the use of microcomputers for census data processing (onu, 1989)

8/14/2019 The Use of Microcomputers for Census Data Processing (ONU, 1989)

1/60

DEPARTMENT OF INTERNATIONAL ECONOMIC AND

SOCIAL AFFAIRS

STATISTICAL OFFICE

WORKING PAPER

THE USE OF MICROCOMPUTERS

FOR CENSUS DATA PROCESSING

UNITED NATIONS

New York, 1989


2/60

ii

PREFACE

The present publication contains a description of the use of a new computer tech-nology, microcomputers, for processing population census and housing data.1

This document has been prepared in response to the continuing concern of theUnited Nations Fund for Population Activities and the United Nations Statistical Officeto provide appropriate and cost effective advice to developing countries in the preparationfor and execution of their censuses of population and housing. The rapid advances ofcomputing technology in the last decade, especially those emerging in the area of micro-computer technology, provide significant new opportunities for developing countries toimprove substantially the effectiveness of their statistical data processing operations.Such improvements are capable of being realized in all phases of operation, from plan-ning and management through data processing to data and analytical services that cannow be delivered to end users of the data.

The publication consists of seven chapters and an annotated bibliography of furtherreferences in the subject and related areas. Chapter 1 is an introduction to the paper andvarious aspects of the subject material. Chapters 2 and 3 focus upon the technologicalfactors that are leading to the growth of microcomputer based markets, the substantial de-centralization of functions that the technology encourages, and the redistribution of re-sponsibilities and requirements that these shifts produce. Chapter 4 discusses the applica-tion of microcomputers to specific areas in census data processing, with emphasis uponthe functionality of the software that should be used in specific areas to obtain maximumassistance from a computer based processing strategy. Chapter 5 raises a number of newissues that arise in introducing microcomputers into statistical and other data processingactivities. Chapter 6 provides a glimpse into a number of areas in which microcomputersmay be useful in the future, and which are likely to be of some importance soon for na-tional statistical offices. Chapter 7 provides a concluding summary of the document.

The work for this and a complementary companion paper 2were sponsored by theUnited Nations Fund for Population Activities.

1This document has been prepared by George Sadowsky, who is currently Director of Academic Computing

and Network Services at Northwestern University, Evanston, Illinois, U.S.A., and was formerly a staff memberof the United Nations Statistical Office as Technical Adviser in Computer Methods. The opinions expressed inthis paper are those of the author and are not necessarily those either of The United Nations or of NorthwesternUniversity.

2 International Statistical Programs Center, Population Census Data Processing on Microcomputers.Prepared for and available from the United Nations under project INT/88/P09. U.S. Bureau of the Census,February 1989.


3/60

iii

CONTENTS

Page

1. INTRODUCTION..............................................................................................1

Background.................................................... ....................................... 1Objective..............................................................................................1

2. MICROCOMPUTERS AND COMPUTING IN DEVELOPING COUNTRIES .....................................3

Background.................................................... ....................................... 3Computing in Developing Countries ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... .. 5

Physical infrastructure...................................................................... 5Human resources infrastructure. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 6Size of market........................................... ....................................... 7Information resources .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 7

Microcomputers and Developing Countries.................................................8

3. STRATEGIC ISSUES IN THE USE OF MICROCOMPUTERS................................................. 10

Overview................................................................................... ......... 10Responsibilities of Microcomputer Operations ... ... ... ... ... ... ... ... ... ... ... ... ... ... . 10Requirements Analysis..... ..... ...... ..... ...... ..... ..... ...... ..... ...... ..... ...... ...... ... 11System Specification and Negotiation with Suppliers ... ... ... ... ... ... ... ... ... ... ... . 11Site Preparation............................................... ..................................... 12Installation and Testing...... ..... ...... ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... . 13System Programming.................... ........................................................ 13System Resource Management.......... ...... ..... ...... ...... ..... ...... ..... ...... ...... ... 14Education and Training...... ..... ...... ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... . 15

Selection of students and curriculum.................................................. 15Conceptual and operational training.................................................. 16Local and foreign training.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 17

Indigenous training capacity .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 17Staff continuity............................................................................... 18Sources of training ...... ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 18

Operating Services...... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... . 19Hardware Maintenance Services...... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 20Applications Implementation and Management ... ... ... ... ... ... ... ... ... ... ... ... ... ... 21Capacity Planning and Configuration Management.. ... ... ... ... ... ... ... ... ... ... ... .. 21Summary.................. ........................................................ .................. 22

4. ADAPTATION OF MICROCOMPUTERS TO CENSUS DATA PROCESSING ................................. 23

Background and History ...... ..... ..... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 23History of census data processing on microcomputers .. .. .. .. .. .. .. .. .. .. .. .. .. . 23

Census Management........................................ ..................................... 24Project planning............................................................................. 24Geographic data base ..... ...... ..... ...... ..... ..... ...... ..... ...... ..... ...... ...... ... 25Cartographic data base................................................................... 26Logistics. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 26Personnel control and payment... ...... ..... ...... ...... ..... ...... ..... ...... ...... ... 26

Cycles of Data Collection and Processing Activity..... ... ... ... ... ... ... ... ... ... ... ... 26Pilot census................................................................................... 26Long and short forms .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 26Post-enumeration survey..... ..... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 27

Data Collection and Recording...... ..... ...... ..... ..... ...... ..... ...... ..... ...... ...... ... 27


4/60

iv

Questionnaire design ..... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... ...... ... 27Data entry.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 28Error correction philosophy.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 29Data control and the unit of processing .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 29

Data Editing.............. ........................................................ .................. 31Determining appropriate edit rule sets... .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 31Performing large scale editing.. ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 31

Data Tabulation ....................................................... ............................ 32Preparation of Census Publications ..... ...... ..... ..... ...... ..... ...... ..... ...... ...... ... 33The Post-Enumeration Survey......... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 34Demographic Analysis ........................................................ .................. 35Data Dissemination and Data Bases ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... . 35

Dissemination and use of machine readable results.... .. .. .. .. .. .. .. .. .. .. .. .. .. . 36Data bases consisting of aggregate tables .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 36Data bases consisting of microdata... .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 37

Choosing a Processing Strategy...... ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... . 38General considerations.................................................................... 38Target computing environment..... ...... ..... ..... ...... ..... ...... ..... ...... ...... ... 38Software strategies ...... ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 39

5. SPECIFIC ISSUES RAISED BY USE OF MICROCOMPUTER TECHNOLOGY................................ 41

Issues of Data Control and Ownership.. ...... ...... ..... ...... ..... ...... ...... ..... ...... . 41Data security .................................................. ..................................... 41Data Integrity ................................................. ..................................... 42Data Confidentiality........ ..... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... ...... ... 43Revised Selection Criteria for Computing Environments.. ... ... ... ... ... ... ... ... ... . 44Micro-Mainframe Connectivity and Co-operative Processing.... .. .. .. .. .. .. .. .. .. .. 45Software as Intellectual Property ..... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 45

6. FUTURE OPPORTUNITIES FOR THE USE OF MICROCOMPUTER TECHNOLOGY ......................... 47

New Methods of Data Dissemination ...... ..... ...... ..... ..... ...... ..... ...... ..... ...... 47Portable Data Entry in the Field...... ..... ...... ...... ..... ...... ..... ...... ...... ..... ...... . 47Geographic Information Systems.......... ..... ...... ..... ...... ..... ...... ...... ..... ...... . 48

Thematic mapping.......................................................................... 49Computer-assisted Cartography....... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 50

Collection mapping......................................................................... 50Socioeconomic Microsimulation ..... ...... ..... ...... ..... ...... ...... ..... ...... ..... ...... 50

7. CONCLUSION................................................ .............................................. 52

8. ANNOTATED BIBLIOGRAPHY................................................. ............................ 54


5/60

5

The Use of Microcomputers for Census Data Processing

1. INTRODUCTION

Background

The year 1990 marks the one hundredth anniversary ofthe first application of automatic computing equipment topopulation censuses in the world. In 1880, frustrated by thelong delays inherent in processing manually census datacollected from a population of then about 45,000,000 per-sons, the U.S. Bureau of the Census commissioned HermanHollerith to design and construct data processing machinerycapable of tabulating census results rapidly and automati-cally. Hollerith's equipment was available in time for theU.S. 1890 census, and was successful in reducing theoverall processing time by approximately a factor of four.

