Introduction

This first background paper surveys the principal classes of data used by modern economic historians, the history and mechanisms of data collection and the reliability of the end product through time.

References cited in this and other background papers are collected together as the final section of background paper 4.
 

The collection of economic statistics in Britain

Introduction

One of the distinguishing characteristics of economic history as it developed from the late nineteenth century (see Coleman 1987) was the use of statistics. The body of data now available, especially for the macroeconomic historian, is huge. Fortunately, much of it has been brought together in a series of collections:

• Feinstein (1972) - the main source for data on the British economy since the mid-nineteenth century.
• Flora et al. (1983; 1987) - two enormous volumes of data on Europe since the early nineteenth century.
• Mitchell (1988) - a very wide ranging collection of British historical statistics.
• Mitchell (1992; 1993) - general collections, the former on Europe and the latter on the Americas.

For those wanting to work on the postwar British economy, the Central Statistical Office (CSO) publishes a number of monthly (Economic Trends, Monthly Digest of Statistics) and annual volumes (Economic Trends, Annual Supplement, United Kingdom Balance of Payments, known as the pink book and United Kingdom National Accounts, the blue book).

The four main classes of macroeconomic data are:

• National income and its components
• Industrial output tables (which cover the three productive sectors: primary, secondary and tertiary).
• Balance of payments accounts (particularly important since historically Britain has been one of the world’s most open economies and an international financial centre).
• Labour market indicators.

The main sources of economic statistics

Economic statistics can be divided into three principal groups:

• Those collected by the government.
• Those arising from the working of business and other economic organizations.
• Those collected by private individuals.

Statistics collected and published by central and local government are of two kinds:

• Those which are obtained from specifically statistical enquiries.
• Those which are collected as part of the working of some administrative control or system of taxation.

Although this distinction can usually be made without difficulty when one enquires about how the statistics started, the two may overlap in later development. Thus statistical series which started merely as an accidental offshoot of some administrative control may later be developed and extended for the knowledge and information they give about some aspect of economic affairs.

Historically, most of our economic statistics are a by-product of administration or law. Statistics of foreign trade originally arose from the control of imports and exports; statistics of income from the operations of the Board of Inland Revenue in assessing income for tax purposes; statistics of the consumption of a wide range of commodities, for example tea, beer and tobacco, from the operations of the Customs and Excise; statistics of employment and unemployment from the schemes of compulsory insurance against unemployment.

Administrative and legal control will not be maintained or continued merely because it yields useful statistics, and changes in administration and law will usually be made without regard to the convenience of those who use the statistics. This means that there are frequently sudden changes in the continuity and coverage of statistics derived from such sources. Thus the abolition of food rationing at the end of the two world wars makes it more difficult to estimate changes in food consumption; and the abolition of exchange controls in 1979 has made it much more difficult to calculate figures of the balance of payments.

Statistics collected directly as part of a statistical enquiry mainly for their informative value are in a different category. For here the questionnaire used, the coverage of the enquiry, and the tabulation of the results can all be designed mainly from the point of view of the usefulness of the information that will be obtained. The first major statistical enquiry of this kind was the Census of Population taken in 1801. Since then a population census has been taken every ten years, except in 1941. The first Census of Production was taken in 1907 and has been conducted annually since 1948. These censuses now form the major statistical enquiries undertaken by government. There are many other smaller enquiries, such as the employment and the family budget enquiries which began under the Ministry of Labour in the interwar period and have developed greatly since, and the monthly statistics of production collected from industry. Some of the statistics first obtained as an offshoot of administration have now been extended in range and detail in such a way that they ought perhaps to be considered as statistical enquiries in their own right. Examples are the collection of foreign trade statistics which was begun by the Board of Trade and the collection of employment and unemployment statistics initiated by the Ministry of Labour between the wars.

Most economic organizations, public and private, need to keep statistical records if they are to run efficiently, and some of these records form a most important source of economic information. For example, most of our information about the movement of wholesale prices comes, from the records of transactions kept by internationally organized produce and raw material exchanges. The records of transactions on the stock exchange, the published accounts of the banks and other public companies, provide valuable information about the operation of a large section of business. Most figures of wage rates and hours of work are based on the collective agreements made between trade unions and employers’ associations or on decisions given by statutory wage-fixing authorities. In many industries, private trade organizations collect statistical information of great economic interest. Much of this is published or passed on to interested government departments for inclusion in official publications. For example, before nationalization, nearly all the detailed statistics we have about iron and steel production and employment were collected by the Iron and Steel Federation from its members.

