Document(s) 10 of 16

The growth of the tertiary sector in India seems to be somewhat out of line with international experience of recent decades. Table 10.1 brings together the data for sectoral changes in the shares of employment for several Asian countries over the last three decades of the twentieth century. The newly industrializing countries of Asia–Korea and Taiwan–had their share of employment in manufacturing increasing much faster than that of the tertiary sector during their initial period of growth in the 1970s. In the next decade tertiary-sector employment grew faster, but the magnitude of the increase relative to manufacturing was not nearly as high as was observed in India during this decade. Only in the 1990s, after Taiwan and Korea had developed into mature industrialized economies, did their tertiary sector become the dominant provider of employment outside agriculture. By contrast India's share of employment growth in the tertiary sector in the seventies was already 60 percent higher than in manufacturing. Since then, the decades of 1980s and the 1990s have seen a virtual stagnation in the share of employment in manufacturing, with the tertiary sector absorbing virtually the entire loss of employment share by the agriculture. The figures also show that other developing countries of Asia–Thailand, Malaysia and Indonesia–do have their larger shares of employment created in the tertiary sector, but the contrast with India is that none of them have a stagnant share in manufacturing in any decade. On the contrary, something between a third and one half of the often large decline in the share of employment in agriculture was taken up by manufacturing. The only country in the sample with an experience close to that of India is the Philippines.

The tertiary sector has been the leading sector of growth in the Indian economy in recent decades, both in terms of output and employment (Table 10.4). The employment elasticity in the sector as a whole in the post-reform period (1993–2000) has been 50 percent higher than in manufacturing sector. Is this growth due to labor being pushed into the sector because of limited growth of jobs in the productive sector or due to labor being pulled into it because of increasing earnings? Are there different trends in different components of the tertiary sector, and between the formal and informal segments of it? What light do the trends in the tertiary sector throw on the process of equitable growth in India?

Table 10.1 Change in the sectoral shares of employment

We should mention at the outset that the Indian statistical series do not allow for the construction of time series of employment and output by formal and informal sectors, however defined. Hence the substance of our analysis in this part will be based on the study of trends in the tertiary sector as whole. We will address the question of absorption of labor in this sector at low- and high-income levels, as well as the earnings gap between 'good jobs' and 'bad jobs' in the sector by looking at the entire distribution of earnings in the sector. But before we come to this analysis it would be useful to give an overview of the structure of employment in the tertiary sector at one time period, i.e., 1999–2000. The 55th round of the NSS, however, included some questions which provide criteria for distinguishing the formal and the informal sub-sectors within the tertiary activities. The broad structure of tertiary employment will be clear from these data.

Formal and informal sub-sectors within the tertiary sector

The 55th round questionnaire obtained information on the type of establishment in which a worker was employed. We grouped workers in all public and semi-public establishments as being in the formal sector. This round of the NSS also reported for the first time the employment size of the establishment in which a worker was employed. We take ten workers as the cut-off point, with those in establishments with ten or more workers being in the formal sector. For the large group of self-employed workers we adopt the usual classification in terms of the workers' education. Those with lower secondary education or less are in the informal sector, and the more highly educated (which would include the professionals) are in the formal sector. These criteria help us to give a rough picture of the composition of tertiary-sector employment for the year 1999–2000 across formal and informal sectors (Table 10.2).

The following points emerge from Table 10.2:

1. The formal sector accounts for a quarter of tertiary employment in the rural areas and rather more than a third in the urban economy. Overall it accounts for 30 percent of all tertiary-sector employment.
   
   
2. Even after the decline of public-sector employment in the post-reform period, this sector still accounts for more than half of formal tertiary employment in the urban areas, and more than two-thirds in the rural.
   
   
3. Females account for a small part of tertiary employment in the formal sector, and surprisingly no more than 10 percent of informal tertiary employment, both in the rural and the urban areas. It should, however, be remembered that we included only UPS workers (principal workers).
   
   
4. The share of the self-employed in the non-public part of the tertiary employment is high, but contrary to expectations it is higher in the formal sector of both the rural and urban economies.

It will be interesting to know how the levels of employment in the formal and informal segments of the tertiary sector compare with those in manufacturing.