Hollerith's electromechanical accounting equipment of

1890 evolved in complexity, capacity, and speed, and foundincreasing uses in other areas of government and business aswell. A major step forward occurred in the 1940's with theinvention of the electronic digital computer. Operations thattook seconds on electromechanical processing equipmentnow took thousandths of seconds on the newly invented all-electronic equipment. The first commercial computingcompany, the Eckert-Mauchly Computer Company, wasfounded in 1947 and was absorbed by the Remington RandCorporation in 1949. So important was the population cen-sus that the first unit of its first computer model, a Univac I,was delivered to the U.S. Bureau of the Census in 1951 andwas used to process the 1950 U.S. Decennial Census ofPopulation.

Today we take for granted the use of electronic comput-ing equipment not only for processing census data, but alsofor processing data from many large scale statistical andbusiness operations. Much of the world's administration,research, and commerce depends critically for success uponcurrent large scale computing equipment that is directly de-scended from the pioneering developments of the 1940's.The resulting computing techniques can provide substantialassistance to developing countries in improving the qualityof governmental administration, the rate of economic devel-opment, the effectiveness with which scarce national re-sources are managed, and the adequacy of the statistical in-formation base for policy planning and implementation.

Major changes have taken place in electronic computingtechnology since the 1940's. Computer hardware has be-come smaller, more reliable, and less expensive each year;over the entire period, the performance-price ratio has in-creased by about 25-30% each year. Likewise, computersoftware, i.e. the programs and user environments that in-struct the hardware to perform useful work, has evolvedsubstantially in the direction of ease of learning and use.Such evolution has made possible the widespread introduc-

tion to and use of computers by individuals who are not

computer professionals and who have little if any technicaltraining.

Prior to the 1980's, statistical data processing was per-formed on either mainframe computers or minicomputers.These computers were large, complex, and expensive. Inaddition to their large capital cost, they required substantialinvestments in site preparation and operation in order tofunction properly. They also required staff withprofessional skills in computer operations, and system andapplication programming in order to exploit theircapabilities usefully.

In the late 1970's and early 1980's, a different computing

environment began to emerge, one that was qualitativelyand quantitatively very different from its predecessors. Mi-crocomputersare characterized by small physical size, rela-tively low cost, tolerance toward a broad range of environ-mental conditions, and a lessened dependence upon special-ized technical personnel. Microcomputing today is oftencharacterized as desktop computing, reflecting the fact thatmicrocomputers are sufficiently small to occupy only a frac-tion of the work surface of a typical office desk. Themethod of use of microcomputers is different; it providespersonalcomputing, in which the new computer systems areoriented to serve one person at a time, employing a userinterface often more powerful than what was previouslyavailable on microcomputers or minicomputers.

The rapid evolution of microcomputing technology andits increasing importance within the overall spectrum of in-formation processing is much more than just a continuationof the overall technical progress in computing technology.Size, cost, reliability and convenience have reached thepoint where there has been a very significant qualitativechange in the nature of computing services available. Theemergence of a mass market in hardware and software andthe resulting explosion of entrepreneurial effort and prod-ucts, as well as the emergence of radically different andmore productive computer interfaces and environments, canbe regarded more as a fundamental revolution ininformation technology than as a continuation of earlier

trends.

Objective

This document addresses the nature of this revolutionand its implications in particular for the use ofmicrocomputers for census data processing in developingcountries. Much of this material is also applicable to otherforms of statistical data processing, and a part is applicableto the use of microcomputers by anyone, either in


6/60

INTRODUCTION

6

developing or developed countries, for any of a wide rangeof applications.

The primary objective of this document is to provide aguide to experience in the use of microcomputer technologyfor the processing of censuses of population and housing in

developing countries. The primary audience for the docu-ment consists of administrators and managers of statisticalactivities who may wish to exploit this technology, and thetext is written with the intent of providing a view for themof the managerial and operational implications of such usebased upon technical and product alternatives that currentlyexist.

There is a substantial secondary audience toward whomthis document is also addressed, consisting of technical andprofessional staff members who may be responsible for de-tailed implementation of microcomputer-based statistical ac-tivities, including censuses of population and housing. Inpractice, these two groups must work together in order toimplement successfully any large scale statistical operationsuch as a census, and their spheres of knowledge must over-lap to some extent so that each group will have anappreciation of considerations affecting the overallprocessing strategy. In order to address the concerns of thesecond type of reader as well as enrich the technical knowl-edge of the first, specific technical material and technicalelaboration of text appears for the most part in footnotesthroughout the document.

This document is a companion to the document, Popula-tion Census Processing on Microcomputers , produced bythe International Statistical Programs Center of the U.S.Bureau of the Census. That document is an extension of thePOPSTAN series, which provides a thorough and com-prehensive operational treatment of census activities in ahypothetical country. In contrast to the ISPC document, thisdocument does not provide a methodical or a cookbookapproach to census data processing activity; rather, it

attempts to reflect actual experience gained in developingcountries that have used microcomputers for census dataprocessing as well as for other statistical activities.

It is important to note that this version of this publicationhas been written in early 1989. Its contents therefore reflect

the state of computing technology, both hardware andsoftware, at that time, as well as the perceived state ofcensus data processing knowledge and practice then indeveloping countries. The material contained in thispublication is therefore, like much information in computingand data processing, subject to relatively rapid decay. In afew years, the products and other options available arelikely to combine to make this publication increasinglyobsolete.

Developing countries are a relatively heterogeneousgroup, each having strengths and weaknesses that affecttheir rate and direction of development. The presentdocument attempts to span a wide variety of skills and needspresent in developing countries, with the knowledge that inits entirety, it is not applicable directly to any one suchcountry. Nevertheless, it is hoped that much of the contentwill be useful for obtaining an effective and operationalunderstanding of the role that microcomputer technologycan play in census data processing, and the various ways inwhich developing countries can harness this technology topromote their interests. In this regard, this documentprovides a technical and organizational overview of the newtechnology and its implications, and a context within whichcountries can understand the alternatives available to themand make informed choices among them.

Finally, this document tells only a small part of a muchlarger story involving statistics, census data processing, andcomputing products and techniques. An annotated bibliog-raphy is provided at the end of this document, with annota-tions providing some guidance regarding the content andrelevance of material in each of the references.


7/60

7

2. MICROCOMPUTERS AND COMPUTING IN DEVELOPING COUNTRIES

Background

The microcomputer environment available today is theresult of technical progress that started in the mid-1940's

with the development of the first electronic digital com-puter. In the intervening four decades, electronic data pro-cessing (computing) equipment has become affordable andusable by most if not all national statistical offices in theworld. This progress has been achieved largely due to arapid and sustained rate of technical progress in the underly-ing electronics technology coupled with continuous intellec-tual and commercial efforts to make computing equipmenteasier to use.1 The cumulative effect of technical progressover 40 years has driven the cost of raw computationalpower and capacity to very low levels, while advances insoftware and the now emerging mass market for computerbased tools has caused software prices to fall almostcorrespondingly.

Two major components contribute to the power and util-ity of computing equipment, hardware and software. Thehardwarecomprising a computing system consists of phys-ical equipment including electronic circuitry, magneticstorage media, input and output devices including devicesfor user interaction such as a keyboard and screen, and de-vices for communication using external transmission facili-ties to exchange information with other computing systems.The softwarethat is executed on a computer system consistsof the specific instructions that tailor the computer to beuseful for specific tasks, such as data editing or datatabulation.

