ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Regrouping of Economic Activities

The authors describe the manner in which the Central Statistical Organisation regrouped economic activities according to a new classification scheme which could be used to compile national accounts statistics with the new base year 1999-2000. This classification scheme was particularly important for the service sector of the economy. The justifiability of the classification scheme is discussed with reference to the relative standard error in estimating gross value added for each group of activities, at all-India and state levels, and for rural and urban areas.

Regrouping of Economic Activities

A Feasibility Study

The authors describe the manner in which the Central Statistical Organisation regrouped economic activities according to a new classification scheme which could be used to compile national accounts statistics with the new base year 1999-2000. This classification scheme was particularly important for the service sector of the economy. The justifiability of the classification scheme is discussed with reference to the relative standard error in estimating gross value added for each group of activities, at all-India and state levels, and for rural and urban areas.

PURNENDU K BANERJEE, AJAY BAKSI, NILANJANA ROY

I Introduction

T
he internal working group (IWG) of the Central Statistical Organisation (CSO) was formed in February 2002 to devise a scheme of classification of economic activities that could be adopted for estimation of gross domestic product (GDP) and other macroeconomic aggregates based on the National Industrial Classification 1998 (NIC 98). The new classification scheme was also needed for compilation of the revised series of national accounts statistics (NAS) with changed base year 1999-2000.

The IWG study was conducted for one segment of the economy where the production method of GDP calculation cannot be used for want of requisite data. “Labour input method” is used for this segment, consisting mainly of the unorganised sector of the economy. The IWG used data from the National Sample Survey (NSS) 55th round employment-unemployment survey (EUS) and enterprise survey (ES) to arrive at the activity groups. A major advantage of using these estimates was that both the surveys were conducted at the same time point with identical scheme of selection of the first stage units.

In the 55th round, nearly 1,98,000 enterprises were surveyed, covering all the non-agricultural activities. However, in this survey, as expected, the sample enterprises were mostly pursuing manufacturing or trading activities. As a result, the service sector activities were not always well represented in the sample. Now, using a small data set may sometimes show a particular pattern which may not reflect the true population characteristics. However, only two years later, during the 57th round, the NSS conducted a survey exclusively on the enterprises belonging to the service sector (excluding trade and finance) with a bigger sample using a sampling scheme focused on the ES. More than 2,40,000 sample enterprises were covered in this survey, as opposed to about 66,000 service sector enterprises during the 55th round.

As, at the time of the study made by the IWG, the results of the 57th round were not available, we decided to put the classification scheme proposed by the IWG related to the service sector under the scanner of the results obtained from the current round. Our primary objective was to find out whether the economic activity groups at all-India level as suggested by the IWG satisfy norms like reasonable relative standard error (RSE). The next subject of investigation was, at what level of disaggregation the estimate of a given activity group can be provided with tolerable error. For example, if we find that the all-India level estimated gross value added (GVA) of activity X is within the set limit of RSE (say, 25 per cent), then we have tried to find out first whether the estimate of GVA for X can be provided for each state/union territory (UT) with reasonable accuracy, and if so, whether we can produce it for the rural and urban areas separately. In the case of the activity groups for which it is possible to generate estimates at all-India level, we tried to check whether it is possible to do the same at a more disaggregated level. It may be noted that, studying the variations in estimated GVA of each of the service sector activities, the IWG, using the variations in the estimates observed during the 55th round, suggested 28 groups of economic activities for this sector.

Given the fact that the sample number of service sector enterprises in the 57th round was about four times the sample size of the 55th round, it is expected that at the same level of disaggregation, the results based on the 57th round captured the inherent and subtle variations of the enterprises belonging to a particular group of activities more reliably.

II IWG Methodology

As already mentioned, the IWG formed productivity homogeneous groups of economic activities whose estimates of worker population ratio (WPR) and gross value added per worker (GVAPW), as obtained from the EUS and ES of the NSSO respectively, were mostly within a specified sampling error. Broadly, its method (called the labour input method) can be expressed in notation as

GVA = P × WPR × GVAPW

where P = census population (adjusted appropriately).

The grouping of the activities was done by the IWG keeping in mind the following. 1 The RSE(GVA) was within 15 per cent. 2 Comparability of the activity group level GVA estimates with those of the earlier series was retained.