General problems in using economic statistics

The main problems that arise in using economic statistics are:

1. To find out exactly what the statistics mean and cover.
2. If a series is being used over a period of time, to make sure that the figures are comparable, or if they are not comparable, to attempt to ensure compatibility by allowing for changes in definition and coverage.
3. To assess the degree of accuracy and reliability in the figures published.
4. To adapt the statistical information to the particular purposes for which one wants to use it, for statistics are rarely published in the form which is exactly appropriate to the use one has in mind.
5. Where there is no published statistical information to try to make estimates on the basis of the data which are available.

There are certain general issues which recur in many different fields of economic statistics. Statistics arising from the administration of some scheme of control or from the operation of the law, inevitably reflect the peculiar and changing features of such administration and law. They cannot, therefore, be used properly without knowledge of the administration and legal system from which they arise.

Thus for most of the twentieth century the figures of income distribution in this country were not collected in a special statistical enquiry. Rather, they arose from the operations of the Board of Inland Revenue in assessing individuals’ incomes for the payment of income-tax and surtax, and therefore reflected the legal and administrative rules according to which such assessments were made. One cannot use the published figures of income distribution sensibly without knowing something about these rules and the changes in them over the period for which one wishes to use the figures. The situation was greatly improved in the 1970s by the establishment of the Royal Commission on Income and Wealth Distribution (Diamond Commission), but this was abolished in 1979, no doubt because the first Thatcher government correctly anticipated the likely distributional consequences of its policies.

A similar situation prevails with the monthly figures of unemployment, first published by the Ministry of Labour. These do not result from a special statistical enquiry into unemployment, but record the number of persons who register as unemployed at the labour exchanges. This registration is much influenced by the coverage of state unemployment insurance and the conditions governing the payment of unemployment benefits. These have varied a great deal since the first state unemployment insurance scheme was introduced before the First World War. Without considerable knowledge of these changes the published figures of unemployment may easily be misinterpreted.

It is not easy to discover how the detailed peculiarities of administration and law affect the meaning of statistics of this kind. Detailed, painful research is often necessary, for the authorities issuing the statistics are themselves not fully aware of the administrative practices which may affect the figures. It is rare for figures of this kind to correspond in definition and coverage for what is wanted in economic and statistical enquiry. And one of the problems in using these figures is to make adjustments to them so that they are more appropriate to the purpose in hand. Thus, until recently, the figures of income distribution published by the Board of Inland Revenue reports cover only incomes assessable to tax, and, therefore, exclude incomes below the  tax-exemption limit and that part of income, such as interest on savings certificates, which is by law exempt from taxation. But for most purposes of economic analysis one wants, if possible, to include these categories of tax-exempted incomes, and one would try therefore to adjust the official figures, difficult though this may be.

Where the statistics are the result of an enquiry specifically undertaken to obtain the information, it is easier to find out what exactly the figures cover, and they are more likely to correspond to what is required for economic analysis. The official reports in which the results of such enquiries are published usually include a full explanation of the exact nature of the enquiry and questions asked, how it was conducted, and how the results were analysed and tabulated. It is always advisable to study the explanatory introduction and notes before using the statistics. Useful hints about the meaning and reliability of the figures can often be got from examining the questionnaire which was used in the enquiry. This also gives some clue to the further material which may be available in the Public Record Office but is not published.

Some of the grossest errors in the use of economic statistics arise from ignoring the most elementary problems of comparability. A regular pitfall arises from ignorance whether the figures being used cover the British Isles, the United Kingdom, Great Britain or only England and Wales. The variation in territorial coverage is the most irritating feature of British official statistics. Even greater confusion is possible in dealing with regional statistics, for although there is now a standard set of regions, these have only been introduced since the Second World War and they are not used in all official statistics.

