The provisional results of the 2017 census have surprised most political commentators, journalists, members of civil society, policy makers as well as demographers, who are questioning the validity of the entire exercise.
This census recorded Pakistan’s population at 207.8 million, which is 10m over the officially estimated population, which grew at an astonishing 2.4% from 1998 to 2017. This contradicts the government’s claim of 2%. (It is reported as 1.89% in the 2015-16 Pakistan Economic Survey.) Another big surprise is that Sindh’s share of people has remained at the same level as it was in 1998 and is only slightly higher than what it was in 1981.
Punjab’s annual growth has been traditionally lower than the national average as well as that of other provinces. Its 2.1% from 1998 to 2017 is attributed to a decline in fertility. Growth rates of 2.9% in Khyber Pakhtunkhwa and 3.4% in Balochistan are substantially higher than the national average, perhaps due to the inclusion of Afghan refugees who have mingled with the locals.
Sindh
Sindh’s population growth during the past 34 years has been substantially reduced. Till 1981, Sindh experienced much higher annual inter-censal growth rates. However, during the 1981-98 and 1998-2017 periods its growth rates seem to have been substantially reduced, from 3.9% during 1972-81 to 2.4% during the 1998-2017 period. The earlier higher growth rate in Sindh during the 1951-81 period was mainly due to migration from other provinces. This led to an increase in Sindh’s share in the country’s population from 18% in 1951 to about 23% in 1981—a 4.5 percentage point increase in 30 years. However, due to slower annual growth rates from 1981 to 2017, only 0.4 percentage points increased in Sindh’s share in the country’s population in 37 years. Thus, Sindh is reported to have maintained about the same share in the national population since 1981.
A combination of factors could have led to Sindh having a slower population growth rate during the past three decades as compared to the first three decades after Independence. These include: its fertility rate has declined more rapidly as compared to that of other provinces; migration from other provinces to Sindh has subsided; Sindh’s population was undercounted in 1998 as well as in the 2017 census.
Fertility
Pakistan is among the few Asian countries where the pace of fertility decline has been the slowest. But an examination of trends in total fertility rate (average number of children born per woman) shows that the decline in all provinces is somewhat similar. However, more recently in Punjab, on average, each woman is bearing half fewer children than women in Sindh. Recent data on KP and Balochistan are not available.
Inter-provincial migration
Since Independence, constant migration from rural to urban areas and from other provinces to Sindh, particularly to Karachi, has been prominent, resulting in an increase in its share in the country’s population.
Migration continued in the 1980s and 90s. During the past ten years prior to the 1998 census, of about 3.8m people who migrated from one district to another, about half migrated within their own province and the other half to another province. However, within each province, the pattern was somewhat different. For example, 71% of migrants who settled in Sindh had originated from districts in other provinces. A majority of migrants in Punjab and KP moved within the province. The question on inter-district and inter-province migration has been dropped in the 1998 census, therefore the pattern cannot be determined. However, it is clear that substantial migration to Sindh, particularly to Karachi, has taken place from other provinces, over the past couple of decades.
Under- or over-counting
Unlike in previous censuses, the 2017 one was conducted on a de jure basis which takes into account permanent residence. Migrants to Sindh from other provinces who have not changed their address on their NADRA cards, have been perhaps counted at their original place of residence.
Besides this, the census results are never perfect and thus under- or over-counting of people is possible. The best way to check the validity of a census result is through the universally accepted procedure of post-census enumeration survey (PES) which is conducted in randomly selected areas, and its results are cross-checked with the census results of the same areas. The PES is routinely conducted in many countries, with varying results.
For example, the PES conducted in the US after the 1950 census disclosed that about 5% of people were missed, which was brought down to 1.8% in 1990. More recently, results of the PES conducted following the 2001 Australia census showed that 2% people were missed. In the 2011 census of India, 2.3% were missed and in the 2011 census of Bangladesh, 4.2% were missed.
In Pakistan, a PES was conducted following the 1961 census, which indicated an under-count of 6.7% and consequently the population was adjusted upwards. A PES was conducted after the 1981 census, but its results were not released. Such an exercise was not carried out after the 1998 census, nor is any planned, after the 2017 one.
Another way of cross-verification of census results is accomplished through demographic modeling, which takes into account trends in demographic indicators (yearly estimates of birth, death, and migration rates when available) between two censuses. While serving as a member of the steering committee set up by Pakistan and the UN to evaluate the 1998 census results, I used data from the annual demographic surveys conducted during 1982 and 1997.
