be of the expected sign. But how are we to interpret the positive coefficients on health and food expenditure? One possibility is that larger health and food subsidies lead to better antenatal care (including nutrition) of mothers, especially at the lower end of the income distribution. This might help more pregnancies to come to term and to avoid miscarriages. To test this hypothesis directly, however, we need more disaggregate data on the composition of health expenditures and on birth rate by income group (who is having more children?).

For the full period 1952–81, the relationship seems to be quite different, except in the respect that health expenditure continues to be significant and positive. Otherwise, the coefficients on food expenditure and educational expenditure become insignificant, the coefficient on income turns from negative to positive (but becomes decidedly insignificant), and a strong negative time trend emerges. Our worries about the use of the birth rate as an appropriate living standard indicator are compounded by this non‐robustness in the face of sample period variation.

To conclude, then, we note that this first attempt at a time‐ series analysis of intervention and achievement does provide econometric support for the hypothesis of a link between the two. The results for the infant mortality rate will perhaps bear emphasizing. Over the period 1960–78, and for the full (p.90) period 1952–81, health expenditure has a very significant effect as an explanatory variable for IMR. Moreover, the estimates for the period 1960–78 indicate that income would have had an insignificant effect on IMR. The estimates for the full period 1952–81 do show a significant effect for income, but this effect is very small. In this context, at least, reliance on income growth alone can be questioned.

We view our results as cautionary rather than definitive. A large and sustained increase in income over a long period might well have an impact on social indicators. Over the short‐ to medium‐run planning horizon, however, developing countries do face real choices between social expenditure and capital investment. Our results would tend to support those who argue for greater benefits at the margin from targeted social expenditure.

2.4. Conclusion

The object of this chapter has been to consider Sri Lanka's record of intervention and achievement in some areas of basic needs provision. We have used two methods of analysis. First, we have provided a descriptive account of intervention over the long run of historical developments this century, and have tried to relate this intervention to achievement by an accompanying narrative of the achievement. This discussion is strongly suggestive that purposive and directed intervention has had remarkable effects on health and education standards both in the early part of the century and in the period after independence.

Complementary to the descriptive approach is our second method of econometric analysis. We have reviewed the current literature on establishing and interpreting Sri Lanka's position as an outlier in a cross‐section of countries. We argue that in the absence of direct information on intervention for countries in the sample, such cross‐section analysis can be problematic. We propose instead that time‐series data for Sri Lanka be used to conduct a more direct investigation of the issues. We have presented a first attempt at such an analysis using data for the 1952–81 period. While our results need to be confirmed by further research, they do suggest that income growth alone would not have achieved for Sri Lanka its enviable basic needs record—the role of direct intervention has been significant.

We are not alone in reaching this conclusion. The central finding of Caldwell (1986: 204), who uses a combination of comparative and intertemporal methods, is that ‘the provision of health services (and, better still, its accompaniment by the establishment of a nutritional floor and perhaps a family planning program) can markedly reduce mortality’. His findings and ours suggest, therefore, that attention should now shift from the question of whether intervention can have a positive impact on basic needs to the more important question of the best patterns and combinations of social welfare expenditure to achieve the maximum impact on basic needs.

(p.91) References

Bibliography references:

Alailima, P. (1985), ‘Evolution of Government Policies and Expenditure on Social Welfare in Sri Lanka during the 20th Century’, mimeo (Colombo: Ministry of Finance and Planning).

Anand, S., and Harris, C. J. (1985), ‘Living Standards in Sri Lanka, 1973–1981/82: An Analysis of Consumer Finance Survey Data’, mimeo (Oxford).

——— ———(1987), ‘Changes in Nutrition in Sri Lanka, 1978/79–1981/82’, mimeo (Helsinki: WIDER).

———and Sen, Abhijit (1984), ‘The Macroeconomy of Sri Lanka after Liberalization’, mimeo (Oxford: St Catherine's College).

Basu, K. (1991), ‘The Elimination of Endemic Poverty in South Asia: Some Policy Options’, this volume.

Bhalla, S. S. (1988a), ‘Is Sri Lanka an Exception? A Comparative Study of Living Standards’, in Srinivasan and Bardhan (1988).

———(1988b), ‘Sri Lanka's Achievements: Fact and Fancy’, in Srinivasan and Bardhan (1988).

———and Glewwe, P. (1986), ‘Growth and Equity in Developing Countries: A Reinterpretation of the Sri Lankan Experience’, World Bank Economic Review, 1.

Caldwell, J. C. (1986), ‘Routes to Low Mortality in Poor Countries’, Population and Development Review, 12.

Edirisinghe, N. (1987), The Food Stamp Scheme in Sri Lanka: Costs, Benefits and Options for Modification, Research Report 58 (Washington, DC: IFPRI).

Fernando, D. F. S. (1985), ‘Health Statistics in Sri Lanka, 1921–80’, in Halstead et al. (1985).

