Professional Documents
Culture Documents
Instant Download Ebook PDF Elementary Survey Sampling 7th Edition PDF Scribd
Instant Download Ebook PDF Elementary Survey Sampling 7th Edition PDF Scribd
Preface ix
Chapter 1 Introduction 1
10.5 Estimating Population Density and Size from Quadrat Samples 319
10.6 Estimating Population Density and Size from Stocked Quadrats 323
10.7 Adaptive Sampling 325
10.8 Summary 327
Exercises 328
Sampling from Real Populations 331
Appendix A
Bibliography, Tables, and Derivations 380
Appendix B
Macros for SAS 392
Appendix C
Data Sets 414
Index 432
Preface
Purpose
Elementary Survey Sampling, Seventh Edition, is an introductory textbook on the de-
sign and analysis of sample surveys intended for students of business, the social sci-
ences, the health sciences, environmental studies, natural resource management, or
other areas that make extensive use of sampling. As such, we mean “survey” in the
broadest sense; applications include, for instance, measured data on selected sample
units, as well as responses from people to questionnaires. It is written to appeal to
students interested in the application of survey sampling, not the statistical theory
that underlies survey design and analysis; the only prerequisite is an elementary
course in statistics.
Approach
This book emphasizes the practical aspects of survey problems. It begins with brief
chapters on the role of sample surveys in the modern world. Thereafter, each chapter
introduces a sample survey design or estimation procedure followed by describing
the pertinent practical problem. These chapters are structured as follows:
• The methodology proposed for solving the problem is described, followed by de-
tails of the estimation procedure, including a compact presentation of the formulas
needed to complete the analysis. The formulas in each estimation procedure
have been programmed into an interactive Excel worksheet.
• In each instance, a practical example is worked out in complete detail. Each
worked example in the text is embedded in the relevant Excel worksheet, so the
student can study the material by both reading and seeing the formulas in action.
• Many exercises are provided at the end of each chapter to give students ample
opportunity to practice the techniques and stretch their grasp of ideas. All data
sets (excepting the very smallest) are available electronically on the accompa-
nying CD.
We have endeavored to avoid making this merely a cookbook. To justify many of
the formulas and to support the choice of particular sampling designs, we have
supplied explanations that appeal to students’ intuition. Simulations demonstrating
the properties of estimators are provided in a few places. Examples and exercises
have been selected from many fields of application. Answers given for some se-
lected exercises may be subject to small rounding errors because of the complexity
of some formulas.
ix
x Preface
Organization
After a brief introductory chapter, the book provides a description of terms pertinent to
survey sampling, along with a discussion of the design of questionnaires and methods
of data collection (Chapter 2). Then, some of the ideas of introductory statistics are
related to basic issues of sample survey design and analysis (Chapter 3). Chapters 4,
5, 7, and 8 present the four most common sample survey designs—namely, simple
random sampling, stratified random sampling, systematic sampling, and cluster sam-
pling, respectively. Chapter 6 brings in the notion of using an auxiliary variable
through ratio, regression, and difference estimation. The remaining chapters deal
with two-stage cluster sampling and other specialized problems that occur in survey
sampling.
Practical aspects of conducting survey samples are emphasized, with sections on
sources of errors in surveys, methods of data collection, designing questionnaires,
and guidelines for planning surveys. Sampling with probabilities proportional to size
is introduced in Chapter 3 and applied to cluster sampling in Chapters 8 and 9. Chap-
ter 10 examines methods for estimating the population size (the number of people,
animals, or plants) in a given circumstance. Brief introductions to a selection of sup-
plemental topics are presented in Chapter 11. (All chapters contain many examples
of how the sampling concepts are used in practice.)
Chapter 12 contains a review of the main sampling designs and a set of exercises
that may require some thought in the selection of appropriate analyses. Appendix A
includes the mathematical derivations of many of the main results in the book. The
understanding of many of these derivations requires a working knowledge of ele-
mentary probability theory. SAS macros to implement methods from the book are
described in Appendix B; certain extensive data sets are detailed in Appendix C.
from rote arithmetic, he or she can focus on the bigger picture of method selection,
properties of methods, and so on.
Beginning with Chapter 4, each chapter has an electronic section (a Word file),
which has links to the tools for all the formulas presented in that chapter. In addition,
all of the numerical examples for each chapter are built in to the relevant tools; the
reader can insert the data from the examples into the tool by a simple click on a but-
ton. In the text itself, an icon is displayed by relevant presentations of formulas and
by numerical examples to remind the reader of the availability of the tools.
Acknowledgments
We are grateful to students of Gerow, whose enthusiastic use of and feedback on the
Excel tools inspired us. Particular thanks are due to Nancy Carter and her students
at the California State University at Chico, who used an early version of the toolkit
and contributed immensely to its advancement by their critical and thoughtful use of
the tools.
