由於人口調查牽涉到許多Deming理論的發展
所以我們姑且一笑讀之
Counting people
Census sensitivity
Dec 19th 2007
Numbers mean power, which is why counting people is so controversial
“GO, NUMBER Israel from Beersheba even to Dan; and bring the number of them to me, that I may know it.” It was not the first census described in the Bible, nor the last, nor yet the most renowned. But for reasons that are obscure, King David's order to Joab, the commander of his army, went against God's will and both men knew it. The count was carried out all the same, and was followed by a heavy punishment: 70,000 Israelites died of the plague before the Lord relented and accepted burnt offerings as a token of David's repentance.
Taking a census thus came to be known as the sin of David, and was long regarded as best avoided. In 1634 Governor John Winthrop of the Massachusetts Bay Colony estimated the local population rather than counting it exactly, telling a correspondent: “David's example stickes somewhat with us.” And when a Census Bill was debated in Britain in 1753, Matthew Ridley, the member of Parliament (MP) for Newcastle, gave a speech saying that there was among the people “such a violent spirit of opposition to this Bill, that if it be passed into a law, there is a great reason to fear, they will in many places oppose the execution of it in riotous manner.”
But nobody needed David's dreadful example to persuade them that censuses were a bad idea. From the point of view of those being enumerated, nothing good could come of being counted. The usual reasons for wanting the numbers were war and taxes. From the sovereign's point of view, such information was crucial: the decision to go to war could be taken only once it was known how many men could be conscripted and how much money levied. So the results were highly sensitive, and an enemy country's numbers would be useful when deciding whether attacking it was prudent, and conquering it worthwhile. The results of a Swedish census in the mid-1700s appear to have been made a state secret because of such fears.
But at the same time men were becoming enamoured of numbers and taking to counting as a way of answering pressing questions of their own. Following the London plague of 1603, weekly Bills of Mortality began to be published, listing all the deaths in the city and, from 1629, their causes. According to John Graunt, whose 90-page book interpreting and drawing conclusions from these bills is generally regarded as the earliest statistical analysis, they were used by the rich to “judge of the necessity of their removall” and by tradesmen to “conjecture what doings they were like to have”. And in 1731 Benjamin Franklin published in the Pennsylvania Gazette, the newspaper he edited at the time, an account of all the ships that entered and left the big northern colonial ports, together with their destinations, so that the reader could “Make some Judgment of the different Share each Colony possesses of the several Branches of Trade”.
It was revolution that renewed the impetus for rulers to count their people. The American war of independence brought a new nation into being, and it was not only one that was made up of separate states, each keen to get full credit for its relative size, but also one whose population was on the move. In order to decide how many representatives each state should send to the new Congress, there was only one thing for it: their populations would have to be counted, and that count would have to be repeated regularly.
America's first census was carried out in 1790, and it was groundbreaking in many ways. It was the first to be mandated in any country's constitution. It was also the occasion for America's first presidential veto, exercised by George Washington on the advice of Thomas Jefferson, whom he had asked to examine the proposals for sharing out congressional seats between the states. Jefferson—a man so fond of enumeration that he once wrote to a friend that he had “ten and one-half grandchildren, and two and three-fourths great-grandchildren”, and that “these fractions will ere long become units”—criticised them for being unclear about how this “apportionment” was to be carried out. He advised Washington that a completely unambiguous method should be chosen and enshrined in legislation.
They want to know what?The fact that this thriving new nation counted its citizens without provoking divine retribution may have given courage to other Christian countries. Over the following decade Denmark, England, the Netherlands, Norway and Sweden all instigated regular censuses of their own.
People started finding the sorts of patterns in the resulting data—life expectancies, crime rates, causes of death, and the mix of religions and races—that are now part of our familiar mental furniture. In the 1800s, for instance, two French statisticians, André Michel Guerry and Adolphe Quetelet, analysed the tables of crimes against individuals and property that had recently started to be published. They were astonished by the hitherto-unsuspected regularities they found. Guerry was particularly struck by the fact that the method by which someone committed suicide could be predicted from his age. The author of an English commentary on his work described his findings thus: “The young hang themselves; arrived at a maturer age they usually blow out their brains; as they get old they recur again to the juvenile practice of suspension.”