Hardware is often characterized by its generation. Firstgeneration hardware was prevalent in the 1950's and con-sisted of circuits based upon vacuum tube technology. As aresult, first generation computers were very large andexpensive, required very large amounts of electrical power,and failed frequently. Second generation hardware in theearly and mid-1960's was characterized by the use ofdiscrete transistor technology, while third generation tech-nology was characterized by some initial small scaleintegration of components in the fabrication process.

Given the substantial progress in electronics technology,it is difficult and perhaps meaningless to categorize currenthardware in terms of precise generation numbers. Each suc-cessive significant identifiable improvement has led toequipment that is smaller and less expensive, consumes lesspower, and is more reliable than its predecessors. However,a very major step forward occurred in the early 1970's withthe invention of solid state integrated circuits that could be

1While there are many measures of rates of change in the underlyingtechnology, a good rule of thumb in the past has been to assume that thecost/performance ratio of computing equipment will be declining byapproximately one-half every two years.

reproduced automatically and in large quantity by precisephotographic techniques, replacing the predominantly man-ual assembly methods that had characterized prior computermanufacture. In particular, such solid state circuits could be

designed that would include all circuitry necessary for acomplete computer processor; such circuits are calledmicroprocessors or microprocessor chips. Similarly,memory circuits and circuits for other computer hardwaretasks were designed and have provided recent and currentcomputer designers with a wide range of miniature circuitson chips from which to construct computing equipment ofall sizes.

While advances in electronics technology havebenefitted all sizes of computing systems, the emergence ofmicrochip technology have made possible economical anduseful small scale computing. The granularityor minimumsize in which computing power can be obtained has changed

radically; while in the 1960's it was not possible to purchasecomplete useful computing systems for less thanUS$250,000, complete systems having far more power cannow be purchased for US$3,000. These inexpensive com-puting systems, based upon microelectronic chips includingmicroprocessors, are generally referred to asmicrocomputersor microcomputer systems.

Microcomputers are best characterized by their physicalsize rather than by the functionality or power of computingthat they provide. Microcomputers are generally placed ona part of a desktop, although sometimes the main unit maybe placed on the floor or on a shelf nearby. Microcomputersare often modular, i.e., they may consist of separate proces-sor, display, and keyboard parts, but they may also be inte-grated. Most microcomputers are meant to be used by oneperson at a time, but there are now multi-user microcomput-ers that are starting to have an impact on statistical dataprocessing operations.2

It is more difficult to characterize what a microcomputeris by its power or functionality, since current microcomput-ers now often have as much computing power as minicom-puters and mainframes that were considered quite powerfulmachines ten or more years ago. Technical progress in theunderlying computing and electronics industries affects allscales of computing machinery, so that the power and func-tionality of any scale of computing is destined to migrate tosmaller scales of computing over time.

2In particular, variations of the Unix operating system have been

implemented that run on larger microcomputers and that provide a time-sharing environment on an inexpensive hardware platform. Although suchsystems are not in common use at present for statistical data processing, thebenefits of shared computing environments that make sharedminicomputers so useful are equally applicable to the microcomputerenvironment. Multi-user microcomputers are therefore likely to play alarger role in the future of statistical and census data processing than theydo in the present in both developed and developing countries.


8/60

MICROCOMPUTERS AND COMPUTING IN DEVELOPING COUNTRIES

8

Low cost is a characteristic of microcomputertechnology that has existed since its beginnings and that islikely to continue. Microcomputers are inexpensive relativeto larger computers; their relatively low cost makes iteconomically practical for them to be used more in a role ofassistant to individuals rather than being regarded as a

scarce resource that must be kept constantly busy to justifyits economic existence, as was generally the case in the past.

Microcomputers are physically compact and relativelylight, and can therefore be transported easily. In addition,microcomputer technology is generally not at the frontier oftechnical progress or extremely fast or capacious, making itrelatively easy to obtain export permission and licenses formicrocomputers and related equipment.

Microcomputers may be connected in such a way thatthey can exchange programs and data electronically. Themechanism by which they perform the exchange is called alocal area network, or LAN. The adjective local arearefersto the geographic compactness of the network; generallysuch networks are installed within the same room or set ofadjacent rooms in a building, and they are almost alwayscontained within the same building. They are meant toserve a specific work group, consisting of persons workingon the same or interrelated tasks. The use of a local areanetwork allows the sharing of resources and data by all mi-crocomputers attached to the network and make statisticaldata processing operations more effective, although it is nota necessary component of using microcomputers.

Local area networks generally connect a number ofsmaller microcomputers to a larger microcomputer whichfunctions as a server. This system provides services to theother computers. It is likely to have a large fixed disk, andso can provide file access and storage services for the othersystems on the network; if so, it is referred to as a fileserver. It may also have a printer attached to it that can beshared via the network connection by the other systems; itcan then be referred to as aprint server. The servers are theprimary mechanism by which resources on the network areshared.

Because of the importance of the services provided bysuch servers to the rest of the network, sometimes two suchservers are included in a network configuration so that ifone fails, the other can take over and provide the services

required to keep the network operational. Such aconfiguration is somewhat similar to having a spare centralcomputer that can be pressed into service almostimmediately should the primary central computer fail.

For the purposes of statistical data processing activities,networks characterized by moderate transmission speed willgenerally be sufficient. Such networks are moderately easyto set up and operate, and are increasingly easy to obtainexport licenses for, for shipment to most countries.

The small size and cost of microcomputers allow dataprocessing tasks to be distributed in a way that was not pos-sible with larger computer systems. Computing power canbe decentralized geographically, since each microcomputercan be made physically independent of others. Computing

power can also be specialized; specific microcomputers canbe dedicated to specific functions rather than requiring alarge central computer to perform several different func-tions. Perhaps most important, demands for additionalcomputing power can be satisfied by acquiring more micro-computer systems rather than having to replace a large com-puter system with a larger one.

Parallel changes in software technology are equally im-portant. Software consists of instructions, written by pro-grammers and users, that computing hardware executes inorder to perform tasks. These instructions are organizedinto modules commonly called computer programs. Theseprograms are executed quickly (at electronic speeds) andunforgivingly; an error in the set of instructions comprisinga program will be executed as specified at the sameelectronic speeds as the correct sequence of instructionswould have been executed. One of the major challenges ofsoftware technology today is to develop methodology thatwill allow the construction of correct programs efficiently.

Early software technology was rudimentary and expen-sive; software specialists called programmers were requiredto instruct computing systems in a language closely corre-sponding to the circuitry of the hardware on which the pro-gram would be executed. Over time, intermediate layers ofsoftware have been created which have allowed users tospecify programs and procedures to computers in a formmore closely resembling the task to be performed. The mostimportant of these intermediate layers are: (1) the operatingsystem, which controls the hardware resources of thecomputer directly; (2) language processors, which allowprograms to be written using language constructs reflectingthe logic of the program rather than the structure of thehardware; and (3) the user interface (includingdocumentation), which allows less technical users to controlthe computer directly.1 For most people familiar withcomputers today, the characteristics of the user interface andthe applications for which programs exist are the mostimportant aspects of a computer system.

Because of their small size and relatively low cost andtheir use in playing games, microcomputers are sometimesregarded as frivolous and not serious alternatives to earlier

1The operating system and the user interface often overlap. An ex-ample is provided by Apple's Macintosh operating system, which hides thebasic operating system as much as possible under a graphic, icon-orienteduser interface that is easy to use and subsumes functionally many operatingsystem characteristics. A somewhat different approach is being taken byIBM/Microsoft in their development of OS/2, in which the basic operatingsystem is readily available, but the Presentation Manager layer is availablefor users who prefer to be a level removed from operating systemcomplexity.