Survey sampling plays an increasingly important role in today’s information
society. We hope this book helps students design better surveys and understand the
subtleties of survey results presented to them. In short, our goal is to improve the flow
of knowledge from data.
Richard Scheaffer
Lyman Ott
Ken Gerow
This page intentionally left blank
1
Introduction
“Twenty-two percent of Americans doubt that the Holocaust ever occurred,” blasted
a news statement in 1993. How could this be? How could so many doubt the occur-
rence of one of the most significant events of the twentieth century—and, perhaps, of
all time? The answer lies in the poll, or survey, itself and in the difficulty of commu-
nicating with language, which always produces some degree of imprecision. The
question actually asked by the Roper Organization was as follows:
Does it seem possible or does it seem impossible to you that the Nazi extermination
of the Jews never happened?
Among the respondents, 22% said “it seemed possible” and another 12% said they
did not know. Only 65% said it was “impossible it never happened.” The double neg-
ative in the question, it seems, confused the respondents to the point that they were
not sure how to phrase their answers. Picking up on this point, the Gallup Organiza-
tion conducted a follow-up poll that asked the question in more detailed form:
The term Holocaust usually refers to the killing of millions of Jews in Nazi death
camps during World War II. In your opinion, did the Holocaust: definitely happen,
probably happen, probably not happen, or definitely not happen?
Among the respondents to this poll, 83% said the Holocaust definitely happened and
another 13% said it probably happened. Only 1% said it definitely did not happen.
Quite a difference! In a separate poll, Gallup asked the Roper question again and
found that 37% of the respondents said it seemed possible that it never happened.
Polls and surveys increasingly guide political, research, and business decisions,
but they can be highly volatile and are little understood. An excellent study of current
trends in collecting and using data is the book Tainted Truth: The Manipulation of
Fact in America by Cynthia Crossen (1994, p. 14), in which she points out that
We are skeptical about statistical and factual information, but not as skeptical as we
think. . . . we respect numbers and we cannot help believing them. Yet, more and
more of the information we use to buy, elect, advise, acquit and heal has been
created not to expand our knowledge but to sell a product or advance a cause.
That’s what surveys do, they basically manufacture news.
1
2 Chapter 1 Introduction
We depend on data to make intelligent decisions, yet the data we see are often
tainted. An old saying on the use and misuse of computers was “garbage in–garbage
out,” but this has become “garbage in–gospel out” as more and more people get into
the numbers game. So what can we do? Part of the answer lies in education. Con-
sumers and producers of data with the serious, unbiased objective to get at the “truth”
must be educated in how surveys work, how good surveys can be designed, and how
survey data can be properly analyzed. That education is the purpose of this book.
Introductory courses stress that modem statistics is a theory of information with
inference as its objective. The target of our curiosity is a set of measurements, a pop-
ulation, that exists in fact or may be generated by repeated experimentation. The
medium of inference is the sample, which is a subset of measurements selected from
the population. We wish to make an inference about the population on the basis of
characteristics of the sample—or, equivalently, the information contained in the sam-
ple. For example, suppose a chain of department stores maintains customer charge
accounts. The amount of money owed by the company will vary from day to day as
new charges are made and some accounts are paid. Indeed, the set of amounts due to
the company on a given day represents a population of measurements of considerable
interest to the management. The population characteristic of interest is the total of all
measurements in the population or, equivalently, the daily total credit load.
Keeping track of the daily total credit associated with charge accounts may seem
to be a simple task for an electronic computer. However, the data must be updated
daily, and updating takes time. A simpler method for determining the total credit load
associated with the charge accounts is to randomly sample the population of accounts
on a given day, estimate the average amount owed per account, and multiply by the
number of accounts. In other words, we employ a statistical estimator to make an in-
ference about the total population. Elementary statistics tells us that this estimate can
be made as accurate as we wish simply by increasing the sample size. The resulting
estimate either is accompanied by a bound on the error of estimation or is expressed
as a confidence interval. Thus, information in the sample is used to make an inference
about the population.
Information from sample surveys affects almost every facet of our daily lives.
Such information determines government policies on, for example, the control of the
economy and the promotion of social programs. Opinion polls are the basis of much
of the news reported by the various news media. Ratings of television shows deter-
mine which shows are to be available for viewing in the future.