Table 10.3 throws light on this question. It is seen that three-quarters of all employment outside agriculture and construction are in the tertiary sector and this percentage is only slightly more in the urban areas. As is to be expected, a larger proportion of tertiary employment is in the formal sector in urban areas. But the rural areas still have a good deal of formal-sector presence.

Employment elasticites by broad sectors

We presented the basic tables in Chapter 3 on employment trends by broad sectors of the economy (Table 3.1) Table 3.1 combined output trends calculated from the National Accounts Statistics with employment trends obtained from the NSS which provides an overview of employment elasticities over a time period for sectors at 1-digit National Industrial Classification (NIC 1987). The employment estimates are based on Usual Principal Status Workers (UPS).

The major points to emphasize from these tables are:

1. Tertiary-sector employment grew faster than manufacturing in all three periods. The differential in the growth rates was much higher with respect to agriculture, particularly between the 50th and the 55th rounds. However, we should remember that the employment growth in the last period was disproportionately affected by the fall in employment-growth rate in the agricultural sector.
   
   
2. Employment growth in the tertiary sector fell in the second half of the nineties relative both to the 1987–1993 period and even the longer 1983–1993 decade. But this was entirely because of the decline in employment in the community and social services. The table shows that compared with the 1983–1993 decade, the decline in employment growth was marginal in financial services. The rate of growth of employment increased in all other groups, particularly strongly in trade.
   
   
3. Employment elasticity mirrored the story of employment growth. The employment elasticity fell slightly in all the other tertiary sectors but was in the last period well above that in manufacturing.

Productivity differentials between sectors

Is the employment growth in the tertiary sector being driven by high demand for labor or is labor entering this sector because of lack of jobs in other production sectors. In other words, is labor being pulled or pushed into this sector? A first cut at this question is to see if there are major productivity differentials or if the productivity differential increasing vis-à-vis the production sectors as revealed by sectoral GDP figures. The data given in Table 10.4 gives an initial answer to this question.

1. The average productivity in the tertiary sector as a whole is pulled up by the high value in the financial sub-sector, but seems to be above the level of manufacturing (in 2000) in most sectors except trade (where it is 20 percent lower). There is a suggestion that the trade – manufacturing differential might have slipped over time. Between 1983 and 2000 productivity in trade relative to its base (agriculture) remained practically constant (in real terms) but went up by more than 40 percent in manufacturing. This allowed manufacturing productivity to go significantly above trade, but it is interesting to see that this differential was established only recently – between the 50th and the 55th rounds.
   
   
2. Not all sub-sectors of tertiary, however, suffered the fate of NIC sub-group 6, i.e., trade. Both finance (group 8) and community and social services (group 9) improved their relative productivity vis-à-vis manufacturing. In the transport services (group 7) the relative improvement of productivity seems to have been under way since the 43rd round. But in the community and social services (group 9) the relative improvement was prominent only between the 50th and the 55th rounds. The surge in salaries in the public sector is reflected in the large increase in productivity between these two rounds in this group.
   
   
3. The above pattern suggests that there is indeed some evidence to support the general perception that some sub-groups, like the trade (group 6), have had a relatively large influx of labor pushing down its relative productivity to some extent, while others, like business services in group 7, have improved their position due to demand factors.

However, a study of trends in average relative productivity can carry us only so far in our understanding about the trends in relative earnings at which labor is being absorbed in the tertiary sector. For a more complete understanding we need to look at the way the entire distribution of earnings (or incomes) have been changing in the tertiary sector in response to the high rate of growth of employment in this sector.

Before getting into further analysis of the tertiary sector on the basis of unit-level data it will be worthwhile to discuss the limitations of the database that we have used.

Limitations of the NSS data

We need to be aware of the limitations of the main source of our data, the NSS, before proceeding further. First, a large share of employment in India is in the 'self-employed' category. There is an inherent difficulty of allocating income accruing from self-employment when more than one earner from the same household is in income-earning activity. Households from different self-employed activities by different members of the household would be typically pooled together. There is no way of distinguishing the individual contributions of individual earners. Hence the income we can deal with is household income, and we can normalize for the size of the household. Further, it is generally accepted that figures on expenditure given by the respondent in the household is more reliable than that of income. Thus we use the measure of household welfare as given by mean expenditure per capita.