9/60


9

forms of computing. This was an attitude adopted in theearly days of microcomputing by many computerprofessionals who were then oriented strongly to using largecomputers, and who in addition perceived the breakdown ofcentral control over computing resources and computerbased operations that microcomputer technology would

eventually lead to.

It is extremely important to recognize thatmicrocomputers are not frivolous and are not toys; nothingcould be further from the truth. Rather, they are the leadingedge of a fundamental set of changes in informationprocessing technology that will result in a widespreaddistribution of computing power in much the same way thatthe development of electrical power led to a widespreaddistribution of mechanical power and energy.

Microcomputer technology is in many respects as power-ful as earlier mainframe computer technology. Mi-crocomputers can execute instructions at very rapid speeds,and the newer models can accommodate substantialimmediate access or primary memory. Microcomputers areless adept currently about providing substantial massstorage. They are also more limited in the area of shareduse, which occurs when a group of users requires concurrentaccess to a shared resource. Although development isoccurring in these and other areas, for many organizationsresidual deficiencies in current microcomputerenvironments will dictate an effective data processingstrategy including the use of both microcomputers andlarger computer technology.

Both computer hardware and software are evolving anddeveloping at a rapid rate. Even though today's microcom-puters are by many standards the equivalent of much largercomputers of the past, they will almost certainly continue todevelop at the same rate in the foreseeable future as theyhave in the past. This expectation has a number of impli-cations for developing countries in the areas of technologyexploitation and training.

More than any other form of computer technology, thecommercialization of microcomputer systems has created amass market and a consumer market in computing hardwareand software. The rapid growth of such a large market hasled to a large and diverse low end computing industry, char-acterized by a large set of products, historically low prices,

and rapid market response to consumer demand. The con-sumer surplus generated by this growth has been enormous,and has benefitted microcomputer users in all sectors and inall countries.

Microcomputer software on the whole is significantlybetter in its orientation toward users compared with main-frame software. Because it is meant to be used by peoplewho are generally not computer professionals, much moreeffort has been invested by software producers to ensure that

their software is much easier to learn and easier to use. Theterm user friendlyoriginated with microcomputer software.Prior to microcomputers, most software was used by dataprocessing professionals, and power and functionality weregenerally more valuable than ease of use. With the growthof microcomputer technology and the mass, non-profes-

sional user market, ease of use often became the decidingfactor between success or failure of a software product.Software producers quickly understood this characteristic oftheir marketplace, and much of the software available todayis easy to learn and easy to use by people who are not com-puter professionals. A very important result is that the timenecessary to train people to use microcomputers is lessenedbecause microcomputer based software is easier to use.

Computing in Developing Countries

Computing in developing countries has historicallylagged the application of computing in developed countries,often significantly. Much of the lag can be explained by thefact that, until fairly recently, computers have representedadvanced technology requiring an infrastructure and a bodyof knowledge that was being created with some difficultyeven in the developed countries. The emergence ofmicrocomputer technology signals a fundamental shift in thenature of the technology and the infrastructure required tosupport it, and has the promise of shortening the technologytransfer lag substantially.

It is useful to examine this issue within the context of thefactors affecting the rate of successful absorption ofcomputer technology within developing countries. Impor-tant and relevant factors include: (1) physical infrastructure;(2) human resources infrastructure; (3) size of market; and(4) access to information resources. Such an examination isuseful in spite of the fact that developing countries as agroup are quite heterogeneous, differing widely in the extentto which they have introduced computer technology, and theextent to which the necessary infrastructure exists for ex-ploiting it. As a consequence, for almost any statement onecan make on this subject, there is highly likely to be a groupof countries for which it may be either not meaningful or notapplicable.

Physical infrastructure. The degree to which physicalinfrastructure is important in exploiting computing tech-nology is now decreasing somewhat in large part due to the

introduction of microcomputer technology. Mainframe andminicomputer systems require a reliable supply of cleanelectrical power and reasonably stable environmental condi-tions such as temperature and humidity levels in order to op-erate reliably. The developing countries' dependence in thepast upon these larger systems created major difficulties ininstalling computers and keeping them in operation. Whilelarge computers still require controlled environmentalconditions, with some exceptions microcomputers can oper-ate under a wider range of environmental conditions,


10/60


10

making the elaborate and expensive site preparations of pre-vious years largely unnecessary today.1

Almost all developing countries now have the necessaryphysical infrastructure to support operations of one or moretypes of computer systems, at least in one or more principal

cities. Where electrical power is unreliable, power regula-tors and/or standalone generating capacity may be requiredfor larger computer systems, making use of such systemsmore expensive than it otherwise would be. Power require-ments for microcomputers can sometimes be met under suchconditions by systems with built-in battery power, or withordinary motor vehicle storage batteries providing anelectrical buffer between an unreliable public power orintermittent private power source and the computinginstallation. While such arrangements involve someinconvenience, use of the technology on at least a limitedscale is still viable.

Transportation and communications links to suppliers ofhardware and software resources are also required for effec-tive use of computers. Where communications andtransport links with foreign countries are weak, somecompensation in the form of additional redundancy ofequipment, a larger spare parts and supplies inventory,special communications equipment and additional travel foreducation and obtaining information may be important.Such operations can add to the expense of using computers,but they make possible their use.

The use of microcomputers has alleviated this situationto a considerable extent, since they are generally used ingroups rather than singly. Providing additional equipmentredundancy can be easily accomplished by providing addi-tional systems; these systems can be used productivelywhen their constituent parts are not needed as spares. Theburden of providing such redundancy is highest for smallcountries which are relatively isolated; their need forcapacity is small to begin with, and they must havesufficient backup because of the cost and time delays inservicing their needs from outside the country. Small andremote countries, such as island nations in the Pacific, areprobably most affected by such considerations. However,countries for which transportation, importation orexportation of goods, or communication are difficult forother reasons may be affected equally or more severely.

Lack of adequate physical infrastructure is rarely a majorproblem today, although remedying deficiencies in it willgenerally increase the costs of providing computing capabil-ity to a country. Providing a stable and reliable supply of

1 Microcomputers are surely hardier and more tolerant of environ-mental extremes than larger systems; however, some conditions jeopardizetheir successful operation. Extreme humidity, for example, may causecondensation on circuit boards. A coastal salt water atmosphere may resultin salt deposits on computer components and media, rendering datairretrievable. Static electricity is a hazard to integrated circuit componentsin environments with inadequate humidity.

electrical power is perhaps the most common problem,which often requires substantial additional effort and cost tosolve. As overall economic development occurs, the abilityof developing countries to provide an adequate physicalinfrastructure for computing activities should improve at amoderate rate.

Human resources infrastructure. The skills required toexploit computing technology effectively depend upon min-imal levels of competence within the country in fields suchas engineering and mathematics. The level of literacyamong the most educated people may be an important fac-tor. In some countries, the ability to work in a foreign lan-guage used by suppliers of computing equipment and soft-ware may be an important determinant of success. Aspectsof the cultural milieu within the country may also be impor-tant. Responsiveness to educational opportunity, thestrength of the work ethic, and attitudes and policies towardachievement, employment, production and productivity areall important in the successful transfer of this technology.

In general, a sufficient human resources infrastructureexists in most if not all developing countries to be able tomake use of computer based systems. However, it is oftenthin, and there is substantial competition for the scarce moretalented individuals within both the public and the privatesector as well as between them. Emigration to better labormarkets, a phenomenon sometimes referred to as the braindrain, causes a depletion of the resources necessary toexploit this technology; in countries initially having alimited set of human resources with which to work, such adrain may cause data processing activities to be in a state ofsuspense for periods of time. The lack of priority given tocomputing in many developing countries, along with dif-ficulty in developing personnel policies adequate to attractand hold qualified specialists in the public sector hurts thedevelopment of the human resources infrastructure.