We usually think of the U.S. Census Bureau (http://www.census.gov/) as attempting
to contact every household in the country in order to count the population. But the de-
cennial census collects far more than simple counts. In the 2000 census, the short-
form questionnaire that went to all households had questions covering only tenure
(whether a housing unit is owned or rented), name, sex, age, relationship to house-
holder, Hispanic origin, and race. The long-form questionnaire, which went to a sam-
ple of one in six households, had the short-form questions plus additional questions
(40 or so) on such topics as the social characteristics of the population, marital
status, place of birth, citizenship, educational attainment, ancestry, language spoken
at home, veteran status, occupation, income, and housing conditions. The resulting
information is used by the federal government in determining allocations of funds to
states and cities. It is used by business to forecast sales, to manage personnel, and to
Introduction 3
establish future site locations. It is used by urban and regional planners to plan land
use, transportation networks, and energy consumption. It is used by social scientists
to study economic conditions, racial balance, and other aspects of the quality of life.
The U.S. Bureau of Labor Statistics (BLS) (http://www.bls.gov/) routinely con-
ducts over 20 surveys. Some of the best known and most widely used are the surveys
that establish the consumer price index (CPI). The CPI is a measure of price change
for a fixed-market basket of goods and services over time. It is used as a measure of
inflation and serves as an economic indicator for government policies. Businesses
have wage rates and pension plans tied to the CPI. Federal health and welfare pro-
grams, as well as many state and local programs, tie their bases of eligibility to the
CPI. Escalator clauses in rents and mortgages are based on the CPI. So we can see
that this one index, determined on the basis of sample surveys, plays a fundamental
role in our society.
The CPI is based on a number of surveys. Consumer Expenditure Surveys pro-
vide information on the buying habits of American consumers, including data on
their expenditures, income, and consumer unit (families and single consumers) char-
acteristics. These are matched with prices for the goods and services found by way of
data collected in 87 urban areas throughout the country and from approximately
23,000 retail and service establishments. Data on rents are collected from approxi-
mately 50,000 landlords or tenants.
One of the most noticeable of the BLS data collection efforts is the Current Pop-
ulation Survey (CPS), a monthly survey of households that provides a comprehen-
sive body of data on the labor force, employment, unemployment, and people not in
the labor force. Each month the CPS collects information on the labor force status of
the civilian noninstitutional population 15 years of age and older, although labor
force estimates are reported only for those 16 and older, using a probability sample
of approximately 60,000 households. Respondents are assured that all information
obtained is completely confidential and is used only for the purpose of statistical
analysis.
The BLS conducts other surveys. The National Compensation Survey provides
comprehensive measures of occupational earnings, compensation cost trends, and
benefit incidence for the purpose of measuring changes in labor costs and average
hourly employer cost per employee. National Longitudinal Surveys are designed to
gather information at multiple points in time on the labor market activities and other
significant life events of several groups of men and women and to serve as an impor-
tant tool for economists, sociologists, and other researchers. The Establishment
Survey collects information on employment hours and earnings for nonagricultural
business establishments. The survey on Occupational Outlook provides information
on future employment opportunities for a variety of occupations, projecting to
approximately ten years ahead. Details of the sampling and related methodologies
used by the BLS can be found in their Handbook of Methods.
Opinion polls are constantly in the news, and the names Gallup and Harris have
become well known to everyone. These polls, or sample surveys, reflect the attitudes
and opinions of citizens on everything from politics and religion to sports and enter-
tainment. Gallup (http://www.gallup.com/) specializes in tracking the public’s atti-
tudes concerning virtually every political, social, and economic issue of the day, in-
cluding highly sensitive or controversial subjects. The organization prides itself in
4 Chapter 1 Introduction
the fact that these polls have always been carried out independently and objectively,
with no polls paid for by special interest groups. Best known for the Harris Poll®,
Harris Interactive® (www.harrisinteractive.com) is a worldwide market research and
consulting firm that has pioneered the use of the Internet for market research.
The Nielson Company (http://www.nielsen.com/) uses sampling in a variety of
interesting and important ways. ACNielsen provides market research, information,
and analysis to the consumer products and service industries. Nielsen Media Re-
search, the famous TV ratings company, provides television-audience measurement
and related media research services. Nielsen NetRatings provides Internet-audience
measurement and analysis, an increasingly important component in the modern age.
Numerous research centers at universities are known for their expertise in sam-
pling, among them the National Opinion Research Center (NORC) at the University
of Chicago and the Survey Research Center (SRC) at the University of Michigan.
NORC (http://www.norc.uchicago.edu/) engages in a variety of studies for govern-
ment agencies, educational institutions, foundations, and private corporations (in-
cluding a study of the Florida voting fiasco of 2000) but is probably best known for
the General Social Survey (GSS). The GSS assesses social changes in contemporary
America through a standard core of demographic and attitudinal variables, plus top-
ics of special interest selected for rotation. The SRC (http://www.isr.umich.edu/src/)
specializes in interdisciplinary social science research involving the collection and
analysis of data from scientific sample surveys, with a good balance among basic re-
search, applied survey-based research, and the propagation of the scientific method
of survey research through teaching and training.