Economic and Political Weekly September 15, 2007 3 Activities with very different productivity (in terms of VAPW) or those growing at uneven pace were not combined. Particularly, the emerging or fast growing industries were kept separate from other activities.

III Methodology of the Present Study

All the activities surveyed during the 57th round of the NSS, grouped according to the grouping scheme of the IWG, were considered for the current study. To compute the RSE of GVA for each of the activity groups, we have used two sources of data as follows.

We have used the RSE of WPR from the 55th round data on EUS. It may be noted, firstly, that the number of households surveyed during the 55th round for the employment enquiry was much greater than the number surveyed during the 57th round. Also, one expects that there will not be much fluctuation in the WPR over a short period of time. Hence, the RSE of WPR based on the 55th round employment survey is expected to be better than the RSE of WPR based on the 57th round.

We have computed the RSE of GVAPW using the ES data of the 57th round. For computing the RSE, the variance of the ratio estimate was computed at each sub-stratum level using the interpenetrating sub-sampling technique. We have considered those activity groups for which the sample size at the all-India level was 1000 or more.

Derivation of the Formula for RSE(GVA)

Detailed derivation of the formula is given in Appendix 1. As convention dictates, we present all the RSEs in percentage form. For this, we will have to multiply both sides of the formula by 100. If we denote the RSE in percentage form as rse (rse = 100 × RSE) then

RSE(GVA)*100=

100*(RSE2 (WPR) + RSE2 (VAPW) + RSE2 (WPR)*RSE2

(VAPW))1/2 rse(GVA)=

100*(RSE2 (WPR)+RSE2 (VAPW)+RSE2 (WPR)*RSE2

(VAPW))1/2 (as an example, in percentage terms, if rse(GVAPW) = 5 and rse(WPR) = 5, (or RSE (GVAPW) = 0.05 = 5 per cent and RSE(WPR) =

0.05 = 5 per cent), then rse (GVA)=100*√(0.052+0.052+0.052*0.052)=7.08)

Here we have defined WPR (PS + SS) of any compilation category specific as WPRi = 1000* No of workers as per PS + SS status in the category i of given sector/total population of that sector. Here both numerator and denominator were estimated from NSS 55th round EUS data. Hence the estimate of WPR is a biased estimate but the bias is negligible. Hence the respective RSE (WPRi ) have been approximated by taking the square root of sample MSE of the sample ratio/estimate of the ratio. While calculating MSE an approximated formula using Taylor series expansion has been used. However, all the calculations are done at the stratum level. The same technique is normally followed in all NSS ratio estimates. Similarly RSE(VAPW) is also calculated for each category of activity from another independent NSS 57th round survey.

On the basis of RSE(WPR) and RSE(GVAPW) obtained from respective sources as mentioned in the beginning of the section we have calculated the RSE(GVA) using the formulae derived above.

We have carried out our study at three stages. Stage 1: At all-India level (rural and urban areas together) Stage 2: At all-India level separately for rural and urban areas Stage 3: At state/UT level separately for rural and urban areas

The different activity groups considered in our study are shown in Table 1.

There are some minor deviations in the groups mentioned above from the grouping suggested by the IWG. These are: IWG has proposed ‘851 + 852’ as one activity group. We have used RSE(WPR) of ‘85’ as a proxy of RSE(WPR) of activity group ‘851+852’ since RSE(WPR) of ‘851+ 852’ was not separately available.

We did not consider the activity group ‘91+853’ for the present study.

We have divided the activity group ‘73+7413+7414+742+74374991’ into two parts, ‘73+7413+7414’ and ‘742+743+74974991’, and tried to verify whether the RSE(GVA) is within tolerable limits for either of these two activity groups or not.

To study the behaviour of homogeneous activity groups as given in Table 1 with respect to 57th round data, we have set some specific limits for RSE(GVA) at different levels of disaggregation. The following are the limits at three stages.