While the allocation of human resources between sectorsand objectives is an internal matter reflecting agovernment's priorities, external assistance can be used toincrease the supply of trained manpower in computer-related skills. If the infrastructure is robust, there will existindividuals who can benefit from specialized trainingabroad. More often, however, the primary need is for morebasic intensive training of larger numbers of people withinthe country. Such training is often necessary to obtain any

degree of effective use of a computer system.

In many countries, the need for such training is not ade-quately met at the present time, in part because it is a laborintensive activity requiring skilled manpower. Given thelimited budgets and multiple objectives of technical assis-tance agencies, both the costs and the opportunity costs ofsuch training are perceived as large. Nevertheless, it is acritical investment for long run effective development ofcomputer based activities that affect not only statistical of-


11/60


11

fices but also many other activities important for nationaldevelopment.

Fortunately advances in both hardware and softwaretechnology have made it easier for developing countries toleverage their existing human resources. Microcomputers

are easier to install, operate and maintain than mainframesor minicomputers. The newer microcomputer applicationpackages are much easier to use than their previous main-frame counterparts and cost less to acquire as well. Thus,persons having limited technical skills can be used to staffmicrocomputer based activities whereas it would have beendifficult if not impossible to have used staff with similarlevels of technical capability in the past.1

Size of market. The current and potential size of themarket for computer systems has been an important deter-minant of the potential profitability of entry into the countryby suppliers. In the past, countries that had limited marketsgenerally received attention from one supplier; somecountries had no suppliers of computing equipment and hadto depend upon sales, support, and service to come fromanother country. Within countries having limited marketsfor computing equipment, the price of conventionalequipment supplied in the past has been quite high.2Smallmarkets have discouraged suppliers from makingdocumentation available in a special local language, whichhas in turn limited the number of people who can locallylearn how to use the supplier's products. Countries withrelatively high rates of growth attracted more than one sup-plier and generally enjoyed the benefits of a more active do-mestic computer industry, including product variety, morecompetitive markets, and greater responsiveness of the mar-ket to demands for specific products and services.

The introduction of microcomputers into developingcountries has modified the situation quite substantially. Mi-crocomputer systems are in general relatively inexpensive,easily shipped, and easy to maintain. The cost of entry intoa microcomputer business in most countries is substantiallylower than for a mainframe or a minicomputer company.Even in the absence of a local supplier, such units can beimported easily and, with some special training, maintainedby lower level staff than would be necessary for larger

1While it is true that maintenance of microcomputer equipment is easierthan maintenance of previous generation equipment, such maintenancedoes require a minimum basic understanding of technology sufficient to

repair basic mechanical and electronic components. Microcomputertechnology is not quite yet at the "plug and play" stage achieved by someconsumer goods and electronics industries.

2Such high prices for conventional equipment have in the past beencited as examples of exploitative behavior, typically by multinationalcompanies in their dealings with developing country economies. Whilesuch exploitative behavior cannot be denied, a sufficient reason for higherthan average prices at that time was the high level of support costs incurredby the supplier in servicing a limited market. There are economies of scalein supporting mainframe computers in compact geographic areas, andmarkets in many countries were too limited to realize such economies.Limited markets also were conducive to the perpetuation of monopolypower, and therefore monopoly pricing, by sole suppliers in a country.

systems.3 Because the cost of microcomputer basedsystems is so low, and because many computing tasks arebetter done using microcomputers, developing countrieswith limited mainframe computing markets now enjoy moresizeable microcomputer markets. Coupled with the low bar-riers to entry faced by suppliers, microcomputers have

proliferated in such environments and have taken oversignificant segments of the computing marketplace.

The most serious constraint upon computer hardwareacquisition have been and continues to be the necessity forand availability of computer hardware maintenance services.At the mainframe and minicomputer level, some developingcountries have sufficient engineering talent to be able todedicate one or more national staff to be trained as mainte-nance engineers, but most do not have such resources. Fur-thermore, training is generally expensive and can result inlocking a country into a particular computer line for sometime. Further specialized training will then be necessarywith every significant shift in computer hardware. At themicrocomputer level, hardware maintenance is considerablyeasier and cheaper, and can be handled in large part bymeans of redundant hardware capacity.

In the less developed countries, the constraint imposedby the necessity for maintenance used to favor large,established manufacturers and the more conventional oftheir product offerings, initially standard batch processingsystems of moderate size and later gaining some limitedinteractive access. The introduction of the microcomputerand the economics of that part of the computing industryshould allow them to leapfrog over some phases of thehistorical development of computers and utilize the moremodern and productive techniques that are becomingcommon practice in developed countries.

Information resources. People living in developed coun-tries live in an information rich environment. A rich varietyof information is available daily from newspapers, maga-zines, books, government publications, technical and schol-arly journals, radio, and television. Information is furtheravailable from friends and colleagues, no matter whetherthey are local or distant. Reliable telephone systemsprovide relatively economical voice communicationbetween most or all parts of the country. Emerging elec-tronic mail systems connect professionals of many kinds to-gether in an effective manner for passing written messages;

3The ability to import microcomputers even in the absence of localsources leads to mixed results, since doing so is likely to retard growth of alocal commercial infrastructure that can supply and servicemicrocomputers. Such an infrastructure is a desirable outcome of theintroduction of microcomputer technology, especially since the probabilityis larger than with mainframe systems that local firms can competesuccessfully in this market. The most desirable outcome of theintroduction of microcomputer technology -- both for statistical officesand within a context of national development -- would be the healthygrowth of an active local industry of microcomputer sales and supportservices, including systems integrators, software specialists, and trainingfacilities and programs.


12/60


12

such systems are growing in number and in size to coverlarger subscriber bases and provide wider geographiccoverage.

The rapid rate of technical progress affecting thecomputing industry, coupled with the growth of personal

computing based on inexpensive microcomputers, hascaused the computing industry to be one of the mostinformation intensive in the world. In developed countries,computing magazines, electronic bulletin boards, computerclubs and user groups, and computing courses and trainingopportunities have all proliferated widely. The market forthese goods and services is thriving in response to a largedemand for information which makes it possible to applythe the technology to a wide variety of disciplines and appli-cations.

Developing countries are limited in varying degrees intheir ability to access this rich flow of information abouttechnical issues, products, and reports of experiences. De-livery of printed material is delayed because of distance; in-ternational air mail rates are high. Foreign exchange may belimited and not available for such material. Local postal de-livery systems may be unreliable, so that printed matter maynot reach its intended recipient. Information in the nationallanguage may be limited. Further, the ability to comprehendsuch material and fit it into a specific environment may beimpaired by a lack of expertise in the country; reasonablequestions that could be answered with a little effort indeveloped countries may go unanswered in developingenvironments, blocking further exploitation of a technique,product, or idea.

Even with the problems within developing countries ofobtaining access to rich sources of information about com-puter technology, the current situation is still considerablybetter than it was when countries had one or just a few com-puter installations and probably only one supplier ofequipment. In the past, knowledge about the computing in-dustry and computing techniques most often came directlyfrom the local supplier of computers; it was in the supplier'sinterest to promote knowledge associated with that sup-plier's products and approach to computing. At that timeonly a limited number of people were qualified to work withcomputers, and foreign training, which was another sourceof knowledge, was expensive and limited to a few individu-als. Access to knowledge is surely better now, with multi-

ple and relatively inexpensive sources of information. Thisbenefit is derived directly from the emergence of the massmarket for inexpensive microcomputer systems.

Information poverty is still one of the more importantand insidious obstacles in developing countries to effectiveexploitation of information processing and other types oftechnology. The lack of adequate information regarding de-velopments in other countries and other environments is of-ten not noticed; in the absence of new information, old tech-

niques and procedures are continued without consciousknowledge of alternatives.1 And, while developingcountries may not be hurt in an absolute sense by lack ofinformation, they are certainly negatively affected by anyrelative measure.