Quetelet was equally amazed by the regularity with which the various types of crime were committed. “We can enumerate in advance”, he wrote, “how many individuals will stain their hands in the blood of their fellows, how many will be forgers, how many will be poisoners, almost as we can enumerate in advance the births and deaths that should occur.” Such regularities, he claimed, left no role for free will in human affairs. “Society prepares the crime”, he wrote in 1832, “and the guilty person is only the instrument.”
Opposition to this line of thinking came from Charles Dickens, who loathed all arguments based on numbers and averages, charging that they were used to legitimise indifference to other people's suffering. His strongest broadside against those who thought that counting people was a good way to answer social questions came in his novel “Hard Times”. When Tom, the hard-hearted Gradgrind's son, is found to be a thief, he uses his father's statistical determinism to shrug off responsibility. “So many people, out of so many, will be dishonest. I have heard you talk, a hundred times, of its being a law. How can I help laws? You have comforted others with such things, father. Comfort yourself!”
The future, though, belonged to those like another ardent social reformer, Dickens's near-contemporary, Florence Nightingale (see article). Best remembered for nursing wounded soldiers in the Crimean war, her sanitary reforms were based on meticulous records of hospital admissions, illnesses, treatments and outcomes. Rather than painting moving pen-portraits of soldiers left to rot on stinking, louse-ridden pallets in a hospital built on an open sewer—as Dickens would no doubt have done—she used death rates to campaign, successfully, for change.
Nowadays, a census is part of the standard equipment of a functioning state. In 1995 the UN called for all member nations to hold a census in the following decade. Yet counting people remains a sensitive business for two reasons, connected with the ambiguous character of government. Where government is oppressive, people want to keep out of censuses, lest information they provide is misused. Where government provides, people want to be in censuses, and to boost their numbers, in order to claim a larger share of the goodies.
Include me out
History offers good reason for worrying about the misuse of information. The Nazis used population records to round up Jews into concentration camps. As a result, Germans are still prickly about being counted. When in the 1980s their government added new questions to the census, there was a public outcry and the constitutional court struck it down on the grounds that it conflicted with a “fundamental right to informational self-determination”. Although the country is planning a census in 2011—its first since reunification—it will not be a full count, but only a sample. Respondents' surnames will be deleted as soon as possible and all data that could identify an individual will be erased once statisticians have finished with them. Questions on race and religion will not be included.
Fears that the data might be used for purposes other than the declared ones may have undermined China's most recent census in 2000. This involved 6m enumerators visiting around 350m households in just ten days. They asked some rather personal questions, such as “How much did you pay for your home?” and “How often do you wash?” But it was the standard ones about the number of residents in each household and their sex and age that provoked the most anxiety. The government wanted to find out whether the country's gender imbalance was primarily due to the abortion and infanticide of females, or whether many of the missing girls were in fact alive and being concealed. To encourage parents to register over-quota children, it reduced the penalties for anyone found to have flouted the one-child law. Some also contend that large numbers of peasants who migrated illegally to the big cities will have hidden from the enumerators, and that there may be as many as 100m uncounted Chinese on top of the 1.3 billion the census found.
It isn't just oppressive governments that misuse information. In early 2007 researchers found proof of what had long been suspected: that during the second world war the American census bureau had played a part in the internment of Japanese-Americans by passing some of their names and addresses to the secret service.
Other people's money
In autocracies, people try to keep out of censuses. In democracies, by contrast, they want to be in them, for censuses mean numbers, and numbers mean money and power. The American census, for example, determines how around $200 billion a year of federal funds is shared out, for everything from education and welfare to highways. Such rich pickings mean that censuses are well worth fighting over.
Although Jefferson ensured the American states could not squabble about the formula used for apportionment, he could not have predicted the partisan rows about how to deal with undercounting. No census counts everyone, and the uncounted are not drawn uniformly from all walks of life. The poor, the homeless, immigrants and ethnic minorities are missed more often. The 1990 census is thought to have missed one native American in eight.
The statisticians' solution is to follow each census with a quality check, surveying representative areas to create a picture of those overlooked in the full count and correcting the figures accordingly. Perhaps unsurprisingly, in light of the profile of those most likely to be missed, Democrats find the intellectual arguments in favour of such adjustment more compelling than Republicans do. Before the census in 2000, the Supreme Court ruled that sampling-adjusted figures could not be used for apportioning congressional seats. But the row over whether they should be used to share out cash raged until 2001, when the census bureau finally declared the raw figures good enough to stand unaltered.
