G’nY. If people were to work on NSSO or Census data what are the primary difficulties of data compatibility would they face?
In India we have two major sources of migration data at the national level; the decennial Population Censuses and the not so frequent nationally representative sample surveys of the erstwhile National Sample Survey Organisation (NSSO) under the Ministry of Statistics. Briefly the Census asks (i) if the birth place of the person was outside the village/town of enumeration and (ii) if the persons had come to the present place from elsewhere, usually called the last place of residence. So you have migration status according to place of birth and by place of last residence. Census also records the reasons for changing the place of last residence, like work/employment, business, education, marriage and moved with household etc.
The Census data provides the number of migrants and their flows to different locations enabling one to understand migration patterns. However, it has limitations in relating migration to household or individual socio-economic characteristics. If researchers were to relate migration to socio-economic characteristics of migrants, they would need more detailed data on these characteristics before and after migration. These are possible in detailed household surveys and not in population censuses.
Sample Surveys on migration and related aspects conducted by the NSSO have been a major source of data accessed by researchers mainly due to its flexibility for analysis and easy access to the micro data collected from households. Unlike the census, these surveys do not have a fixed periodicity. Migration surveys with comparable definitions and coverage were done in 1983, 1987-1988, 1993, 1999-2000 and in 2007-2008. The last survey, for example, collected data on household migration, out-migration, in-migration and seasonal short-term migration. The details of location of last usual residence, nature of migration and reason for migration, duration, economic activity before and after migration, remittances sent and how it was used were asked from the surveyed households. In addition the short duration movements of the people for employment or in search of employment for a period of 30 days to 6 months, number of spells of duration of 15 days or more, destination for longest spell and industry of work were also collected.
The NSSO data is much more comprehensive and help understand the migration situation better within the limitations enforced by the nature of a sample survey such as sampling errors especially when one is looking at a disaggregated analysis.
G’nY. Do you think that in releasing data years later (for instance the 2011 data on migration of the Census was released recently) makes the data lose much of its relevance?
In the Indian geo-political context migration is a sensitive topic, especially the Census numbers. Therefore some amount of detailed examination of the data before publication is always expected. However, the final tables from Census 2011 were made available only recently. Some provisional tables were released earlier. Questions were then raised by Prof Amitabh Kundu and myself on the inconsistencies in the provisional data and these were withdrawn and replaced later. Multiple versions of data is a challenge
There are views that the internal migration scenario has changed substantially in recent times. This was a topic in one of the Economic Surveys of the Government. Indirect estimation using advanced data analytics from alternate data sources like rail ticket sale, were cited as evidence for this. These estimation procedures depends on a lot of assumptions. A combination of decennial census and periodic surveys are essential to track the migration scene more effectively.
That the Census 2011 data on migration came very late and the failure of NSSO to conduct another survey after 2007-2008 are major setbacks for migration studies in India.
G’nY.Can you advise whether recording only one reason for migration in the Census is insufficient? For instance, women migrate on account of exogamous marriage, but they can start work immediately. However, this would not be counted as a reason for migration. Do you think any rethinking/modification is needed in
Some amount of simplification always take place while collecting data on reasons for migration. The head of household is the proxy respondent for other members in the household and the survey instrument facilitates this simplification. Migration affects people differently in the same household. Marriage migration may shift women away from workforce just as it may also improve their chances of entering workforce in the new place. The instructions given in NSSO are usually to assign the primary reason for the movement to the new place. A better solution to understand this is to record the activity status of respondents before and after migration as is done in the NSSO.
G’nY. Can NSSO capture a more nuanced migration data in the future as the Census does not adequately capture seasonal and/or short-term circular migration? The period of short term migration is gradually changing from the lean period in the agricultural season to irregular episodes of circular migration. Should such complex migration episodes be meaningfully captured?
One cannot expect the census to capture very detailed data in view of the scale of operations. But even the NSSO has limitations due to its sample size in capturing all possible instances. The complexity of migration may not be adequately represented in the selected sample due to its small size.
Large number of micro level studies have documented the prevalence of short term migration in India. The Census enumerates people on change of place of residence and no data on their mobility pattern are collected. While the NSSO does attempt to collect information on short term migrants, there are constraints in covering the entire range of short term migrants. Large number of migrant workers may go out of their villages and towns for work for periods more than six months without planning to change their place of residence permanently. Theoretically, they should be enumerated in their new place of residence since the stay exceeds six months. This would happen only if such a place of stay is fixed for the entire period of migration and the structure that house them are listed in the survey operation. For migrant workers, especially in construction sector or having mobile work places, the place of stay may not be fixed although the stay away from home is more than six months. Since they are staying away from their households exceeding six months, they cannot claim to be members of the household or their place of origin as the usual place of residence. The population census does cover the houseless population, but the surveys usually do not.