The issue underlying information poverty is how toovercome it to provide a sufficient flow of information intodeveloping country environments so that individuals andgovernments can make good choices from among realisticalternatives. There are a number of approaches that couldassist countries in obtaining a richer and more up-to-dateflow of relevant information for statistical data processing.They include: (1) stronger partnerships with national univer-sities and similar organizations; (2) greater exploitation ofboth short and long term visits by experts from othercountries; (3) effective sharing and exploitation of locallyavailable technical skills and knowledge; (4) establishmentof an informal statistical data processing newsletter; and (5)recognition of the importance of release time for continuedon-the-job learning.

Microcomputers and developing countries

From its inception, microcomputer technology has beenappropriate technology for developing countries. Their size,cost, modularity, ease of transport and environmentalrequirements, as well as the variety and power of user-ori-ented software, make them appropriate for the smaller andsimpler environments found in many developing countries.

Support requirements for the mainframe and minicom-puter systems that were previously installed in developing

countries were significant, and computer suppliers wouldgenerally offer products that were somewhat older and morewell understood than the leading edge products being intro-duced in developed countries at the same time. Such a strat-egy worked well in part also for the developing country; theequipment and software supplied was well understood andmore stable than when it was first introduced. Supplierscould afford to offer and support such systems far from theirgeographic areas of technical expertise because the systemswere well understood and trouble free relative to newermodels.

With the rapid development of the microcomputer indus-try, the minimum practical lag between the introduction of

current technology in developed and developing countrieshas been compressed to a matter of months instead of

1 Readers who wear eyeglasses and whose vision is changing may feelsome sympathy with this condition. Deterioration of corrected vision isgenerally sufficiently slow that it is not noticed untilnew eyeglasses withimproved correction are first worn. Often the effect of first using such neweyeglasses is startling; objects and details snap into sharp focus and thewearer is amazed that such improvement is possible. With respect toinformation, the effect is similar; it is often difficult to understand thelosses due to not having relevant information until you have it, and onlythen is it possible to assess the opportunity cost of not having had it.


13/60


13

years.1 This lag no longer depends very much upon theability of the original manufacturer to support its products ina specific geographic region, since support can now often bedelivered either within the country or with the assistance ofa third party, perhaps an international aid agency, eitherlocally or remotely. Instead, the lag depends more upon the

ability of trained personnel in the country who are able toexploit the microcomputer systems and train others to do so

1This argument is of course valid only if the funds exist to acquire theequipment in the first place. Assuming however that funds areavailable,the range of appropriate information technology products that can now besupported within developing countries is relatively larger than it has beenin the past because the newer technology requires less dependence upon theoriginal supplier of the equipment.

also. This function can initially be provided by foreign ex-pertise and later, as the results of training accumulate, by lo-cal expertise. For the first time, emerging computingtechnology can be placed at the disposal of developingcountries almost as soon as it is readily available and under-stood in the developed countries.2

2Progress in the computer base for statistical data processing depends

upon advances in hardware and software, the rate of technical innovation,and the rapidity with which these advances are translated into marketableproducts at affordable prices. The rate at which the technology isexploited within any given environment is another matter, and reflects thestatus of the local infrastructure, especially the human resourcescomponent, and other conditions between countries.


14/60

14

3. STRATEGIC ISSUES IN THE USE OF MICROCOMPUTERS

Overview

Microcomputer technology is appropriate for many areasof statistical data processing, including census data pro-

cessing. As the capacities and speeds of microcomputersystems increase, an increasing proportion of statistical dataprocessing activities become tractable by applying this newtechnology. In addition, the personal interactive nature ofthe microcomputer and available software for it make it inmany respects a qualitatively better computer platform onwhich to base such processing activities. Microcomputerscan be used effectively for data collection, editing, tabula-tion, analysis, publication, control, and management. Thistechnology provides the potential for a fundamentalbreakthrough in the effectiveness of many statistical dataprocessing operations.

Nevertheless, microcomputers are not a panacea for data

processing problems, just as the older mainframes did notsolve all problems. Microcomputer technology brings withit a new set of responsibilities which are not immediatelyobvious to an initial user. It is necessary to understand theseresponsibilities as well as the new opportunities offeredby exploiting the technology and develop an ap-propriate strategy for dealing with them. This section dis-cusses these issues within the context of statistical data pro-cessing and operations in general, and is directly applicableto census data processing activities.

Responsibilities of Microcomputer Operations

The introduction of microcomputer technology in anorganization implies a significant transfer of responsibilityfrom a central computer support staff to individual users ofthe microcomputer systems. The transfer of responsibilityhas both benefits and costs for the user and for the organiza-tion.

In organizations that use mainframe and minicomputersystems, both the suppliers of the equipment and the centralcomputer support staffs provide a variety of functions thatmay be partially or totally invisible to the users of the sys-tem. These functions include:

Performing a requirements analysis for either thespecific dedicated activities or for the organization as awhole and specifying the characteristics of the system tobe acquired;1

1This task is considerably simpler for the microcomputer user. Not

only are the tasks simpler and more homogeneous, but given the relativelylow level of expenditure, often several different small computers can bepurchased. Such diversity of hardware or software may be more effectivein meeting institutional needs than earlier purchases of single systems.

System specification and negotiation with suppliers;evaluation of responses to the request for proposals tosupply hardware and software components;

Site preparation, including designing and constructing oradapting the physical site for the computer system, sup-plying electrical power and possibly backup power asrequired, and providing air conditioning and other envi-ronmental requirements;

Installation and testingof the system, including co-ordi-nating and assisting in delivery, installation, and accep-tance testing;

Providing system programmingskills and services, gen-erating the operating system and updating it periodically,and installing application program packages;

Providing system resource management, including man-agement of peak loads, priority scheduling, and alloca-tions for scarce resources such as disk storage space;

Providing education and training, including userconsultation and education of various kinds, and possiblya more turnkey operation including system analysis,design, programming, testing, implementation and doc-umentation;

Providing operating services , including machine opera-tion, tape and disk volume storage and control, filebackup, and an inventory of expendable supplies;2

Providing hardware maintenance services, includingfault diagnosis (either internally or through a externalsupplier);

Providing applications implementation and management,including choosing and implementing specific choices ofprograms in the computing system, as well asestablishing operational controls over their use andmanaging staff use of those programs; and

Assisting in long term capacity planning and configura-tion management; monitoring the market for additionalor replacement hardware and software to maintain an ap-propriate and cost effective data processing operation.

Users of microcomputers are faced with the same set ofresponsibilities, although at a much lower level, and oftenwith substantially less risk to themselves or to their orga-

2The importance of device and file backup, routinely performedcentrally at most mainframe installations, is routinely ignored by mostmicrocomputer operators, and at their own peril. Since the data processedon such machines is institutional, distributed data base management mustensure that microcomputer users are not only conscious of the possibility ofdata loss through failure to backup, but also perform backups on schedule.


15/60

STRATEGIC ISSUES IN THE USE OF MICROCOMPUTERS

15

nization if non-optimal decisions are made. Nevertheless,these responsibilities have been by and large transferred tousers albeit on a considerably smaller scale, and users mustbe aware of them if decentralized microcomputer use is tobe viable within an organization. This issue becomes im-mediately visible when malfunctions occur, and it is up to

the user to diagnose the problem and solve it. Organizationswith non-programmer users may want to retain a centraldata processing services staff for dealing with more difficultproblems that may arise, while leaving the operational re-sponsibility for substantively oriented work with the user.

The following sections deal with each of the issuesraised above.

Requirements Analysis

Most classical computing centers were created and oper-ated to serve a spectrum of needs. Such centers werecreated either for the use of one organization or for severalorganizations. In either case, the problem of understandingthe combined requirements of all users and uses of thecenter was among the first to be dealt with.