In Britain the 2001 census was corrected using such sampling techniques, but some areas still complain that they are being undercounted. In the decade between censuses, the national statistical office updates the figures with estimates of external and internal migration. Some towns with lots of foreigners are convinced that large chunks of their population are being overlooked, losing them millions of pounds of government money. The MP for Slough, a town to the west of London full of Poles, poetically told Parliament in January that the increasing “amount of shit that goes through our local sewers” was evidence that her constituency was being sold short.
Some groups are begging to be counted because they hope to prove their importance and increase their influence with government. Church leaders in Britain were gratified when the 2001 census, the first to ask about religion, found that 70% of the population identified themselves as Christian. They reckoned this was a large enough majority to justify their religion's special place in state-run education. Whether the figures bear the weight that has been put on them is, however, questionable. Church attendance figures suggest that few of these self-identified Christians attend services more than once a year, or indeed at all. And the irreverence with which some respondents treated the question can be judged by the 390,000 people—0.7% of the population—who answered it by claiming to be Jedi.
Dashed hopes of gaining a higher profile caused despondency among gay and lesbian groups when it was decided not to add a question about sexual orientation to Britain's next census, to be held in 2011. Government statisticians were keen, but they were stymied by the difficulties of phrasing a question that would neither offend people nor leave them in the dark about what was being asked. They were mindful of the reaction to a question on self-perceived sexual identity in a recent survey by the Metropolitan Police. The most common query it provoked from respondents was: “What's heterosexuality?”
That British gays wanted to stand up and be counted says much for national tolerance. It is hard to imagine the same happening in Nigeria. There, the dispute is about who will get the oil money. The country has not had an uncontested census since gaining its independence in 1960. Civil war and poorly trained enumerators have been formidable obstacles to a reliable count. The most recent census, in 2006, omitted questions about religion and tribe following demands from leaders in the largely Muslim north. Christians in the south, who believe they have been undercounted in the past, threatened a boycott unless the questions were asked. The census found 140m people: 75m in the oil-poor north and 65m in the revenue-poor south. Politicians in the north endorsed the figures; those in the south did not.
In India, the arguments are not about money, but about a different sort of resource: the jobs and university places handed out under the government's affirmative-action programme. Many dalits, or untouchables, have tried to escape the discrimination they face in Hinduism by converting to Christianity or Islam. But this means no longer being eligible for the programme—a restriction on religious freedom, say some, and indeed court challenges are wending their way through the system. Rather than being neutral in this dispute, the census form took sides by allowing respondents to indicate their caste as dalit only if they claimed to be Hindu, Sikh or Buddhist.
The shakier the state, the fiercer the rows about censuses, for numbers affect how power is distributed. In Iraqi Kurdistan, for instance, a census due to happen in 2007 was delayed. It was intended to correct figures distorted by Saddam Hussein's “Arabisation” programme, which forced many Kurds to leave and others to declare themselves Arab. If and when an Iraqi census happens, it will not only give a truer picture of the ethnic mix in a contested region, but will also have consequences for billions of dollars in oil revenues.
Lebanon has not held a census since 1932, when it counted the number of adherents to various religions in order to share out power under a system known as confessionalism. Since that time its demography has changed and the politically favoured Christians are now believed to be in a minority. Plenty of powerful people are keen to keep that quiet, so the prospects for a new official count are dim.
Counting can be even more dangerous than being counted. In 1936 Stalin told his officials that the following year's census would find a total population of 170m—a figure that took no account of his slaughter of millions in famines and purges. But the enumerators found only 162m people, and also revealed other unwelcome facts, including that nearly half the population of this avowedly atheist country was religious. So Stalin denounced the count as a “wrecker's census” and had the census takers either imprisoned or shot. A new count in 1939 came up with a similar total, but this time officials wisely classified the results and gave Stalin his figure of 170m.
That Stalin insisted on this charade is a backhanded testimony to the way counting introduces people to themselves. “The interest and significance of the census for the community lie in this,” wrote Leo Tolstoy of the Moscow census of 1882: “that it furnishes it with a mirror into which, willy nilly, the whole community, and each one of us, gaze.” The faces that look back can surprise us still.
沒有留言:
張貼留言