When statistics are improved in scope and definition one usually has to pay a price in terms of comparability with the past, for it is rarely possible to go back and revise earlier figures on the improved basis. Thus the introduction in 1948 of the standard industrial classification for most government statistics was a great step forward. But the improvement itself to some extent destroyed comparability with statistics before 1948. There is also the problem that periodically statistics are rebased on a new constant price basis, which destroys comparability over long periods.

It is never easy to assess the exact degree of accuracy and reliability in any set of published statistical data, and official publications give little help in such an assessment. There are three main kinds of error.

1. Errors occur in the original statistical material, either because people who should do not provide returns, or because some provide figures which are inaccurate.
2. Errors arise in statistics which are not merely summaries or tabulations of original data, but attempt to make secondary estimates on the basis of incomplete statistical material. Typical statistics of this kind are the figures of the national income and expenditure or the balance of payments.
3. There is the extent of error in using statistics as if they measure something slightly different from what they actually measure. A typical example is the error involved in using figures of the average value of imports, based on customs declarations, as a measure of changes in the price of goods imported into this country.

The error in the original statistical material can usually only be assessed, and then often with difficulty, from detailed knowledge of the source of the statistics and the way they were analysed and tabulated. One often gets a useful clue about the reliability of figures by looking at the original form or questionnaire and asking whether the questions are so framed that reliable answers can and will be given.

Thus the Census of Population volumes give figures of the non-employed population in this country in three categories: those out of work, the unoccupied who are retired, and other unoccupied people. These figures are based on the answers given about each person’s occupation in the census form. The notes to this column on the census form state ‘if occupied for payment or profit, state precise occupation or calling. If out of work or wholly retired, state usual former occupation, and add "out of work" or "retired".’ How would this question be treated by a person aged 60 who has lost their job and is not sure whether to try to live on their pension or to look for another job? Would they call themselves ‘retired’ or ‘out of work’? There is no clear dividing line in such cases between ‘out of work’ and ‘retired’, and the way in which individuals in similar circumstances may describe themselves may be quite different. If one examines the original questionnaire in this way one must come to the conclusion that for certain categories and age groups the figures of those ‘out of work’, ‘retired’, or ‘otherwise unoccupied’ must be treated with great caution.

Where statistics arise from law or administration, their accuracy and reliability will in part be a reflection of the efficiency of the administration. If, for example, the statistics arise from the operation of taxation, one can assume that the taxation authorities have the power to investigate the accuracy of the information on the basis of which such taxation is assessed. The accuracy of the statistics will therefore reflect the honesty of those subject to the taxes and the efficiency and thoroughness with which the authority exercises its power of investigation (Those interested might consult Smithies’ (1984) study of the black economy in Britain).

If there is an ad valorem duty on imports, the Customs and Excise can require information to satisfy themselves that the importer is not trying to evade payment of tax by undervaluing the goods, and can, if necessary, insist on its own assessment for tax purposes. It is this valuation for customs duty which is used in the trade statistics published by the Board of Trade. If the customs authority is lax in checking import valuations, then importers may be tempted to undervalue goods in their customs declarations, and in countries where this happens figures of import trade would be correspondingly unreliable. In the same way statistics of income derived from taxation authorities reflect in part the efficiency of the taxation authorities in assessing incomes for tax purposes and the honesty of the citizens in declaring their full incomes as required by law.

Where the published statistics depend on estimation, one ought to try to discover the element of estimation in the figures and the exact way in which the estimate was made. This is often extremely difficult because British official statistical publications give little information about how estimates are arrived at. The annual National Income Blue Book is full of estimates, but there is little indication of the methods used in arriving at them. At present one must proceed from general knowledge of the available material to work out for oneself how the official statisticians probably made their estimates and then attempt to assess how reliable they are.

Assessment of error and accuracy cannot be expressed in precise numerical terms, since it is usually a matter of personal judgement. The assessment of a margin of error in statistical terms would imply much more accurate knowledge of the error in the figures than is usually available to the statistician. Indeed, the kind of information which would usually enable one to quote precise figures of margins of error would often itself enable one to reduce the extent of error in the figures. Thus the kind of knowledge which would enable one to state with precision the margin of error due to evasion in the Board of Inland Revenue’s estimates of income, would itself enable the Board to reduce the amount of evasion.