The first step in analyzing the requirements that will beplaced upon a computer system is to examine existing activ-ities that are likely to benefit from being implemented onthe new system. Such applications may already be runningon another computer system, or they may be activities thatare either performed manually or not at all.

Analyzing requirements that will be placed upon a com-puter system has a hidden dimension that is difficult for thenon-computer professional to appreciate fully. That dimen-sion is the analysis of latent demand, i.e., demand notforeseen at the time that the requirements for the computerare assembled. Experience indicates that over the life cycleof a computer system, which is in the range of 5-10 years,the requirements not foreseen at the beginning of the cyclebecome increasingly numerous and can come to dominatethe total activity for which the computer is used. Suchgrowth in unforeseen applications can eventually force achange or upgrade of the computer system because of theimportance of the originally unseen capabilities of the sys-tem.

Such a process represents a healthy learning response to

using a new and more powerful tool. It is important, how-ever, not to pay too high a cost for the experience. Duringthe process of analyzing requirements which the computerwill be expected to fulfill, it is therefore advisable to obtainprofessional assistance in deciding which activities, existingor not, might benefit from the presence of the computer. Tothe extent that latent applications can be identified as earlyas possible, the computer system will be more responsive tothe users' requirements over a longer period of time, and

will provide a more stable and responsive computingenvironment.

The analysis of requirements for procurement ofmicrocomputer systems is somewhat easier, since such sys-tems are often selected for specific tasks. It is often con-

venient to provide a number of microcomputers to perform anumber of different tasks. In this case, the hardware andsoftware configurations of the microcomputers may be spe-cialized to adapt them suitably for the different tasks. As la-tent demand is realized, additional systems can be acquiredin the short run without incurring a significant cost penalty.Nevertheless, from the point of view of setting a directionfor the acquisition of microcomputers and the way they areused as a whole to attack the problems of an organization, arelatively complete and professional requirements analysisis likely to lead to an acquisitions strategy, possibly phasedover time, that satisfies the needs of the organization well.

For processing a census of population and housing, sucha requirements analysis generally begins with the softwarethat is to be used for the various phases of data processing,which in turn is determined by the amount of processing tobe performed and the size and variety of the outputsrequired. Considerations of data volume generally dictatethe number and size of computer systems that will berequired to perform the work in a targeted period of time.Excess hardware capacity will be determined in part by thecountry's ability to repair computers locally and in part bythe amount of trouble anticipated in effecting replacementparts and maintenance services from sources outside thecountry.

System Specification and Negotiation withSuppliers

Regardless of the scale of computing equipment used,requirements identified during the above analysis must beturned into a specification of what type and amount of com-puting equipment is necessary to accomplish the job. Thespecification must include hardware which has the speed,capacity, and devices to support the required activities, andsoftware which has the functionality and capacity to per-form the applications required. The choices of specifichardware and software are interdependent, making theselection process sometimes difficult.

When there are two or more possible suppliers and theprocurement is of non-negligible value, public sector orga-nizations are generally required to go through a process ofqualifying suppliers and obtaining competitive bids for thesystem. Often such a procedure is based upon an officialdocument called a request for proposals (RFP) whichreflects the results of the requirements analysis as well asservice, training, and related considerations deemednecessary as a part of the procurement. The content of theRFP is critical to assuring that the proposal ultimately


16/60


16

accepted really meets the requirements well. Dependingupon the legal requirements that pertain to specificprocurements of equipment, an incomplete or inaccurateRFP documents can lead to having to accept a system that issub-optimal or inappropriate for the tasks at hand.

Microcomputer procurements are generally simpler, al-though the same cautions are in order. The requirements ofthe activity to be performed must be translated intohardware capacity and software functionality. Once it isdetermined which set of software packages would best beused to process the census data, hardware requirements canbe specified. With computer hardware, there is generally atradeoff between buying from a recognized industry leaderand an off-brand imitation at lower price. If the latteralternative appears attractive, then either the bid documentor the contract with the supplier must be written in a way toprotect the user from incompatibilities that interfere withuse of the hardware and software together. Off-brandsystems may require off-brand spare parts; if so, this mustbe foreseen at the time of procurement. Later expansion ofthe system, including adaptation to new hardware productsand operating systems,1 may be useful and the bid orcontract document should include the purchaser'sexpectations in this regard.

Most developing countries using microcomputers to pro-cess their population censuses will use hardware that runsthe MS-DOS operating system. With the currently availableMS-DOS based software, there are several strategies thatcan be used. Although hardware requirements will tend tobe somewhat similar, the approaches used by differentediting and tabulation programs may require more or lesshardware to do a specific job.2 The total hardware specifiedshould reflect the requirements of the particular softwarestrategy chosen for the data processing.

Site Preparation

Site preparation for mainframe and minicomputers is of-ten a complex and technical task. Mainframes require aphysical environment in which temperature and humidityare controlled within moderately narrow limits. Any sub-stantial amount of dust in the air may interfere with physicalstorage media such as disk or magnetic tape. The stability

1For example, currently most systems in developing countries are IBMPCs or PC clones running the PC-DOS or MS-DOS operating system. As

programs become larger and users' needs evolve, there will be migrationfrom MS-DOS to the OS/2 operating system. This migration will beslowly forced upon users who want to uti lize new programs and newversions of programs, which to an increasing extent will not fit into theprogram address space provided by MS-DOS. The upgrade path for PC-compatible equipment may therefore be important, depending upon thetasks to be performed and the expected lifetime of the applications and theequipment.

2The fact that a specific set of programs requires less hardware than analternative strategy does not necessarily mean that it should be chosen. Theadditional hardware required may reflect more efficient operations, or itmay allow additional flexibility and options in the production process andin its outputs.

and quality of electrical power supplied to the computermust meet the requirements of the computer system. Theremay be a sufficient number of cables interconnecting thevarious cabinets comprising the computer system that araised floor may be useful or necessary for safety. Whileminicomputers are generally less demanding of the environ-

ment in which they operate, preparation of a site for aminicomputer is non-trivial.

Large computer suppliers have been aware of thedifficulties faced by individual users in preparing a site fortheir computer systems, and provide significant educationand site planning and preparation assistance to their cus-tomers. They perform this function both for their customersand in their own self-interest; their systems will not performeffectively if the site is inadequately prepared.

Microcomputers are more insensitive to variations intheir environment than are larger computer systems. Theycan tolerate wider swings in temperature and humidity; ingeneral, they will work well in any environment in whichtheir human operator is moderately comfortable. On theother hand, microcomputer diskettes are more fragile, andcan be harmed by dust, sand, salt water condensation, andother particulate matter carried in the air.

Microcomputers require a stable electrical power supply,but the amount of power required by each individual unit isrelatively low. A typical microcomputer consumes as muchpower as a bright incandescent electric bulb. Power outagesand serious power fluctuations can cause loss of data anddamage to the computer equipment, so that users must en-sure an electrically clean and steady power source for theirequipment. Different types and brands of microcomputersvary in their ability to tolerate electrical disruptions, and it isthe purchaser's and user's responsibility to ensure that thequality of power supplied matches what the microcomputersystem requires so that neither data nor equipment will becompromised or damaged.