When we are comparing levels of household welfare across sectors we need to identify the principal occupation of the household. This poses problem both conceptually and in terms of execution. The conceptual problem arises from the fact that a significant number of households will have more than one earner, and not all earners will be in the same category of occupation. The secondary earners might not be all wage earners. If they are working in the self-employed sector, they will be pooling their earnings with other earners of the household to create the household's pot of earnings. By assigning all the household income effectively to the principal occupation of the household we might be exaggerating the income – and the expenditure which it sustains – originating from this occupation.

In terms of execution, one of the major problems faced in the 55th round of the NSS is that, unlike in the earlier rounds, households were not classified in terms of their detailed occupational or industrial code of their main source of earnings. We first have to match household type (given in household file) to the individual workers' file which provides the code for occupation, industry, work status, etc. We generated household type for each individual worker. Thus through an arduous process we could identify main earners in most of the household and then assign main earners' industry – occupation code to the household's main earning source. The occupation – industry distribution of households will differ somewhat from that of individual earners to the extent that our matching has been unsuccessful particularly in households where more than one principal earner belongs to a different industry – occupation. The difference in the proportions of employment in the tertiary sector obtained on the basis of households and two definitions of the individual worker (usually principal and usually principal-cum-secondary status) are given in Table 10.5.

Evidence on the marginal absorption of labor

We can get some idea about the question posed – how far labor is being pulled rather pushed into the tertiary sector – by looking at the share of labor in the tertiary sector at different parts of the distribution of income. Specifically, we can look at the proportion of the main earners working in the tertiary sector in different quintiles of the distribution of household expenditure per capita for successive rounds.

Table 10.6 gives the share of household employment across different rounds. It shows that the share of tertiary sector in household employment increased over the successive rounds. Table 10.7 seeks to throw light on the question as to where the jobs were created – at the low end or uniformly across household quintile ranges.¹ The data are presented in Figure 10.1 which shows the changes in the distribution more clearly, separately for the rural and the urban areas.

A major change seems to have taken place in the post-liberalization period (between 50th and 55th rounds) both in the rural and the urban areas, compared with the movement in the pre-liberalization period (between the 38th and the 50th rounds). In the earlier pre-liberalization years more jobs in the tertiary sector seem to have been created in the higher quintiles. The slopes of the graphs increased with the quintile groups between 1983 and 1993 (the 38th and the 50th rounds)–more prominently in the rural areas, and except for the highest quintile in the urban economy. But between 1993 and 2000 (the 50th and the 55th rounds), the graph for the rural sector shows a more or less parallel movement outwards, with some suggestion that the movement was larger in the 1–2, as well as the 5th quintiles. In the urban sector the differential movement by quintile groups was quite striking at the two ends of the distribution. There is a sharp increase in the share of tertiary earners both at the lower (2nd) and the highest (5th) quintiles at the expense of the middle (3rd and 4th) quintiles.

Figure 10.1 Employment share of the tertiary sector by quintile groups, different rounds.

Note
Panel A is rural and panel B is urban.

The fact that more tertiary-sector employment has been created at the lower quintiles does not mean that there has been immiserizing growth of the tertiary sector in the sense that labor pushed into this sector has depressed earnings in the sector absolutely. The mean of the distribution might have increased over the period. There is a suggestion that the distribution of incomes in the sector might have deteriorated, particularly in the urban areas, with the incomes of the low earners falling relative to the high earners. But to shed more light on this specific question we need to look directly into the changes in the distribution of income (or household welfare in our case). This we do in the next section.

Evidence on the distribution of average per capita consumption expenditure (APCE) in the tertiary sector

The Kernel density functions for the three rounds have been graphed, separately for the rural and the urban areas in Figure 10.2. Both the distributions have shifted to the right in the post-liberalization years – much more perceptibly so in the post-liberalization years than between the previous two rounds. Further the outward movement is more striking in the urban economy. This is our first important conclusion: in spite of tertiary-sector jobs being created disproportionately in the lower quintiles, particularly in the urban areas, the evidence suggests that levels of earnings have gone up significantly including at the lower part of the distribution.