Table 1: Activity Groups Considered in the Study

No NIC 98 Activity Group

1 551 Hotels, camping sites 2 552 Restaurants, bars and canteens 3 6021 Scheduled passenger land transport 4 60221 Non-scheduled passenger land transport by

motor vehicle 5 60222 Other non-scheduled passenger land transport 6 60231 Freight transport by motor vehicles 7 60232 Freight transport other than by motor vehicles 8 61 Water transport 9 6302 Storage and warehousing

10 63-6302 Supporting and auxiliary transport services

11 6412 Courier services

12 64204 Cable operators

13 642-64204 Other communication

14 70 Real estate activities

15 71 Renting of machinery and equipment

16 72 Computer and related activity

17 7411 Legal activity

18 7412 Accounting, bookkeeping

19 73+7413+7414 Research activities

20 742+743+749-74991 Architectural, engineering, technical, business other than auction activity

21 80903+80904 Coaching, activities of individuals providing tuition

22 80-80903-80904 Other educational activities

23 851 + 852 Human health activities, veterinary activities

24 90 Sewage, sanitation and related activities

25 92 Radio, television, cultural, recreational activities

26 9301 Washing and cleaning of textile and fur products

27 9302 Hairdressing and other beauty treatments

28 9303+9309 Funeral and related activities

Table 2: Activities for Which RSE(GVA) Exceeded 15 Per Cent

Activity No NIC 98 Description rse(GVA)

8 61 Water transport 20 9 6302 Storage and warehousing activity 16 11 6412 Courier services 24 19 73+7413+7414 Research and management consultancy 20

Economic and Political Weekly September 15, 2007

Stage 1: For all-India (considering both rural and urban areas together) we tried to identify the activity groups whose RSE(GVA) was (1) within 15 per cent and (2) within 25 per cent. Stage 2: Then, while evaluating the performance of the RSE(GVA) for each of the activity groups, separately for the rural and urban areas at all-India level, we considered a 25 per cent limit tolerable. Stage 3: Finally, for each activity clearing Stage 2 above, we computed the RSE(GVA) at the state level, separately for rural and urban areas, and checked its efficacy with respect to the tolerance limit for RSE (GVA) at 50 per cent.

IV Results of the Study

As already mentioned, the study on the behaviour of the rse of aggregate GVA was examined for 28 activity groups in three different stages. The results, as observed, are depicted below.

Stage 1

At the all-India level, considering all the enterprises together, as also separately for the enterprises not registered under the Companies Act, 1956 (this may be considered a better representative of the unorganised sector), RSE(GVA) was within 15 per cent for all the activity groups except those shown in Table 2.

Research and management consultancy was considered together with architectural, engineering, technical, business other than auction activity, as a single activity group in the recommendations of the IWG. Even if we break this up into two separate activity groups as spelt out, estimates of aggregate GVA fall within the tolerable limit of sampling fluctuation (15 per cent RSE) for architectural, engineering, technical, business other than auction activity.

At all-India level, the RSE of GVA was highest for courier services (6412) at 24 per cent. Water transport (61) too failed to clear the 15 per cent level of RSE(GVA). Keeping these two as separate activity groups is therefore not supported by the 57th round data if sampling fluctuations are to be kept within tolerable limits. For water transport, the survey was unable to capture enough enterprises as the usual unorganised sector surveys of the NSS do not have any separate stratum for the coastal areas or for areas lying beside rivers. The same cannot be said, however, for courier services, as the number of such enterprises in the all-India sample size was more than 1000. It could be that sampling fluctuations here remained large as this is still an emerging area where the enterprises’ activities are yet to stabilise. As we shall see subsequently, the rse(WPR), particularly for the rural areas, was very large for courier services. Probably, for the emerging areas, our usual method of estimation can provide good results if we conduct regular focused surveys on this.