Power analysis equipment is available for rent or for salethat will monitor existing electrical power precisely and willprovide the data for an analysis of current power quality inany given location. The results of this analysis can be usedto determine what types of power conditioning equipment orstandby generating capacity, if any, will be required to pro-vide acceptable power for the computer systems.3 This ap-

3The general subject of electric power conditioning is too broad to bediscussed in detail here. There is a range of possible power problems thatcould either disrupt computer operations or damage data or equipment.Voltage stabilizersare used to make constant the voltage of a reliablesource of power that varies in voltage. Surge protectorsare used to clipshort surges or drops in power often caused by other electrical devices onthe power line or by external events such as stormy weather.Uninterruptible power suppliesare often used to bridge short periods ofpower outage, using a bank of storage batteries that constantly buffer spareelectrical capacity. Independent electric generators, often attached todiesel motors, provide continuous power subject to the availability of petrolto run the diesel or other engine that powers the generator. Dependingupon the electric power environment that the computers needs to operate,


17/60


17

proach is generally effective, but has one major weakness; itis based upon the assumption that future power quality willmirror past quality. If in the future there is a significantdegradation of power quality, then any power conditioningmethods used may not be sufficient to cope with the de-graded input quality.1

Fortunately it is easier and probably less expensive fordeveloping countries to prepare a site for installing micro-computer systems for census processing than it was for theirmainframe predecessor. While physical hazards to the datamedia (sand, salt, dirt) are still present, the temperature con-straints and to some extent humidity constraints will be eas-ier to meet. Local area networks may provide a way tominimize the detrimental effects of dirty environments byminimizing the transport of diskette media. Unstablepower, unfortunately, will continue to be a problem whichstatistical agencies need to address.

Installation and Testing

Mainframe systems are almost always installed by thesystem supplier. Such an installation is technically rela-tively demanding, and the cost of making a mistake is po-tentially high. Minicomputers are less difficult to install,but the task is best done by technical specialists. Accep-tance testing for such systems is generally done jointly bythe customer and the supplier, and usually includes verifica-tion that all installed hardware and software operatefunctionally as promised, as well as a more lengthy test tomeet a minimum downtime requirement.

Microcomputer installation is almost always offered asan additional service by the supplier, and the price of suchservice is not included in the price of the computer. Manymicrocomputers are relatively easy to set up and test, and re-quire only a moderate amount of knowledge about the sys-tem. However, some installations may include installingspecial electronic circuit cards or other special deviceswhich may pose some difficulties.2 If the microcomputerone or more of these techniques, as well as others, may be required toprovide power that is sufficiently clean to operate the computer systems.

1Unfortunately, the cost of bad electrical power compounds itself in thefollowing manner. In addition to obtaining whatever power conditioningequipment is required to condition the electrical power appropriately, theorganization must also be prepared to support the power conditioningequipment itself should it malfunction. Such support can be provided inthe same manner as supporting microcomputers, through maintaining asmall inventory of spare units on site. For this reason, it may be be betterto condition power to microcomputers using a number of transformers,regulators, and micro-UPS units rather than relying upon a powerconditioning system based upon non-redundant large units. Powerconditioning equipment is notoriously heavy, so that significant transportflexibility can be had by using multiple units. While the cost of multiplesmaller units may be somewhat greater, the higher degree of certainty ofcontinuous operation provided by having several identical units is almostalways worth the additional cost if the application is at all time critical.

2The manner in which a microcomputer is packaged can affect ease ofinstallation. Some types of microcomputers are relatively easy to openphysically so that additional components can be added and other changesmade. Some microcomputers, on the other hand, are designed to remainclosed and are consequently more difficult to modify. Indeed, such

system is made up of parts from different manufacturers,then the possibility of some level of incompatibility maycomplicate the installation and testing. Installation andtesting has a software dimension, which is discussed below.

If a microcomputer system is procured at a remote or for-

eign location, it may be wise to specify testing at the pointof purchase as a part of the procurement procedure. Suchtesting serves to identify initial problems before theequipment has been shipped and when it can easily be re-turned to the supplier for a replacement at little cost andwith minimum delay.3

Statistical agencies should assure themselves that theprocurement process for microcomputer systems and systemsoftware includes an acceptable strategy for installation andtesting of the systems. Such a period is also useful for edu-cating staff who will supervise the computer operation whenit is commissioned. These staff members should ensure thatthey understand exactly how the systems are working andwhy they are working correctly before they formally acceptthe systems from the organization responsible for their pro-curement.

System Programming

System programming refers to the process of understand-ing and modifying the instructions comprising the operatingsystem under which the computer operates. The operatingsystem is a software program, generally supplied by orthrough the manufacturer of the computer hardware, thatgives the user effective control over system resources.Classical operating systems generally include commands ofthe type "print, "copy, "create file, etc. Operating sys-tems support execution of programming languages, applica-tions programs, and all other software. They are anessential part of a computer configuration.

Mainframe computer centers and, to a somewhat lesserextent, minicomputer systems have one or more technicalstaff members who understand the operating system struc-ture and modify and update it to meet the special needs ofthe software and users of that computer. They also installnew software and ensure that it works well with theoperating system of the computer. This is a highly spe-cialized skill; operating systems for large computers arevery complex and contain millions of instructions.

modifications made directly by the user may void the manufacturer'swarranty. However such voiding of the warranty may be unimportantwhen the maintenance strategy for the equipment relies more upon localresources than upon return to a supplier or a supplier's representative at aremote location.

3A strategy of assembling and testing each microcomputer suppliedthrough its technical co-operation backstopping activities was adopted bythe United Nations Statistical Office during 1979-1984. Such a pretestingpolicy remains valuable in cases where the system configuration issufficiently new or different and may not be understood or work properly atits final destination.


18/60


18

Some of the tasks necessary for installing and maintain-ing a microcomputer environment can be considered in therealm of system programming. It is necessary to: (1) installthe operating system on a fixed disk or diskettes; (2) makechoices regarding how the system is to be configured; (3)configure a startup sequence for the computer; (4) install

software links to special pieces of equipment that may beattached to the microcomputer;1(5) load and manage mem-ory resident software, ensuring no memory conflicts withapplication programs; (6) configure application software tomatch specific characteristics of the microcomputer,2 espe-cially the video and printer interfaces; and (7) update the op-erating system periodically as new releases are issued, en-suring that the new release works consistently within thecontext of the changes made and the application programsbeing run.

In a large computer operation, the system programmerserves as the local expert on software issues. When a com-puter procedure consistently does not work, very often thecause is software based. Typically the system programmeris best situated technically to analyze the problem, deter-mine why it does not work, and either change the procedureor create a work-around solution. Many such errors can beremedied by training users, but some require substantialanalysis. In a microcomputer environment, the same kind ofproblems will occur, and it will be important to have at leastone technically trained staff person who can understand theinteraction of the system and application software as well asthe structure of the hardware well enough to be able to assistusers. Failure to have such skills available reduces users toguesswork, inefficient and limited use of the computer, andpotential loss of information.

Most system software and applications software productsare being revised or updated on a continuous basis by theirmanufacturers. Periodically, a specific versionor releaseofsuch a product is made available to individuals andorganizations that have previously purchased the product.New releases of software products are useful; they containnew features and functions that users have been asking for,as well as eliminating problems, sometimes referred to asbugs, that were detected in previous versions.

However, installing new versions of software may causenew problems to appear, possibly because ofincompatibilities with other software products on the same

system. Such modules should therefore not be installedwhile a large application (such as a population census) is

1A concrete example is provided by the addition of device driverswithin an MS-DOS environment and placing appropriate commands withintheCONFIG.SYS system file.

2Even the installation of some advanced applications in simplemicrocomputer environments may cause some difficulty. Often the mosteffective way of obtaining prompt and expert assistance is to contact thesupplier's technical staff, which is likely to be in another continent. Havingreliable communications links such as telephone, telex, and other electronicmail linkages with other countries is very useful on such occasions.

being processed without extensive testing to ensure thatsubstitution of the new module for the old has no adverseside effects. If the new functionality offered by the newrelease is not required and there have been no previousproblems with the software, it may be better to freeze thehardware-software configuration for a time and defer the

installation of the new release until after the entire largeactivity has been completed.

Census agencies should assign this function to one ormore staff members at the beginning of the process ofprocuring the equipment. These staff members should beresponsible for ensuring that they learn as much as possibleabout the aspects of system soft

the use of microcomputers for census data processing (onu, 1989)

Documents