There is often a tendency to exaggerate the extent of reliability in published statistics. For when figures are well presented in neatly printed tables, the reader is usually inclined to assume that they are accurate, especially if the publication is an official one. And official publications rarely give sufficient information to enable the user to distinguish easily between the more and less reliable figures, unless they are themselves an expert in the subject. It is true that in some publications one can gather from the notes and explanations which accompany the tables that the figures are no more than ‘reasoned guess-work’ or are subject to ‘a wide margin of error’. But statistics and the notes which explain and qualify them are easily parted, and the same figure may appear in another publication without any mention of the cautionary warnings about error (one of the classic errors identified by Platt 1989 in his diatribe against quantitative economic historians). For example, there are voluminous notes in the annual Balance of Payments Pink Book explaining and defining the terms used and in some cases indicating the uncertainty in the figures. But most of these qualifications and notes have disappeared, perhaps inevitably, when the figures are given in the summary tables on the balance of payments in the Annual Abstract of Statistics. The unwary reader of the Abstract might assume, quite wrongly, since there is no mention of error, that all the figures are equally reliable or unreliable.

There is no easy way of learning how to use economic statistics correctly. There are no simple rules which will enable one to become an efficient expert in the use of the language. At every stage there are difficulties and pitfalls, and one can only acquire skill by patience and caution in actually handling the statistics oneself.
 

The political use of statistics

Introduction

Johnson (1988, p. 7) rightly notes that ‘In a democratic society official statistics should be beyond reproach’, that they should be a ‘public good rather than a political tool’, but that ‘Like all democratic ideals, this one is only imperfectly realized in the UK and many other Western democracies.’ The political use of statistics is currently very topical, but in fact has a long lineage.

The political abuse of statistics is inevitable in modern democracies where governments have monopoly powers over the supply of certain classes of data, and where there is great political advantage from blurring the distinction between normative information and propaganda (for example, the current political abuse of the unemployment statistics). This can be illustrated by examining budgetary data which is another class of statistical information where great care has to be taken since there is much advantage for government in obscuring the true position; in creating fiscal illusions as a means of gaining political and economic advantages. Let’s take a couple of examples to illustrate the continuity of this problem and the dangers it poses for the economic historian.

Case study I: interwar Britain

As some of you may be aware during the interwar period there took place an important economic debate over the role of the government budget: whether that budget should be small and balanced, or whether, as Keynes maintained, deficit-financing should be used to increase aggregate demand and reduce unemployment.

During the interwar period there was no statutory definition of the budget balance, and while no government was prepared to countenance deficit-financing they had to meet the twin pressures of maintaining confidence, which was dictated by financial markets, and incessant demands for new expenditures, particularly social expenditures to meet the unemployment crisis. The result was that government’s sought to maintain the appearance of balanced budgets, whilst in reality resorting to all sorts of devices and cosmetic actions to hide expenditures and artificially boost revenues.

 TABLE BG.1A Central government receipts, expenditure and budget balance, 1929/30-1939/40
 

SOURCE: Middleton (1985a, table 7.4).

We can see the effects of their actions in table BG.1A. The conventionally defined budget balance is here Bc (where T and G are receipts and expenditure respectively, and Y, the denominator in panel 2 of the table, is GDP), and it is this that they tried to keep in small surplus to reassure financial opinion. In reality, when we exclude their cosmetic actions, the real balance of the budget was B. Note that there are some large differences between Bc and B, particularly in the early and late 1930s. Why were they able to do this; simply because as one Treasury official noted in a private paper:

These budgetary practices have important implications for economists’ ability to measure the fiscal stance - a measure of the impact of the budget on the economy - as well as for the maintenance of democratic political systems.

Case study II: postwar Britain

Let’s now take a more recent example of the problems associated with budgetary data (see Middleton 1988, pp. 108-10), that concerning the effects of privatisation proceeds on the Public Sector Borrowing Requirement (PSBR), the official measure of the balance of public sector accounts.

The economic objectives of the Thatcher government first elected in 1979 were strongly influenced by an interpretation of postwar British economic history which saw Britain’s growth performance as constrained by the excessive growth of the public sector. As a consequence the incoming government had an explicit objective - which was codified in the Medium Term Financial Strategy (MTFS) announced in the March 1980 budget - of reducing the growth of public expenditure as a means of reducing both inflation and the PSBR.