Businesses conduct sample surveys for their internal operations, in addition to
using government surveys for crucial management decisions. Auditors estimate ac-
count balances and check on compliance with operating rules by sampling accounts.
Quality control of manufacturing processes relies heavily on sampling techniques.
One particular area of business activity that depends on detailed sampling activ-
ities is marketing. Decisions on which products to market, where to market them, and
how to advertise them are often made on the basis of sample survey data. The data
may come from surveys conducted by the firm that manufactures the product or may
be purchased from survey firms that specialize in marketing data. The Market Re-
search Corporation of America, for example, provides many types of marketing data
through the use of surveys, but some of the more interesting results come from its
National Menu Census. This survey samples families and observes their eating pat-
terns for two weeks. As many as 4000 families may participate during a year. Data
are obtained on the number of times a particular food item is served, how it is served,
how many people eat the item, and many other details, including what happens to the
leftovers. Such details are important for product development and advertising.
Many interesting examples of the practical uses of statistics in general and sam-
pling in particular can be found in Statistics: A Guide to the Unknown (Tanur, 1989).
In this book, you might want to look at some of the methods and uses of opinion
polling discussed in the articles “Opinion Polling in a Democracy” by George Gallup
and “Election Night on Television” by R. F. Link. Those interested in wildlife
ecology should read “The Plight of the Whales” by D. G. Chapman. Find out how
interrailroad and interairline billing is handled economically through sampling by
reading “How Accountants Save Money by Sampling” by John Neter.
Introduction 5
Because the objective of modern statistics is inference, you may question what
particular aspect of statistics will be covered in a course on sample survey design.
The answer to this question is twofold. First, we focus on the economics of purchas-
ing a specific quantity of information. More specifically, how can we design sam-
pling procedures that reduce the cost of a fixed quantity of information? Although
introductory courses in statistics acknowledge the importance of this subject, they
place major emphasis on basic concepts and on how to make inferences in specific
situations after the data have been collected. The second distinguishing feature of our
topic is that it is aimed at the particular types of sampling situations and inferential
problems most frequently encountered in business, the social sciences, and natural
resource management (timber, wildlife, and recreation) rather than in the physical
sciences.
Even the terminology of the social scientist differs from that of the physical sci-
entist. Social scientists conduct surveys to collect a sample, whereas physical scien-
tists perform experiments. Thus, we acknowledge that differences exist from one
field of science to another in the nature of the populations and the manner in which a
sample can be drawn. For example, populations of voters, financial accounts, or an-
imals of a particular species may contain only a small number of elements. In con-
trast, the conceptual population of responses generated by measuring the yield of a
chemical process is very large indeed. (You may recall that the properties of estima-
tors and test statistics covered in most introductory courses assume that the popula-
tion of interest is large relative to the sample.) Limitations placed on the sampling
procedure also vary from one area of science to another. Sampling in the biological
and physical sciences can frequently be performed under controlled experimental
conditions. Such control is frequently impossible in the social sciences, business, and
natural resource management. For example, a medical researcher might compare the
growth of rats subjected to two different drugs. For this experiment, the initial
weights of the rats and the daily intake of food can be controlled to reduce unwanted
variation in the experiment. In contrast, very few variables can be controlled in com-
paring the effect of two different television advertisements on sales for a given prod-
uct; no control is possible when studying the effect of environmental conditions on
the number of seals in the North Pacific Ocean.
In summary, this book is concerned with the peculiarities of sampling and infer-
ence commonly encountered in business, the social sciences, and natural resource
management. Specifically, we consider methods for actually selecting the sample
from an existing population and ways of circumventing various difficulties that arise.
Methods for designing surveys that capitalize on characteristics of the population are
presented along with associated estimators to reduce the cost for acquiring an esti-
mate of specified accuracy.
Chapter 2 presents some of the basic terminology of sampling, as well as a dis-
cussion of problems arising in sample survey design. Chapter 3 reviews some of the
basic concepts encountered in introductory statistics, including the fundamental
role that probability plays in making inferences. Simple random sampling, familiar
to the beginning student, is carefully presented in Chapter 4; it includes physical
procedures for actually selecting the sample. Subsequent chapters cover economical
methods for selecting a sample and associated methods for estimating population
parameters.
6 Chapter 1 Introduction
In reading this book, keep in mind that the ultimate objective of each chapter is
inference. Identify the sampling procedure associated with each chapter, the popula-
tion parameters of interest, their estimators, and the associated bounds on the errors
of estimation. Develop an intuitive understanding of and appreciation for the benefits
to be derived from specialized sampling procedures. Focus on the broad concepts and
do not become hypnotized by the formulas for estimators and variances that some-
times are unavoidably complicated. In short, focus on the forest rather than the trees.
Work some exercises, and the details will fall into place.