I took into account people who had migrated from Pakistan to the Gulf countries, Europe and the US during the 1981-98 period. I estimated that in the 1998 census, about 6m fewer people were counted, and most were missing from Sindh. Unfortunately, the annual Pakistan Demographic Survey was discontinued after 2007, thus proper estimates of demographic indicators from 2008 to 2017 are not available. This makes it difficult to estimate the population of the country and its provinces.
Unlike in the censuses from 1951 to 1998, when questions related to inter-district and inter-provincial migration were routinely asked, these questions were dropped in the 2017 census. Therefore, recent trends and patterns of migration in the country and between provinces cannot be estimated.
Since Independence, Sindh has been receiving a highly disproportionate number of migrants from other provinces, resulting in an increase in Sindh’s share in the country’s population from about 18% in 1951 to about 23% in 1981. However, both the 1998 as well as 2017 census indicate that Sindh’s share has remained at 23%.
Data indicates that the average number of persons living in each household in Sindh has been declining from 6.8 in 1981, to 6.1 in 1998 and to 5.6 in 2017. A similar trend is also noted in Balochistan where average household size declined slightly from 7.2 in 1981 to 7 in 2017. On the other hand in Punjab, it increased from 6.3 in 1981, to 7 in 1998 and was recorded as 6.4 in 2017, while increased substantially from 6.8 in 1981 to 7.9 in 2017.
While it is possible that due to the slow decline in fertility and perhaps a slow increase in housing construction, household crowding has been increasing. But one is puzzled that a similar trend is not noted in Sindh, particularly when it has been receiving migrants. Could the lower average household size reported for Sindh in the 1998 and 2017 censuses be a reflection of under-reporting of Sindh’s population?
When the average household size from the 1998 and 2017 censuses is compared with that in the yearly household sample surveys (by the PBS during the past 30yrs), Sindh’s was higher than that of Punjab’s in most years. In the most recent household sample surveys (2011-16), due to lower fertility levels in Punjab, the average household size in the province declined to 6.1. It is 6.5 in Sindh. It appears that an average of one person was missed in each household in Sindh and 0.5 persons were missed in Balochistan, for some unexplained reasons. On the other hand, on an average of 0.3 additional persons in Punjab and 0.7 in KP were reported in each household.
Estimates suggest that in 2017 it is likely that about 5.7m and 2.8m people were over-counted in Punjab and KP, respectively. On the other hand, 7.9m and 0.6m people were under-counted or missed in Sindh and Balochistan. Assuming that there was no under-count, the revised estimated share of Sindh and Balochistan would increase to 26.9% and 6.2%, respectively and that of Punjab and KP would decline to 50.2% and 13.3%.
If we take into account average household size as reported in household surveys conducted during 2011-16, population of each province will have to be adjusted. The share of each province in the total population will change. Sindh and Balochistan’s share will increase and that of Punjab and KP will decline. Thus, if the 2017 census would have been conducted correctly, the share of Sindh in the NFC award and in representation will increase substantially.
In the absence of international standards of cross-validation, if the 2017 census results are accepted at face value, it could lead to discontent in Sindh and Balochistan on the one hand, and disastrous policy planning on the other. it would not be in the national interest to accept the 2017 census results at face value until the validity of the census is checked through a post-enumeration survey in randomly selected areas. This should be done by a third party (e.g. jointly by the PBS with input from demographers and university statistics departments).
If the PES were followed in 2017, it would have been possible to determine any under- or over-counting. In meetings with senior PBS officials, demographic experts (including myself), had proposed a PES following the 2017 census. This was opposed over fears the census results may not match PES results.
It is not too late to conduct a PES in about 2% of randomly selected census enumeration blocks all over the country. This can only be done if the PBS cooperates. It has been reported in the press, that when the Sindh government requested the PBS to provide documents to check the validity of results, the request was denied. This is not only against the principle of transparency, but in case there is a genuine discrepancy in the numbers, it is will also jeopardize policy outcomes.
The author was a professor of Demography and Public Policy and at present is VC of Malir University of Science & Technology in Karachi. He has co-authored three books on demographic issues in Pakistan and Muslim countries. Earlier, on behalf of the UN, he served as an Advisor to Pakistan Census Organization. During 2012-14, he was a member of the Governing Council of the Pakistan Bureau of Statistics and in that capacity chaired a committee on census. mehtabkarim@gmail.com
This census recorded Pakistan’s population at 207.8 million, which is 10m over the officially estimated population, which grew at an astonishing 2.4% from 1998 to 2017. This contradicts the government’s claim of 2%. (It is reported as 1.89% in the 2015-16 Pakistan Economic Survey.) Another big surprise is that Sindh’s share of people has remained at the same level as it was in 1998 and is only slightly higher than what it was in 1981.