The graph also confirms what has been suggested by the evidence discussed in the last sub-section: that there has been some increase in the inequality in the distribution in the urban sector – perhaps not at all in the rural economy. Further information on the changes in distribution can be found from the decile and quartile ratios reported in Table 10.8.

The conclusions emerging from two tables are as follows:

1. As far as the rural areas are concerned there has been a decided improvement in the distribution. Inequality decreased in magnitude in the lower half   of the distribution – judged both by the decile and the quartile ratios. There has been a smaller improvement in the top half; both the P90/P50 and the P75/P50 ratio moved down a bit.
   
   
2. In the urban economy, there is an evidence of the deterioration in the distribution at the lower part of the distribution. The P10/P50 ratio deteriorated particularly between the 50th and the 55th rounds – when we saw there was such a pronounced increase in the absorption of labor in low-income tertiary jobs. But the deterioration is not by any means large.

Figure 10.2 Kernel density functions of APCE in the tertiary sector, different rounds.

Trends in poverty and inequality in the post-liberalization years

It has been noted in the earlier chapters in Part I that, while the incidence in poverty has fallen both in the rural and the urban areas in the post-liberalization years, the reduction in poverty in the urban economy has been accompanied by a perceptible increase in inequality (see Chapter 2). The graphs of APCE given in Chapter 3 (Figures 3.4a and 3.4b) clearly bring out the change between the successive NSS rounds in the urban and rural sector.

The material presented in the two previous sections above suggests that the increase in inequality in the urban sector (and not so much in the rural) has been driven by the trends in the distribution of incomes in the tertiary sector. The point has relevance to the wider literature on the impact of liberalization in inequality.

It has been expected on the basis of standard trade theory of the Heckscher–Ohlin type that greater openness of an economy would tend to increase the relative returns to those factors of production which are in abundance in the economy concerned. Thus a less developed economy, where labor rather than capital is the more abundant factor, will see an increase in the relative return to labor – leading to a more equitable trend in the distribution of income. The experience of many developing countries after the recent spate of liberalization has, however, belied this expectation. Economists have tried to explain the observed increase in inequality in less developed economies by modifying the Heckscher–Ohlin model to allow for the inclusion of two types of labor – skilled and unskilled. Liberalization in this extended model leads to an increase in demand, not of unskilled labor but of more skilled labor which is demanded by the manufactured products in the sector open to international markets. In other words the industries which have a spurt in growth following liberalization demands labor of a type which might be less skilled than labor in manufactured goods produced by advanced countries, but they are more skilled than the general mass of unskilled labor which is in abundant supply in less developed countries. Thus the increase in skill differential in the latter drives the observed increase in inequality (Acmogolu, 2002).

The discussion in this chapter suggests that the mechanism described in the literature would be more pertinent if we incorporate the tertiary sector in the discussion. In other words the relative increase in demand for more skilled labor after liberalization comes as much, if not more, from the growth of some parts of the tertiary sector, as from the traded manufacturing sector. Clearly this effect can come only from the sub-sectors of the tertiary activities which deal with services to the globalized part of the economy. These contrast with those branches of the tertiary sector which are 'non-traded' catering to the needs of the domestic economy. As far as the latter are concerned, we would like to know if they show any evidence of 'immiserizing growth' which the aggregate view of the tertiary sector does not reveal–i.e., is labor being 'pushed' into the sector with falling incomes because of lack of opportunities in the production sectors.

The next section, therefore, goes into a discussion of trends in income distribution in different branches of the tertiary sector.

Shifts in the KDF distribution in different sub-sectors of tertiary activity

How do the shifts in the APCE distribution compare in different sub-sectors of the tertiary activities? We can go a fair distance by looking at the picture for the four major one-digit sectors distinguished in the National Industrial Classification (NIC). This is done in Figure 10.3.

The NIC Group 8 (business services) would contain the bulk of the services catering to the traded part of the economy, while group 6 (trade, hotels and restaurants) would comprise the bulk of the private non-traded services. Group 9 includes community, social and personal services, but is also heavily represented by government activities, including administration.

Two points stand out in the picture presented in Figure 10.3. First, the shift in the distribution between the two rounds is more pronounced for the urban areas than the rural ones even when we look at the disaggregated tertiary sub-groups.

Figure 10.3 Kernel density functions by major sub-groups of the tertiary sector.