2
Elements of the Sampling Problem
2.1
Introduction
The objective of sample surveys is to make inferences about a population from in-
formation contained in a sample selected from that population. The inference often
takes the form of estimating a population mean (such as mean income per household)
or proportion (such as proportion of voters favoring a certain issue), and these are the
two types of problems that are discussed most in this book. We begin by considering
the particular problem of sampling from a finite collection of measurements (popu-
lation), and then observe what happens as the population size gets larger. In most
cases, the inference is in the form of an estimate of a population parameter, such as a
mean, total, or proportion, with a bound on the error of estimation. For those more
interested in methodology than theory, intuitive arguments are given whenever pos-
sible to justify the use of estimators.
In the first part of our discussion of the sampling problem (Section 2.2), we
introduce certain technical terms common to sample surveys. Next, in Section 2.3
we discuss how to select a sample from the population.
Each observation, or item, taken from the population contains a certain amount
of information about the population parameter or parameters of interest. Because in-
formation costs money, the experimenter must determine how much information he
or she should buy. Too little information prevents the experimenter from making
good estimates, whereas too much of it results in a waste of money. The quantity of
information obtained in the sample depends on the number of items sampled and on
the amount of variation in the data. This latter factor can be controlled somewhat by
the method of selecting the sample, called the design of the sample survey. The de-
sign of the survey and the sample size determine the quantity of information in the
sample pertinent to a population parameter, provided that accurate measurements are
obtained on each sampled element. Several sample survey designs are introduced in
Section 2.3.
If accurate measurements are not obtained on each element of the survey, then
other errors are introduced. Sources of these errors are discussed in Section 2.4.
7
8 Chapter 2 Elements of the Sampling Problem
2.2
Technical Terms
Technical terminology is kept to minimum in this book; however, certain common
terms must be defined. Let us introduce these terms by way of an example. In a cer-
tain community, an opinion poll was conducted to determine public sentiment toward
a bond issue in an upcoming election. The objective of the survey was to estimate the
proportion of voters in the community who favored the bond issue.
DEFINITION 2.1
An element is an object on which a measurement is taken. ■
DEFINITION 2.2
A population is a collection of elements about which we wish to make an
inference. ■
The population in our example is the collection of voters in the community. The
characteristic (numerical measurement) of interest for each member of this popula-
tion is his or her preference on the bond issue.
An important task for the investigator is to carefully and completely define the
population before collecting a sample. The definition must include a description of
the elements to be included and a specification of the measurements to be taken be-
cause these two components are interrelated. For example, if the population in the
bond issue study consists of registered voters, then we may want to collect informa-
tion on whether or not each sampled person plans to vote in the upcoming election.
Sampling the desired target population is not always possible, and the investigator
may have to collect additional information so that answers can be provided for ques-
tions of interest. In our example, if the population available for sampling is a list of
residents of the community, then information about whether each sampled person is,
in fact, a registered voter should be collected.
DEFINITION 2.3
Sampling units are nonoverlapping collections of elements from the population that
cover the entire population. ■
2.2 Technical Terms 9
In the bond issue example, a sampling unit may be a registered voter in the com-
munity. However, a more efficient process may be to sample households, which are
collections of elements, in order to obtain information on voter preferences. If house-
holds are the sampling units, they must be defined so that no voter in the population
can be sampled more than once and so that each voter has a chance of being selected
in the sample.
As the definition states, sampling units should be nonoverlapping. However, sit-
uations do arise in which the nonoverlapping condition is virtually impossible to
achieve. Field plot samples taken, for example, in studies of animal habitats are often
circular. The circular pattern is a convenient one to lay out and has advantages in
terms of the amount of walking necessary to study the plot. Obviously, circular plots
cannot cover a field without some overlap. The intent here is to suggest that the over-
lap should be as small as possible for efficient sampling.
If each sampling unit contains one and only one element of the population, then
a sampling unit and an element from the population are identical. This situation
arises if we sample individual voters rather than households within the community.
DEFINITION 2.4
A frame is a list of sampling units. ■
If we specify the individual voter as the sampling unit, a list of all registered vot-
ers may serve as a frame for a public opinion poll. Note that this frame will not in-
clude all the elements in the population because updating the list daily is impossible.
If we take the household as the sampling unit, then a telephone directory, a city di-
rectory, or a list of household heads obtained from census data can serve as a frame.
All these frames have some inadequacies. The lists will not be up to date. They
will contain many names of unregistered household heads, and hence a sample
drawn from the lists will contain many units that are not in the population. Also,
some registered voters may not appear on any of these lists. It is hoped, however, that
the gap between the frame and the population is small enough to permit inferences to
be made about the population on the basis of a sample drawn from the frame.