Table 4: Activities for Which rse(GVA) Exceeded the Predetermined Limit

No NIC 98 Description rse(GVA)
Rural Urban
8 61 Water transport 30 28
9 6302 Storage and warehousing 33 29
11 6412 Courier services 58 28
Table 5: Activities for Which rse(GVA) is Acceptable for Urban
But Not Rural Areas
No NIC 98 Description rse(GVA) Urban
1 551 Hotels, camping sites 15
12 64204 Cable operators 19
14 70 Real estate activities 18
16 72 Computer and related activity 11
18 7412 Accounting, bookkeeping 13
Table 6: Difference between rse(WPR) and rse(GVAPW) for
Activities for Which RSE(GVA) Exceeds 25 Per Cent
No NIC 98 Description rse(WPR) - rse(GVAPW)
Rural Urban
1 551 Hotels, camping sites 27 4
8 61 Water transport 6 23
9 6302 Storage and warehousing 18 22
11 6412 Courier services 48 1
12 64204 Cable operators 39 17
14 70 Real estate activities 28 11
16 72 Computer and related activity 32 0
18 7412 Accounting, bookkeeping 20 8

Table 3: Activity Groups for Which RSE(GVA) Was Less than 25 Per Cent

No NIC 98 Description 2 552 Restaurants, bars and canteens 3 6021 Scheduled passenger land transport 4 60221 Non-scheduled passenger land transport by motor vehicle 5 60222 Other non-scheduled passenger land transport 6 60231 Freight transport by motor vehicle 7 60232 Freight transport other than by motor vehicle 10 63-6302 Supporting and auxiliary transport services 13 642-64204 Other communication 15 71 Renting of machinery and equipment 17 7411 Legal activity 19 73+7413+7414 Research, management, consultancy 20 742+743+ 749-74991 Architectural, engineering, technical, business other than auction activity 21 80903+80904 Coaching, activities of individuals providing tuition 22 80-80903-80904 Other educational activities 23 851 + 852 Human health activities, veterinary activities 24 90 Sewage, sanitation and related activities 25 92 Radio, television, cultural, recreational activities 26 9301 Washing and cleaning of textile and fur products 27 9302 Hairdressing and other beauty treatments 28 9303+9309 Funeral and related activities Rural 6 10 8 9 6 9 24 18 9 23 NA 17 12 5 9 20 16 11 8 8 rse(GVA) Urban 5 10 5 10 13 14 14 9 11 9 24 9 9 5 6 21 10 11 13 12
Economic and Political Weekly September 15, 2007 3747

As RSE(GVA) was less than 25 per cent for all the activity groups, we proceeded to Stage 2 to examine the nature of RSE(GVA) at the next level of disaggregation. One reason for continuing with all the activity groups is that, even when performance of a particular group is not good at the all-India level, it is quite possible that it may exhibit low RSE(GVA) in one sector and high RSE(GVA) in another.

Stage 2

As most of the activity groups had shown an RSE(GVA) of less than 15 per cent and all the 28 activity groups had satisfied the 25 per cent level of RSE(GVA) at all-India (combined) level, we proceeded to examine whether it was also possible to get estimates for them at all-India level separately for rural and urban areas. We considered that at this level, obtaining estimates with RSE(GVA) less than or equal to 25 per cent would be quite good. It may be mentioned that, RSE(GVA) exceeds this value even when, say, RSE(WPR) is 15 per cent and the corresponding RSE(GVAPW) is 20 per cent.

For the 20 activity groups shown in Table 3, RSE(GVA) was less than 25 per cent for both rural and urban areas at all-India level.

If we analyse the performance of RSE(GVA) separately for the rural and urban areas, we observe that, in general, the estimates for urban areas had lower RSEs than their rural counterparts. The RSEs of rural estimates are lower for most of the transport segment and a few traditional service activities, with the sole exception of hairdressing and beauty parlours. For almost all the activities which are newly emerging, the urban estimates have performed better in terms of RSE(GVA) than their rural counterparts. For all activities where RSE(GVA) of rural areas were more than RSE(GVA) of urban areas (10 among the 20 mentioned above, including research), the RSE(WPR) is more than the RSE(GVAPW) in the rural areas. These activities probably cannot be adequately captured in a generalised survey on employment and unemployment as they were found to be rarely pursued in the rural areas, yet to crystallise into a definite pattern in the rural economy.

For three activity groups, the rse(GVA) for both rural and urban areas exceeded our predetermined limit. These activities are shown in Table 4.

Thus, preparing separate estimates for these three activities is difficult even on the basis of the 57th round data.

For the remaining five activity groups, the RSE(GVA)s for rural estimates are much higher than for the corresponding urban estimates. Although the RSE of the estimate for the two areas put together was more than 25 per cent, the fact remains that separate estimates, if required, can be prepared for the urban but not for the rural areas by this method. These are shown in Table 5.