Note
6: Trade, hotels and restaurants; 7: transport, storage and communication; 8: finance, real estate and business activities; 9: community, social and personal services.

Second, the shift is least for the NIC group 6 (trade, hotels and restaurants) in both the rural and the urban areas, and the most striking for groups 8 (business services) and 9 (community, social and personal services). Further, in the groups showing the larger outward shifts, the shift in the urban areas is more prominent. Nowhere is there any evidence of any increase in the incidence of low-income groups.

KDF distributions for regular wage earners in the tertiary and other sectors

It might be useful to look at the KDF functions for the three rounds exclusively for regular wage earners (see Figure 10.4). The incomes of these respondents are more easily obtained in the NSS survey. A study of the change in the distribution of their earnings over the three rounds of the survey is a useful supplement to the changes in the household welfare by the classification of 'main earners' presented above.

Figure 10.4 KDF distributions for regular wage regions by major sector, and rural and urban areas: three rounds.

Note
Panel A is rural and panel B is urban.

Two points need to be emphasized:

1. There is a rightward shift in the KDF in the successive rounds for both the rural and the urban areas, but it is clear that the shift is largest for the tertiary-sector regular wage earners. The ordering of the primary and secondary sectors are, however, rather different for the rural and urban areas. In the urban areas the shift seems to be larger for the primary rather than the secondary sector, presumably because of the development of different types of high-value primary activities. In the rural areas, however, the outward shift in the secondary sector is more pronounced relative to the primary sector.
   
   
2. The shape of the KDF in the tertiary sector is altered in rather the same way in the rural and the urban areas in the later rounds, even though the movement is stronger for the urban economy. There is marked flattening of the curve suggesting a wider dispersion of earnings and larger proportions of workers with higher earnings. There is a clear reduction of the proportion of people with low earnings, but interestingly both in the rural and urban sectors, the mode seems to have moved to the left (even though much reduced in its density). This might suggest that there is a sizable influx of low wage workers – earning rather less than in the 38th round in real terms. However, this phenomenon might really mean that there is a larger influx of younger or less educated workers along with others who earn much more.

The last point carries with it an implication that "dualism' has increased in the tertiary sector, and might indeed be stronger in the tertiary than in the secondary or manufacturing sectors. We cannot be sure about this hypothesis unless we control for the quality – in particular the human-capital attributes – of the workers entering these sectors.

Is 'dualism' higher in the tertiary sector? Earnings differentials (net) as between sectors in different points of the distribution

Our purpose is to know how the earnings in the tertiary sector relative to the earnings in the other two sectors, in particular manufacturing, vary in different parts of the distribution. 'Dualism' in terms of the gap between low and high earners in manufacturing is high in the Indian economy and has also been discussed in Chapter 9. If the dualism is stronger in the tertiary sector, then we would expect to find the 'net' tertiary-manufacturing differential, after controlling for the other major determinants of earnings (like human-capital attributes) to increase as we move up the scale in the earnings distribution. We ran quantile regressions for the 55th round of the NSS to estimate the net differential at the five quintiles of the distribution. Dummies for the sectors (with primary sector as base) were used in the regressions along with a set of other explanatory variables. The latter included education, age, sex, urban–rural location and regions.

Figure 10.5 Estimated coefficients of (dummy) variables from quantile regressions: APCE.

The exercise was done separately for the APCE of households (for in which the characteristics of the 'main earner' were used for the explanatory variables) and for the daily earnings of regular wage earners. There were some differences in the sets of explanatory variables used in each case. (Model description is given in the Appendix.)

The coefficients of the tertiary and manufacturing sector dummies at the different quintiles are given in Table 10.9, and they are graphed in Figures 10.5 and 10.6. There are apparent differences in the shapes of the distribution. This is primarily because for the wage sector secondary wages are below tertiary wages (remembering that the base in each case is primary sector earners), while for the APCE of households the values for the tertiary and the secondary sectors are all above the primary. This rather intriguing difference is probably because secondary wage earners in the middle range of the distribution (q25 to q75) earn less than those in regular primary employment. The relatively high wages observed in the latter are due to public-sector and similar government employment in the primary sector.

Figure 10.6 Estimated coefficients of (dummy) variables from quantile regressions: regular wage earners.