 Table BG.1B UK general government expenditure, 1974/5-1987/8
 

SOURCE: Middleton (1988, table 7.3).

UK general government expenditure as a percentage of GDP is shown in table BG.1B for 1974/5-1987/8. As can be seen the first Thatcher government was spectacularly unsuccessful in reducing the share of public expenditure in GDP, and that it wasn’t before 1983/4 that the public expenditure ratio fell below that of the last year of the previous Labour administration (1978/9). Thereafter, the ratio has fallen steadily, in part because of the recovery of the real economy, but also because of privatisation proceeds. In Britain, the accounting convention is that the sale of state assets are classified as negative expenditures. More properly, they are capital receipts and should not be used as an offset against total public expenditure.

This has important implications for the course of the public sector in Britain over recent years. Between 1979/80 and 1986/7 total receipts from privatisation amounted to £12.5bn, with a further £15bn in prospect for the following three years. If these receipts had been allocated as positive revenues, public expenditure would have been much higher over these years. For example, in 1986/7, when the published share of public expenditure in GDP was 43.25%, adjustment for privatisation receipts yields a figure of 44.0%, which is, for example, some three-quarters of a percentage point above the figure of 1978/9, the final year of the previous Labour government.

Growth rates and other summary measures

As much of this course is concerned with changes in the value of time series variables, such as national income in £bn, it is appropriate here to make some observations about growth rates and other simple summary statistics employed by economic historians.

We begin by establishing what is meant by changes in a variable over time, for here there is much confusion, particularly about percentages. You must be clear in your own mind about the difference between a change expressed as a percentage and in terms of percentage points. For example, if the mortgage rate increases from 10% to 11%, by how much has it increased? - 1% or 10%. Answer = 10% or 1 percentage point.

 Table BG.1C Public authorities’ current expenditure on goods and services and GDP at constant prices (£m), 1924-37
 

 SOURCE: Calculated from Feinstein (1972, table 5).

Next, let’s introduce the concept of the elasticity measure which will be familiar to those who have done ‘A’ level economics. Consider table BG.1C which details a national income measure (Gross domestic product, GDP) and one component of public expenditure (Public authorities’ current expenditure on goods and services, PACE) over the period 1924-37. One way of presenting this would be to show the PACE/GDP ratio, and how this rises over the period (col. 3). But another way would be to calculate an elasticity measure; this shows us for each increment of GDP how much public expenditure rises, i.e. it is a measure of sensitivity.

The elasticity of PACE with respect to GDP is defined as:
 

•  PACE/PACE
   __________
•  GDP/GDP

The elasticity is thus positive and greater than unity. For any given percentage increase in GDP, PACE will increase by just over twice the increase in GDP in percentage terms.

One other general observation about growth rates is appropriate at this stage. It is most important to appreciate that macro data can be as much affected by cycles as well as by the trend. Therefore, it is important to select initial and terminal dates with care. For example, in the case we have just looked at our result would be much influenced by the initial and terminal dates as both the numerator (PACE) and denominator (GDP) are highly cyclical in their behaviour. In this case, as with most, it is appropriate to select initial and terminal dates which represent trade cycle peaks (or troughs).

Conclusions

Clapham (1933, p. 416), one of the founding fathers of quantitative economic history, laid down some rules for assessing quantitative information:

 Every economic historian should ... have acquired what might be called the statistical sense, the habit of asking in relation to any institution, policy, group or movement the questions: how large how long? how often? how representative?

To this we should add:

1. Develop a sense of doubt about data;
2. Why, how and for what purpose was the data collected;
3. How appropriate are the initial and terminal dates in a time series; and are the data affected by price changes; if so, is it possible to adjust them onto a constant price basis;
4. A numerical value is rarely useful (e.g. British exports increased by £15m over 1851-2); much more useful is to express the change as a proportion of some other variable - normally GDP is the most useful here.
5. Watch out for deliberate deceptions on the part of the writer; after all they might have been using the rules issued to Cambridge economics graduate students in the mid-1970s (reproduced as figure BG.1A).

  FIGURE BG.1A  How to slay ‘em with science
 
 

 
 
To IT-MA home page
To Department of Historical Studies home page.

(c)R. Middleton 1997. Last modified 30 June 1998.