The difference in RSE of estimates between rural and urban areas for activities shown in Table 5 is remarkably high for a few cases. For example, in computer and related activity, it is 47 per cent for the rural estimate but only 11 per cent for the urban. Two more cases where this difference was acute were activities of cable operators and real estate. This seems entirely plausible, as these activities, being urban-centric, get captured more in the urban areas, providing us with a more stable urban estimate. For hotels and accounting and bookkeeping activities, the rural RSE(GVA) was lower, but the difference in RSE(GVA) with the urban estimate was still large.

If we present the differences between RSE(WPR) and RSE(GVAPW) of the activities where RSE(GVA) is more than 25 per cent for either the rural or the urban estimates, we get a better insight into the phenomenon. Table 6 shows that the RSE(WPR) has been considerably higher than the RSE(GVAPW) in nearly every case, often by 20 percentage points or more particularly for newly emerging activities. This raises a question on blanket use of the labour input method currently in vogue for the urban-centric activities.

With some more exercises, it can probably be established that the aggregate GVA estimate from the ES can be used straightaway with correction factors obtained through comparison of workers available from this survey and the workers obtained from the population census. For example, from the total number of workers available from the population census, the number of workers in agricultural and manufacturing activities can be deducted to get the total number of workers in services. At the next stage, the number of workers from organised services (as estimated by the Central Statistical Organisation) can be deducted to arrive at the number of workers from unorganised services. This figure divided by the total number of workers available from the unorganised service sector survey by the NSSO gives one such correction factor. We did not attempt an exercise on these lines mainly due to paucity of time.

Stage 3

In the sub-section ‘Stage 2’ above, we have mentioned the activities for which estimates of RSE(GVA) were within 25 per cent for both the rural and urban areas. In the next stage, we tried to find out for how many of these activities, the state/ UT level estimates had an acceptable degree of variability.

Table 7: Activities for Which RSE (GVA) at State Level Is Tolerable for Rural and Urban Areas Separately

No NIC 98 Description No of States/UTs Where rse(GVA) ≤ 50 No of State/UT Level
Estimates Available
Both Rural and Urban Only Urban Only Rural Rural Urban
2 552 Restaurants, bars and canteens 14 5 3 26 27
3 6021 Scheduled passenger land transport 4 3 1 10 14
4 60221 Non-scheduled passenger land transport by motor vehicle 17 7 2 27 27
6 60231 Freight transport by motor vehicle 14 4 4 26 26
7 60232 Freight transport other than by motor vehicle 7 3 5 20 20
22 80-80903-80904 Other educational activities 9 4 6 24 27
28 9303+ 9309 Funeral and related activities 13 2 1 17 21
3748 Economic and Political Weekly September 15, 2007

As preparation of gross state domestic product (GSDP) for indi vidual activities is becoming more and more important, particularly with the growth in service sector activities, providing direct estimates at the state level for as many activities as possible can be one of our prime goals.

Many of the service sector activities have shown an urban inclination in the sense that they are concentrated more in the urban areas of the country. Expecting that this behaviour will also be observed at state level, we felt it necessary to find to what extent state level estimates can be generated for the urban areas separately. The “tolerable limit” of RSE(GVA) for this stage of exploration was kept at 50 per cent. It may be noted that, using the formula for calculation of RSE(GVA) as mentioned above, if both RSE(WPR) and RSE(GVAPW) equal 35 per cent, the resultant RSE(GVA) exceeds 50 per cent.

However, at this stage, we found only seven activities for which the RSE(GVA) was within our set limit for most major states for both rural and urban areas. This meant that generating state level estimates separately for both rural and urban areas can be attempted only for these seven activities, which are listed in Table 7.

In addition to the above, the RSE(GVA) of other communications was within the tolerable range for the urban areas of 12 of the 23 states. For the rural areas, the tolerable limit was met in 12 out of 24 states in human health activities, veterinary activities and 11 out of 18 states in hairdressing and other beauty treatments. It may be noted that three states, Uttar Pradesh, Bihar and Madhya Pradesh, were bifurcated politically during the period between the 55th and 57th round surveys. Thus, the above state level comparison could be readily made for all the states/UTs except these three states.