Punjab’s annual growth has been traditionally lower than the national average as well as that of other provinces. Its 2.1% from 1998 to 2017 is attributed to a decline in fertility. Growth rates of 2.9% in Khyber Pakhtunkhwa and 3.4% in Balochistan are substantially higher than the national average, perhaps due to the inclusion of Afghan refugees who have mingled with the locals.
71% of migrants who settled in Sindh had originated from districts located in other provinces. On the other hand, a majority of migrants in Punjab and KP moved within the province
Sindh
Sindh’s population growth during the past 34 years has been substantially reduced. Till 1981, Sindh experienced much higher annual inter-censal growth rates. However, during the 1981-98 and 1998-2017 periods its growth rates seem to have been substantially reduced, from 3.9% during 1972-81 to 2.4% during the 1998-2017 period. The earlier higher growth rate in Sindh during the 1951-81 period was mainly due to migration from other provinces. This led to an increase in Sindh’s share in the country’s population from 18% in 1951 to about 23% in 1981—a 4.5 percentage point increase in 30 years. However, due to slower annual growth rates from 1981 to 2017, only 0.4 percentage points increased in Sindh’s share in the country’s population in 37 years. Thus, Sindh is reported to have maintained about the same share in the national population since 1981.
A combination of factors could have led to Sindh having a slower population growth rate during the past three decades as compared to the first three decades after Independence. These include: its fertility rate has declined more rapidly as compared to that of other provinces; migration from other provinces to Sindh has subsided; Sindh’s population was undercounted in 1998 as well as in the 2017 census.
Fertility
Pakistan is among the few Asian countries where the pace of fertility decline has been the slowest. But an examination of trends in total fertility rate (average number of children born per woman) shows that the decline in all provinces is somewhat similar. However, more recently in Punjab, on average, each woman is bearing half fewer children than women in Sindh. Recent data on KP and Balochistan are not available.
Inter-provincial migration
Since Independence, constant migration from rural to urban areas and from other provinces to Sindh, particularly to Karachi, has been prominent, resulting in an increase in its share in the country’s population.
Migration continued in the 1980s and 90s. During the past ten years prior to the 1998 census, of about 3.8m people who migrated from one district to another, about half migrated within their own province and the other half to another province. However, within each province, the pattern was somewhat different. For example, 71% of migrants who settled in Sindh had originated from districts in other provinces. A majority of migrants in Punjab and KP moved within the province. The question on inter-district and inter-province migration has been dropped in the 1998 census, therefore the pattern cannot be determined. However, it is clear that substantial migration to Sindh, particularly to Karachi, has taken place from other provinces, over the past couple of decades.
Under- or over-counting
Unlike in previous censuses, the 2017 one was conducted on a de jure basis which takes into account permanent residence. Migrants to Sindh from other provinces who have not changed their address on their NADRA cards, have been perhaps counted at their original place of residence.
Besides this, the census results are never perfect and thus under- or over-counting of people is possible. The best way to check the validity of a census result is through the universally accepted procedure of post-census enumeration survey (PES) which is conducted in randomly selected areas, and its results are cross-checked with the census results of the same areas. The PES is routinely conducted in many countries, with varying results.
For example, the PES conducted in the US after the 1950 census disclosed that about 5% of people were missed, which was brought down to 1.8% in 1990. More recently, results of the PES conducted following the 2001 Australia census showed that 2% people were missed. In the 2011 census of India, 2.3% were missed and in the 2011 census of Bangladesh, 4.2% were missed.
In Pakistan, a PES was conducted following the 1961 census, which indicated an under-count of 6.7% and consequently the population was adjusted upwards. A PES was conducted after the 1981 census, but its results were not released. Such an exercise was not carried out after the 1998 census, nor is any planned, after the 2017 one.
Demographic modeling
Another way of cross-verification of census results is accomplished through demographic modeling, which takes into account trends in demographic indicators (yearly estimates of birth, death, and migration rates when available) between two censuses. While serving as a member of the steering committee set up by Pakistan and the UN to evaluate the 1998 census results, I used data from the annual demographic surveys conducted during 1982 and 1997.