Some sampling schemes may involve multiple frames. In sampling voters, we
could start by sampling city blocks, then sampling housing units within those blocks,
and finally sampling voters within the selected housing units. One frame, then, is a list
of city blocks, and the second frame is a list of housing units within those blocks. The
second frame may not be available until the blocks are selected and studied in some
detail. As another example, estimation of crop yields in a state may involve sampling
from a list of growers to be interviewed and a list of fields to be objectively measured.
A frame needn’t be a list only. A map of an Arctic island might represent a study
area, from which random locations will be selected for estimating the abundance of
nesting geese on the island. Any suitable representation of the population of interest
that allows random selection of units can serve as a frame.
DEFINITION 2.5
A sample is a collection of sampling units drawn from a single frame or from multi-
ple frames. ■
Another random document with
no related content on Scribd:
of Cortés, that he had come of his own free-will, and at the intimation
of the gods, to stay awhile with his guests. He told them to calm the
people with this assurance, and to disperse the gathering.
FOOTNOTES
[490] They had now been four days in Mexico, without going farther than the
palace, says Bernal Diaz. A page named Orteguilla, who had already acquired a
smattering of Aztec, was sent with the interpreters to ask this favor. Hist. Verdad.,
69.
[491] Soldiers who had been in Rome and Constantinople declared that never had
they seen so large and orderly a market, with so large an attendance. Bernal Diaz
indicates the site of the plaza to have been where the church of Santiago de
Tlatelulco was erected, and this still remains under the same name, over a mile
north-west-by-north of the central plaza of Mexico. Hist. Verdad., 70-1. The old
maps of Mexico already spoken of give the same site, and Alaman’s investigations
point out correctly the street which led and leads to it, although he has failed to
notice the above authorities, which give the very site. Disert., ii. 282-5.
[492] It has been generally accepted that the temple in the centre of the city was
visited, but Bernal Diaz, who is the only narrator of this excursion, states distinctly,
in several places, that the pyramid ascended was situated in the Tlatelulco
market-place, ‘adonde está aora señor Santiago, que se dize el Tlatelulco.’ Hist.
Verdad., 70-1. The description of the temple court and interior is somewhat
confused, and evidently combines points which belong to the central temple.
[494] Bernal Diaz, Hist. Verdad., 70-1. Sigüenza y Góngora, the well-known
Mexican scholar of the seventeenth century, follows Bernal Diaz. Anotaciones
Crit., MS., 1-2.
[496] He refers to this promise in the second letter to the king, saying, ‘porque
certifiqué á V. A. que lo habria preso ó muerto ó súbdito.’ Cortés, Cartas, 52.
[497] ‘En la verdad era así é lo tinien acordado,’ affirms Tapia, Rel., in Icazbalceta,
Col. Doc., ii. 579. ‘Estas nueuas, falsas, o verdaderas,’ is the non-committing
phrase of Gomara. Hist. Mex., 123. Ixtlilxochitl takes firmer ground. ‘According to
an original letter in my possession, signed by the three heads of New Spain, and
written to his Majesty the emperor, our master, they exculpate Motecuhzoma and
the Mexicans of this and other charges, declaring them inventions of the
Tlascaltecs and of some Spaniards who feared that they would never see the hour
when they might leave the city and place in security the riches they had obtained.’
Hist. Chich., 296. Clavigero adopts the view that Cortés called for such testimony
from certain chiefs among his allies, whose dislike of the Mexicans would be sure
to prompt it. ‘Per giustificar vieppiù il suo attentato, e muovere i suoi Spagnuoli ad
eseguirlo, fece chiamar parecchie persone principali de’ suoi alleati (la cui
informazione dovrebbe sempre essergli sospettosa).’ Storia Mess., iii. 90-1.
Vetancurt has a story that, a drought prevailing at the time, the Spaniards induced
the heavens by means of masses and prayers to send rain. This made the priests
and idols jealous, and the emperor was prevailed on to rid himself of the
Spaniards. The Tlascaltecs learned of the plot and reported it. Teatro, pt. iii. 130.
[498] According to Bernal Diaz the members of this council suggested not only the
seizure but the reasons for it. Cortés responded that he had not been oblivious of
the danger, but saw not how the seizure could be effected. The captains proposed
to beguile the prince to their quarter and detain him. If Cortés hesitated they were
willing to undertake the task. The old soldier is evidently misled, as he was in the
scuttling affair, to assume too much credit for himself and his fellow-soldiers.
Cortés had no doubt adopted his common tactics, so frequently admitted by
Bernal Diaz himself, of inspiring his comrades to suggest what he had resolved
on. This is proved by the promise made to the emperor in his first letter, four
months before, to capture Montezuma.