But as far as the tertiary–secondary differential is concerned the results are the same for APCE and daily wages. The differential is all along higher for the tertiary-sector workers. The gap between the two sectors increases in the middle range and diminishes somewhat only at the highest quarter of the distribution.

We conclude that dualism is quantitatively more important in the tertiary sector when we compare the earnings of the lowest quintile with those in the higher quintile – except that the difference is reduced for the highest quintile. There is then some support for the popular perception that the tertiary sector is home to a body of low earners more so than the secondary sector.

Conclusion

The structure of employment observed in the NSS survey year of 1999–2000 (the 55th round) shows that the formal sector accounted for a quarter of tertiary employment in the rural areas and one third in the urban areas. Even after the decline in public-sector employment in the post-reform period this sub-sector still accounts for more than half of formal tertiary employment in the urban areas and more than two-thirds in the rural. Around one-half of employment in the informal segment of the tertiary sector is accounted for by the self-employed in both the areas. Regular wage earners are more important in the urban sector, the rest (25 percent in the urban, and 33 percent in the rural) being casual wage-workers.

In the absence of time-series data for the formal and the informal sectors we are obliged to analyze the trends in the low- and high-paid employment in the tertiary sector by looking at the changes in the entire distribution of earnings in this sector over time. We have looked at the issue from several angles and for different variables representing income levels. As mentioned the self-employed constitute a very large part of the tertiary sector. By definition the individual earnings of the self-employed are not recorded for each worker. All the earnings of the household members are pooled together. The variable most relevant to look at, then, is a measure of household welfare – which in the simplest formulation is mean household per capita expenditure (APCE). The industry affiliation of the household is given by the occupation of the main earner. This may create some errors for multiple-earner households whose earners follow different occupations.

The movement of the distribution of APCE for the successive rounds brings out two important points: (i) there is an outward shift in the distribution in the tertiary sector, so that earnings at all levels have increased; and (ii) there has been proportionately larger increase in the number in the first and the fifth quintiles of the distribution – with relatively less absorption of labor in the middle range. This implies an increase in inequality in the bottom half of the distribution – a trend more prominent in the urban economy. Disaggregating the tertiary sector by its 1-digit components, it is seen that these effects are mild in trade (group 6) but much more striking in business services and in the community and social services,

We looked specifically at regular wage earners whose individual earnings are recorded. The outward movement of the earnings distribution over successive rounds (and particularly during the 1987–1993 and 1993–1999 periods), as well as the 'flattening' of the curve, is more striking for the tertiary sector than either the primary or the secondary. It is also more prominent for the wage-earners than the welfare index for all tertiary households (APCE) which we had used. Thus we conclude that while there is no evidence for the incidence of low incomes in the tertiary sector to increase in any absolute sense, more jobs are being created in the bottom and the topmost part of the distribution.

This last point suggests an increase in 'dualism' in the tertiary sector. We have seen in the last chapter that dualism was particularly striking in Indian manufacturing compared with other Asian economies, and it had most likely increased in recent years. Our quintile regression analysis was meant to see how the earnings differentia between tertiary and the manufacturing sectors compare at different parts of the earnings distribution. The results for the 1999–2000 round of the NSS show that the differential, after controlling for human-capital attributes and location of the labor, increases from the lowest quintile to the fourth – and only in the highest is there some reduction in the 'net' differential. This is true for both the APCE measure and for regular wages. We conclude that dualism has become higher in the tertiary sector than in manufacturing.

Appendix

In the last section, both sets of regressions were simultaneous quantile regression with bootstrapping standard errors. The quantile regressions were simultaneously run at five quantile points namely 5, 25, 50, 75 and 95.

Both regressions were based on NSS unit level data of 55th round. The regression with APCE as dependent variable was estimated at household level and it had 92,282 observations. The regression with wage of the regular workers as dependent variable was estimated at individual level and it had 52,439 observations.

In the following tables, we present variable descriptions of both regression models.

In addition, all independent variables were interacted with urban to control urban influence on them.

Dependent Variable: ln (APCE).

In addition, all independent variables were interacted with urban to control urban influence on them.

Dependent Variable: ln (Wages of Regular Workers).

Document(s) 10 of 16