Gavan, J. D., and Chandrasekera, I. S. (1979), The Impact of Public Foodgrain Distribution on Food Consumption and Welfare in Sri Lanka, Research Report 13 (Washington, DC: IFPRI).

Glewwe, P., and Bhalla, S. S. (1987), ‘A Response to Comments by Graham Pyatt and Paul Isenman’, World Bank Economic Review, 1.

Gunatilleke, G. (ed.) (1984), Intersectoral Linkages and Health Development: Case Studies in India (Kerala State), Jamaica, Norway, Sri Lanka, and Thailand, WHO Offset Publication No. 83 (Geneva: WHO).

——— (1985), ‘Health and Development in Sri Lanka: An Overview’, in Halstead et al. (1985).

Halstead, S. B., Walsh, J. A., and Warren, K. S. (eds.) (1985), Good Health at Low Cost, Proceedings of a Conference held at the Bellagio Conference Centre, Bellagio, Italy, 29 Apr.–2 May (New York: Rockefeller Foundation).

Isenman, P. (1980), ‘Basic Needs: The Case of Sri Lanka’, World Development, 8.

——— (1987), ‘A Comment on “Growth and Equity in Developing Countries: A Reinterpretation of the Sri Lankan Experience,” by Bhalla and Glewwe’, World Bank Economic Review, 1.

Jayawardena, L. R., Maasland, A., and Radhakrishnan, P. N. (1987), ‘Sri Lanka’, Country Study 15, WIDER Series on Stabilization and Adjustment Policies and Programmes (Helsinki: WIDER).

Kelegama, S. B. (1990), ‘The Consequences of Economic Liberalization in Sri Lanka’, unpublished D.Phil. thesis, University of Oxford.

(p.92) Marga Institute (1984), Intersectoral Action for Health: Sri Lanka Study (Colombo: Sri Lanka Centre for Development Studies).

Ministry of Plan Implementation (1982), Evaluation Report on the Food Stamp Scheme, Publication No. 7 (Colombo: Food and Nutrition Policy Planning Division).

———(1983?), Nutritional Status: Its Determinants and Intervention Programmes, Final Report (Colombo: Food and Nutrition Policy Planning Division).

———(1984), Nutrition Strategy (Colombo).

———(1985), Health and Nutrition Sector Report, National Science and Technology Policy for Sri Lanka, vol. vii (Colombo).

Orde Browne, G. St J. (1943), Labour Conditions in Ceylon, Mauritius, and Malaya, Cmd. 6423 (London: HMSO).

Perera, P. D. A. (1985), ‘Health Care Systems of Sri Lanka’, in Halstead et al. (1985).

Pyatt, F. G. (1987), ‘A Comment on “Growth and Equity in Developing Countries: A Reinterpretation of the Sri Lankan Experience,” by Bhalla and Glewwe’, World Bank Economic Review, 1.

Rasaputra, W. (1986), ‘Public Policy: An Assessment of the Sri Lanka Experience’, mimeo (Colombo: Central Bank of Ceylon; and Helsinki: WIDER).

Ratnayake, R. M. K. (1985), ‘A Survey Paper on Nutrition Situation in Sri Lanka’, mimeo (Colombo: Food and Nutrition Policy Planning Division, Ministry of Plan Implementation).

Ravallion, M. S. (1987), ‘Growth and Equity in Sri Lanka: A Comment’, mimeo (Canberra: Australian National University).

Sahn, D. E. (1987), ‘Changes in the Living Standards of the Poor in Sri Lanka during a Period of Macroeconomic Restructuring’, World Development, 15.

Sen, A. K. (1981), ‘Public Action and the Quality of Life in Developing Countries’, Oxford Bulletin of Economics and Statistics, 43.

——— (1988), ‘Sri Lanka's Achievements: How and When?’, in Srinivasan and Bardhan (1988).

Srinivasan, T. N., and Bardhan, P. K. (eds.) (1988), Rural Poverty in South Asia (New York: Columbia University Press).

Tinker, H. (1974), A New System of Slavery: The Export of Indian Labour Overseas, 1830–1920 (London: Oxford University Press).

Notes:

(*) An earlier version of this paper was presented at a seminar at WIDER, Helsinki, in Aug. 1986. We are grateful to Saman Kelegama and Madhura Swaminathan for research assistance, and to Jean Drèze, Keith Griffin, Lal Jayawardena, Heather Milne, Martin Ravallion, Abhijit Sen, and Amartya Sen for comments.

(1) Perera (1985) gives an intriguing historical account of health care systems in Sri Lanka from 300 BC to the present. A detailed overview of health and development in Sri Lanka in the modern period is provided by Gunatilleke (1985).

(2) In his review of health and development in Sri Lanka, Gunatilleke (1985: 112) notes that ‘The year 1947 marks a very significant turning point. The infant mortality rate dropped steeply from 141 per thousand to 101 and the crude death rate from 19.8 to 14. This steep decline in mortality recorded in the period of one year was described by WHO as “an unparalleled achievement in world demography”.’