I took into account people who had migrated from Pakistan to the Gulf countries, Europe and the US during the 1981-98 period. I estimated that in the 1998 census, about 6m fewer people were counted, and most were missing from Sindh. Unfortunately, the annual Pakistan Demographic Survey was discontinued after 2007, thus proper estimates of demographic indicators from 2008 to 2017 are not available. This makes it difficult to estimate the population of the country and its provinces.
Unlike in the censuses from 1951 to 1998, when questions related to inter-district and inter-provincial migration were routinely asked, these questions were dropped in the 2017 census. Therefore, recent trends and patterns of migration in the country and between provinces cannot be estimated.
Since Independence, Sindh has been receiving a highly disproportionate number of migrants from other provinces, resulting in an increase in Sindh’s share in the country’s population from about 18% in 1951 to about 23% in 1981. However, both the 1998 as well as 2017 census indicate that Sindh’s share has remained at 23%.
Data indicates that the average number of persons living in each household in Sindh has been declining from 6.8 in 1981, to 6.1 in 1998 and to 5.6 in 2017. A similar trend is also noted in Balochistan where average household size declined slightly from 7.2 in 1981 to 7 in 2017. On the other hand in Punjab, it increased from 6.3 in 1981, to 7 in 1998 and was recorded as 6.4 in 2017, while increased substantially from 6.8 in 1981 to 7.9 in 2017.
While it is possible that due to the slow decline in fertility and perhaps a slow increase in housing construction, household crowding has been increasing. But one is puzzled that a similar trend is not noted in Sindh, particularly when it has been receiving migrants. Could the lower average household size reported for Sindh in the 1998 and 2017 censuses be a reflection of under-reporting of Sindh’s population?
When the average household size from the 1998 and 2017 censuses is compared with that in the yearly household sample surveys (by the PBS during the past 30yrs), Sindh’s was higher than that of Punjab’s in most years. In the most recent household sample surveys (2011-16), due to lower fertility levels in Punjab, the average household size in the province declined to 6.1. It is 6.5 in Sindh. It appears that an average of one person was missed in each household in Sindh and 0.5 persons were missed in Balochistan, for some unexplained reasons. On the other hand, on an average of 0.3 additional persons in Punjab and 0.7 in KP were reported in each household.
Estimates suggest that in 2017 it is likely that about 5.7m and 2.8m people were over-counted in Punjab and KP, respectively. On the other hand, 7.9m and 0.6m people were under-counted or missed in Sindh and Balochistan. Assuming that there was no under-count, the revised estimated share of Sindh and Balochistan would increase to 26.9% and 6.2%, respectively and that of Punjab and KP would decline to 50.2% and 13.3%.
If we take into account average household size as reported in household surveys conducted during 2011-16, population of each province will have to be adjusted. The share of each province in the total population will change. Sindh and Balochistan’s share will increase and that of Punjab and KP will decline. Thus, if the 2017 census would have been conducted correctly, the share of Sindh in the NFC award and in representation will increase substantially.
Way forward
In the absence of international standards of cross-validation, if the 2017 census results are accepted at face value, it could lead to discontent in Sindh and Balochistan on the one hand, and disastrous policy planning on the other. it would not be in the national interest to accept the 2017 census results at face value until the validity of the census is checked through a post-enumeration survey in randomly selected areas. This should be done by a third party (e.g. jointly by the PBS with input from demographers and university statistics departments).
If the PES were followed in 2017, it would have been possible to determine any under- or over-counting. In meetings with senior PBS officials, demographic experts (including myself), had proposed a PES following the 2017 census. This was opposed over fears the census results may not match PES results.
It is not too late to conduct a PES in about 2% of randomly selected census enumeration blocks all over the country. This can only be done if the PBS cooperates. It has been reported in the press, that when the Sindh government requested the PBS to provide documents to check the validity of results, the request was denied. This is not only against the principle of transparency, but in case there is a genuine discrepancy in the numbers, it is will also jeopardize policy outcomes.
The author was a professor of Demography and Public Policy and at present is VC of Malir University of Science & Technology in Karachi. He has co-authored three books on demographic issues in Pakistan and Muslim countries. Earlier, on behalf of the UN, he served as an Advisor to Pakistan Census Organization. During 2012-14, he was a member of the Governing Council of the Pakistan Bureau of Statistics and in that capacity chaired a committee on census. mehtabkarim@gmail.com