V Conclusions

The data of the ES conducted during the 57th round of the NSS broadly supports the grouping of service sector activities made by the IWG.

There was, in general, a noticeable difference in performance level between the rural and urban estimates at the all-India level.

For most of the service activity groups, generation of state level estimates of aggregate GVA directly from these data sets is discouraged by the variability of such estimates.

The RSEs of the estimates for rural areas are much more than those of their urban counterparts at the all-India level, particularly for urban-centric activities. The variability of the estimates of WPR has a bigger contribution to this than the variability of the GVAPW estimates.

It appears that the aggregate GVA at the state level for a particular activity can be satisfactorily estimated directly from the data sets only if it has a significant presence in that state.

For a few very common activities like most of the transport activities, restaurants, educational activities except coaching centres and individuals providing tuition, human health activities, veterinary activities, and funeral activity, separate estimates for the rural and urban areas of most major states can be obtained within tolerable limits of error.

In view of the large RSEs of many of the WPR estimates, using the direct estimate of aggregate GVA from the ES adjusted for the undercount of workers as available from the census may perhaps produce better estimates for many service sector activities.

VI Limitations of the Study

The sampling design of the 57th round of the NSS necessitates computation of the joint inclusion probabilities of each pair of first stage units to estimate the RSE(GVAPW) precisely. As this requires a lot of computational time as well as a host of collateral information, we have instead traversed the beaten track of computing the RSE based on approximate variances using the interpenetrating sub-sampling technique.

The method of computing estimates of aggregate GVA using a non-stochastic multiplier for undercount of workers in the ESs could not be taken up due to paucity of time.

EPW

Appendix I: Derivation of the Formula for RSE(GVA)

In the labour input method (as used by NAD) GVA = P × WPR × GVAPW, where GVA = gross value added, WPR = worker population ratio GVAPW = gross value added per worker, P = census population (assumed to be non-stochastic) Let GVA = Z, WPR = X, VAPW = Y Var(Z)=E(Z-E(Z))2 = E(XY-E(XY))2 = E(XY-E(X)E(Y))2 assuming WPR and VAPW as independent, since they are estimates based on different surveys. Var(Z) = E(XY-XE(Y)+XE(Y) - E(X)E(Y))2 =E(X(Y-E(Y))-E(X)(Y-E(Y)) +E(X)(Y-E(Y))+E(Y)(X-E(X))2 =E((X-E(X))(Y-E(Y)) +E(X)(Y-E(Y))+E(Y)(X-E(X))2 Var(Z)= Var(X)*Var(Y)+Var(Y)*E(X)2+Var(X)*E(Y)2 ---- (*) As RSE(GVA)= SE(GVA)/E(GVA), dividing (*) by E2(Z) we get the following RSE2(Z)=RSE2(X)+RSE2(Y)+RSE2(X)*RSE2(Y) => RSE(Z)= √(RSE2(X)+RSE2(Y)+RSE2(X)*RSE2(Y)) => RSE(GVA)= √(RSE2(WPR)+RSE2(GVAPW)+RSE2(WPR)*RSE2 (GVAPW))

Email: purnendukb@yahoo.com

[We are deeply thankful to Samiran Mallick, director, DPD, NSSO, for providing the RSE of WPR based on the 55th round; National Accounts Division, CSO, for the report of the internal working group; P Chaudhury, director, SDRD, NSSO, for editing and suggestions on the paper; J K Kar and A K Verma, directors, SDRD, NSSO, for their encouragement in taking up this analysis; and K V Rao, director general and chief executive officer, NSSO, for suggesting the paper for the seminar. The views expressed in this article are personal and do not represent the opinions of the organisations with which the authors are associated.]

Back Volumes

Back Volumes of Economic and Political Weekly from

1976 to 2006 are available in unbound form.

Write to:

Circulation Department,

Economic and Political Weekly

Hitkari House, 284 Shahid Bhagat Singh Road,

Mumbai 400 001.

Economic and Political Weekly September 15, 2007

Dear reader,

To continue reading, become a subscriber.

Explore our attractive subscription offers.

Click here

Comments

(-) Hide

EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

Back to Top