G’nY. A segment of the migrant labour live at worksites – from construction sites to makeshift labour camps. Is it relevant to capture such mobile but invisible workers (and their families)?
A problem in household surveys and also to an extent in population census, is locating migrants who do not set up independent households. It is known that many migrants live in temporary structures around construction sites or in labour camps that are not usually registered as households in records. They are not easily accessible to enumerators. To an extent NSSO overcomes this problem by asking the surveyed households if any member has out-migrated for work. This is possible only when the entire household has not moved to such temporary living facilities. Temporary migration such as for casual or temporary work has a time dimension that may not be easy to specify
Identifying migrant labour or itinerant artisans living in temporary facilities or taking shelters in purely temporary structures along highways etc. is extremely important as they constitute the most deprived sections of migrants. Most often they fail to avail welfare entitlements due to migration and lack of portability of these schemes across states. As noted earlier, there are practical issues in national level surveys to locate and list migrant workers. The basic difficulty arises due to the need to survey ‘households’ which presume a structure of certain permanency. With some effort they can be identified from the household of origin as out-migrants, provided the entire household has not shifted.
National surveys have the object of providing valid estimates at national and regional levels. Micro studies of different industries and locations are useful to provide corroborative evidences on labour migration that national level surveys cannot provide.
G’nY. Reasons of migration should have multiple choices. For instance–natural disasters, including floods and drought, which may be added as also social and economic tensions – what are your views?
No doubt these are important reasons but many a time these are localised making it difficult to be sufficiently represented in national level studies. Many a time we see that certain categories of migrants like short term migrants are rather small but quite significant for some states. The census would be a better place to record them. As per the 2011 Census, 53.3 million migrants had moved due to reasons not specified in the census questionnaire (viz. employment, business, education, marriage, birth and accompanying household movement). Between 2001 and 2011 about 20 million migrants belonged to this unspecified reason category. It may be difficult to expect a sample survey to produce reliable numbers when these reasons are broken down further. We have to depend on area specific studies to understand and highlight
G’nY. Significantly, NSSO’s consumer expenditure data seems to be facing some difficulties in the current context. What seems to be the primary concerns?
Consumer expenditure data have always been at the centre of critical debate ever since it came to be used to derive poverty estimates. Many of the issues like the growing divergences with the private consumption expenditure derived from national accounts are well known. Besides technical issues, practical issues like its ability to capture the consumption of those in the high income bracket have been in radar of critics. There has been an increasing feeling among survey practitioners that collecting consumption expenditure data on all items of goods and services consumed by the household are getting very difficult due to respondent resistance and the recall issues. Thus nobody would be surprised if the data showed up some inconsistencies with other sources. If we are looking at the distribution of households by expenditure classes then this problem is a little less serious compared to aggregate level expenditures.
While these issues are nothing new, the decision to junk the survey altogether on the basis of issues regarding data quality without specifying the details create doubts as it follows a series of similar steps in respect of other official data. Alternatively it is likely that the survey done immediately after the demonetisation exercise shows a drastic decline in consumption. In any case, the most desirable action would have been to release the report and data for researchers to examine it scientifically, paving the way for methodological improvements
Unfortunately data related issues have acquired a political dimension as we saw in the reluctance of the Government to release the labour force survey results before the elections. We have also seen the caveats under which the estimates of households having access to toilets from another survey has been released recently. The survey estimates were significantly lower than the claims of the government. Independent surveys are necessary to validate administrative data and are not meant to defend the official figures.
My primary concern is the manner in which national statistics is losing its independence and public trust leading to loss of credibility of numbers put out by statistical agencies. Statistical agencies should not be immune to criticisms. But this should not be an excuse for them to supress statistical reports and data. Transparency can only strengthen them and the autonomy of these agencies need be safeguarded through legislative measures.
G’nY. Has there been any methodology change suggested or undertaken in the consumer expenditure data set that might affect the comparability of consumer expenditure over the previous years?
There were many methodological studies to improve the collection of consumer expenditure data. The most discussed were the studies on reference period (recall period). Yet another study was done to collect data for different groups of items from different households to reduce survey time and respondent burden. The comparability of data was the main factor for not making substantial methodological changes as that would have raised debates on the resulting poverty numbers either way. This actually happened in 1999-2000.
Household consumer expenditure is usually accepted as a proxy for household income, which is admittedly more difficult to collect. But given the present context it may be worthwhile to attempt to collect income data from households in place of consumption.
G’nY. Is there any relationship between migration and citizenship?
Migration reporting including period since migration is also likely to be affected if the respondents have concerns of their ultimate use by government agencies for possible administrative and political purposes. Exercises to establish citizenship or prepare citizen registers for the population will have a bearing on the reportage of migration related issues.