(3) See Gavan and Chandrasekera (1979: 27–9) for details of changes in the food subsidy scheme between 1952 and 1977.

(4) Another indication of the deteriorating nutritional status of the population emerges from the anthropometric data collected in two surveys on pre‐school children undertaken in 1975/76 and 1980/82. The 1975/76 survey was conducted by the Ministry of Health in Sri Lanka, with technical assistance from the US Center for Disease Control in Atlanta, Georgia. The 1980/82 survey was conducted by the Food and Nutrition Policy Planning Division of the Ministry of Plan Implementation, Sri Lanka. The findings of these surveys have been reviewed in, inter alia, Ministry of Plan Implementation (1983?: ch. II), Ratnayake (1985), and Sahn (1987). The conclusion of Sahn (1987: 813) is that ‘The percentage of children suffering from acute malnutrition was higher in 1980/82 than in 1975/76. Overall, there was a 64 per cent increase in wasting in the rural sector from one survey to the next. The increase was especially high among the 6–11 month old age cohort. This undoubtedly reflects a combination of a decline in dietary intake, more episodes of infection, and less favourable birth outcomes conditioned by the mother's health and nutritional status. The prevalence of concurrent wasting and stunting is also higher in 1980/82 than in 1975/76 …’

(5) Anand and Harris (1987) identify several problems with the Edirisinghe (1987) methodology for estimating mean calorie consumption by decile and sector. (These are apart from problems with his cleaning of the CFS 1978/79 and CFS 1981/82 food quantity files, and with ensuring the comparability of food items between the two surveys.) For example, Edirisinghe (1987: 37–9) estimates per capita daily calorie consumption as an unweighted mean across households of household per capita calorie consumption. Obviously this is not a meaningful average of individual calorie intakes. For details of the biases caused by this and other problems with the Edirisinghe methodology, see Anand and Harris (1987).

(6) As noted in Anand and Harris (1985), the detailed food quantities file of the CFS 1973 data is unfortunately no longer available.

(7) A non‐econometric but nevertheless ‘cross‐country’ approach is also employed by Caldwell (1986). Using World Bank data he shows that ‘some countries reach health levels far above those that would be dictated by their economies and others fall far below. Thus the superior health achievers are characterized by average per capita income levels one‐ninth of those of the poor health achievers, but, nevertheless, record half the infant mortality level and an expectation of life at birth ten years higher.’ (Caldwell 1986: 173)

(8) It should be stated at the outset that this specification has a number of problems other than the ones which we deal with below. For example, the distribution of income as well as its average level may be expected to influence social indicators; and there may be interaction effects between Y and E. Furthermore, dynamic considerations and the role of the ‘stock’ of E, as opposed to its flow, may also be important. These shortcomings are recognized in the literature, but the bulk of the discussion is organized around the Bhalla‐Glewwe specification given in equation (2.1). A major problem with the more complete specification is that the data requirements—e.g. on the distribution of income and the stock of E—are greater.

(9) Dropping the t‐subscript, the OLS regression of (2.2) yields the residual

e ˆ i = H i H ˆ i = α + β Y i + δ E i + λ i + u i ' ' α ˆ β ˆ Y i

using (2.1) and H ˆ i = α ˆ + β ˆ Y i . But since the OLS regression passes through the sample means Y , H , we have

α ˆ = H β ˆ Y = α + β Y + δ E + λ + u ' ' β ˆ Y

using (2.1). Substituting in the equation above for e ˆ i , and taking expectations, gives the expression (2.6) for E ( e ˆ i ) .

(10) In fact, Bhalla seems committed to such a positive correlation between E t and Y i. In the course of formulating his living standards model, Bhalla (1988 a: 101) specifies the relationship E i t = β ' Y i t + e i t ' , which with β ' > 0 will in general imply a positive correlation between E t and Y i. By contrast, Isenman (1980; 1987) and Sen (1981; 1988) are not committed to any such correlation.

(11) The significance level chosen for the discussion here is 5%.

(12) For the sample period 1952–81 the absolute value of the coefficient on ln PCY is greater than that on ln HEXP. However, these coefficients, unlike those in the equations in non‐log form, are elasticities: they indicate the impact of a proportionate change in the independent variable. A 1% decrease in HEXP will only allow an increase of approximately 0.019% in income in the hands of the population, since the average ratio of HEXP to PCY over the sample period is 0.019 (see Table 2.4). Hence the relevant comparison (in terms of ‘bang‐for‐a‐buck’) is between the coefficient (−0.4300) on ln HEXP and 0.019 times the coefficient (−0.764000) on ln PCY—which implies a factor of 30 for the trade‐off.

(13) See e.g. Basu (1991: n. 6) who argues that the birth rate should not be included as a living standard indicator.