[499] Bernal Diaz, Hist. Verdad., 73, followed by a number of other writers, states
that the letter was received at Mexico the morning after this meeting, but it has
been shown that he must be wrong. He mentions as one of the statements in the
letter that the Totonacs were in revolt, and it is probable that Cortés may have said
so to show the soldiers that retreat was cut off, and that seizure was the only
recourse.
[501] The neglect of this less speedy plan does indicate that the rumors of danger
were credited to a great extent at least.
[502] Hist. Verdad., 74. Cortés in his quiet way writes to the king that, after
passing six days in the great city, and finding that for various reasons ‘it suited the
royal interest and our safety to have this lord in my power, and not wholly at liberty,
lest he change in the proposal and desire manifested to serve Y. H.... I resolved to
seize and place him in the quarter where I was.’ Cartas, 88-9.
[503] ‘Mandó que su gente dos á dos ó cuatro á cuatro se fuesen tras él ... é con
él entramos hasta treinta españoles é los demas quedaban á la puerta de la casa,
é en un patio della,’ says Tapia, who appears to have been one of those who
entered. Rel., in Icazbalceta, Col. Doc., ii. 579.
[504] ‘Cõ armas secretas,’ says Gomara, Hist. Mex., 123, and probably they did
bear extra weapons beneath their cloaks.
[505] ‘Y otras hijas de señores á algunos de mi compañía.’ Cortés, Cartas, 89. But
the customary mark of favor was to give them from his harem. See Native Races,
ii.
[506] ‘Per non dar disgusto al Re, e per avere occasione di farla Cristiana,’ is
Clavigero’s excuse for the acceptance. Storia Mess., iii. 93. Brasseur de
Bourbourg, and some other writers, assume that Cortés declined; but the original
authorities all say or intimate that he accepted. Even Cortés himself writes in his
letter to the emperor, ‘despues ... de haberme él dado algunas joyas de oro y una
hija suya,’ etc. Cartas, 89. ‘Le persuadió,’ says Ixtlilxochitl, Hist. Chich., 290.
Gomara is even more explicit, and Herrera says that Montezuma insisted, ‘porque
queria tener nietos de hombre tan valeroso.’ dec. ii. lib. viii. cap. ii. The affair is
perhaps less important in itself than as index to the character of Cortés, who could
accept so intimate an offer with one hand while he prepared a blow with the other.
It might also be made to indicate that Montezuma could have had no base designs
against him when he made the uncalled-for offer of intrusting a daughter (if such
she was) to his keeping. Still the imperial character would not have suffered had it
been shown that this was but an artifice to lull his intended victim into a false
security.
[507] Some authors, like Herrera and Torquemada, say that he denied all
knowledge of the occurrence, calling it an invention of enemies.
[508] Ixtlilxochitl, Hist. Chich., 297. ‘Q̄ tenia la figura de Vitzilopuchtli,’ Gomara,
123; and so says Bernal Diaz. Tapia states that Montezuma told Cortés to send
two Spaniards with the messengers; but he doubtless declined to risk two lives on
such a trip. Rel., in Icazbalceta, Col. Doc., ii. 583-4.
[509] ‘Ingrato rey!’ exclaims Gallo, in commenting upon this surrender of a devoted
officer. Hombres Ilust. Mex., i. 318.
[510] Bernal Diaz makes Cortés accuse the emperor of perfidy, and of having
instigated the Cholula massacre. Hist. Verdad., 74. But this he would hardly do,
since his purpose was clearly to persuade, not to arouse anger.
[511] ‘No querria començar guerra, ni destruir aquesta Ciudad ... que si alboroto, ò
vozes daua, que luego sereis muerto de aquestos mis Capitanes, que no los
traigo para otro efeto,’ is Bernal Diaz’ blunt version. Hist. Verdad., 74.
[512] ‘Auia tenido platica de su idolo Huichilobos ... que convenia para su salud, y
guardar su vida, estar con nosotros.’ Id., 75.
[513] ‘Estuuieron mas de media hora en estas platicas.’ Id., 74. Tapia and others
say four hours, which is unlikely.
[514] ‘Porque mas vale que desta vez asseguremos nuestras vidas, ò las
perdamos.’ Bernal Diaz, Hist. Verdad., 75.
[515] ‘Era, ó muy falto de ánimo, ó pusilánime, ó muy prudente.’ Oviedo, iii. 289.
‘En él se cumplió lo que de él se decia, que todo hombre cruel es cobarde,
aunque á la verdad, era ya llegada la voluntad de Dios.’ Ixtlilxochitl, Rel., 411.
[516] Bernal Diaz intimates clearly enough that no demonstration was made till
after his arrival. Hist. Verdad., 75. And so does Cortés. ‘Llorando lo tomaron en
ella [the litter] con mucho silencio, y así nos fuimos hasta el aposento donde
estaba, sin haber alboroto en la ciudad, aunque se comenzó á mover. Pero sabido
por el dicho Muteczuma, envió a mandar que no lo hubiese; y así, hubo toda
quietud.’ Cartas, 90. Ixtlilxochitl, however, allows Montezuma to stay long enough
in his palace, after ordering the litter, to enable the lords and nobles to come and
offer their services. A delay like this, which the Spaniards certainly never could
have permitted, might have given time for the tumultuous gathering which he
describes. Hist. Chich., 297. Prescott, in following this version, makes the emperor
so far overstep his usual dignity as to ‘call out’ to the people to disperse. ‘Tambien
detuvieron consigo á Itcuauhtzin, gobernador del Tlatilulco,’ says Sahagun, while
the leading nobles ‘cuando fue preso Mocthecuzoma le desampararon y se
escondieron.’ Hist. Conq., 25.
The seizure has, like the equally prominent episodes of the massacre at
Cholula, and the scuttling of the fleet, aroused no little comment in justification or
condemnation. ‘Now that I am old,’ says Bernal Diaz, ‘I stop to consider the heroic
deeds then performed, and I do say that our achievements were not effected by
ourselves, but were all brought about by God; for what men have existed in the
world who, less than 450 soldiers in number, dared to enter into so strong a city as
Mexico, larger than Venice, and so remote from Castile, to seize so great a lord?’
Hist. Verdad., 76. ‘Never Greek or Roman, nor of other nation, since kings exist,
performed a like deed, only Fernando Cortés, to seize Motecçuma, a king most
powerful, in his own house, in a place most strong, amid an infinity of people,
while possessing but 450 companions.’ Gomara, Hist. Mex., 124. Commenting on
this, Torquemada adds that ‘it was indeed a deed for daring never seen, and must
be attributed to God rather than to human heart.’ i. 458. Solis of course fails not to
extol the genius and daring of his hero, whose deed ‘appears rather in the light of
a fable’ than in consonance with simple history. Hist. Mex., i. 448. ‘A deed which
makes one tremble even to conceive, and much more to carry out. But God had
so determined it.’ Ixtlilxochitl, Hist. Chich., 296. ‘History contains nothing parallel to
this event, either with respect to the temerity of the attempt, or the success of the
execution,’ etc. Robertson’s Hist. Am., ii. 60. ‘An expedient, which none but the
most daring spirit, in the most desperate extremity, would have conceived.’
Prescott’s Mex., ii. 159. ‘An unparalleled transaction. There is nothing like it, I
believe, in the annals of the world.’ Helps’ Cortés, ii. 351. Clavigero is less carried
away by the incident, for he sees therein the hand of God. Nevertheless, he
sympathizes with Montezuma. Storia Mess., iii. 95, etc. Pizarro y Orellana finds
the deed eclipsed by the similar achievement, with a smaller force, under his
namesake Pizarro. Varones Ilvstres, 89-90. And later Mexican writers, like
Bustamante, see, naturally enough, nothing but what is detestable in the incident,
for according to the native records which form their gospel, Montezuma was
guiltless of any base intents. Unfortunately for them, these very records paint him
a blood-thirsty despot who punishes the slightest offence against himself, even
when merely suspected, with the most atrocious cruelty; one who is continually
seeking his aggrandizement at the expense of inoffensive, peace-loving tribes,
who oppresses not only conquered peoples, but his own subjects, with
extortionate taxes and levies to satisfy his inordinate appetite for pomp and for
new conquests. These records also admit that he had repeatedly sent sorcerers, if
not armies, to entrap and destroy the Spaniards. He who looked calmly on
hecatombs of his own subjects, slaughtered before his very eyes, would not
hesitate to condemn strangers for plotting against the throne which was dearer to
him than life itself. The Spaniards may have anticipated events considerably, but
there is no doubt that numerous personages, from Cuitlahuatzin downward, were
bitterly opposed to their enforced guests, and they would sooner or later have
realized the rumors which the allies began to circulate. Placed as he was, Cortés’
duty to himself, to the men intrusted to him, to his king, and to the cause of
religion, as then regarded, required him to give heed to such rumors, and, after
weighing their probability, to take the precautionary measure of seizing the
monarch, since retreat not only appeared fraught with disaster and dishonor, but
would be regarded as a neglect of opportunity and of duty. With Cortés, naught but
the first steps in assuming the conquest, and in usurping certain credit and means,
can be regarded as crimes, and the former of these was forced upon him by
circumstances of his age and surroundings. Every project, then, conceived by him
for the advancement of his great undertaking must redound to his genius as
soldier and leader. Of course, among these projects appear many which did not
advance the great object, and which must be condemned. But where do we find
greatness wholly free from stain?
CHAPTER XVIII.
DOUBLY REFINED DEALINGS.
1519-1520.