Discuss the limitations in the use of index data in producing Williamson and Lindert?s results. Comment in reflection on Crafts and Flinn?s criticism.

Flinn argues that Lindert and Williamson’s turning-point in 1820 arises, not so much out of real differences as out of their method of selecting data.  ‘Rather than work from annual money wage series they have selected data relating to a few isolated years.’ They provide statistics for the years 1797, 1808, 1810, 1815, 1819, 1827 and 1835 only, a choice, which Flinn argues, lessens their accuracy and prevents them from seeing the main problems of measurement provided by Flinn himself. He goes on to say that, the fluctuated prices in the period of the French Wars disfigured measurement and therefore cannot be ignored when analysing.  In particular, the years 1812-13 represented a sharp peak of prices in both the short and the long run – years, which Lindert and Williamson failed to consider. ‘Trends taken from dates part-way up or part-way down these steep slopes can be very misleading and radically altered merely by shifting the chosen date a year or two either way.’ Lindert and Williamson’s isolated years run straight into this problem, especially for reasons regarding the years 1810 and 1815 as the figures produced for 1810 and 1815 can clearly be argued as inaccurate.

Yet another problem of isolated years in an age of violent short-run fluctuations is the possibility of their being poor indicators even of the medium run. Lindert and Williamson’s ‘best-guess’ cost-of-living index in Table 4 (page 11 of their article) shows that some of the years they have selected for wage data, fall victim also to the above problem.  ‘Both 1815 and 1835 turn out to be rather special years of short-run price troughs, 1815 being 9.2% below the mean and 1835 7.6% below its comparable mean.’  The price level in the final year of 1851 cannot even be tested similarly since the cost-of-living index does not even extend to that year. These problems would not matter according to Flinn if the purpose of their research were to analyse short-run fluctuations, however we are here concerned with secular trends, therefore their results are misleading.

Lindert and Williamson’s results show that there was a peak in money wage between 1810 and 1815 and that thereafter a relatively small decrease. The price fall was the main vehicle for this peak, at least until the 1830s for raising real wages.  ‘Because their new ‘best-guess’ cost-of-living index shows almost exactly the same price level for 1819 (182.9) as it does for 1815 (182.6), while price rises before 1810 were offset by very comparable money wage rises, they are able to conclude that the rise in real wages could not have begun until 1820.’  Therefore, according to Lindert and Williamson, after prolonged wage stagnation, real wages nearly doubled between 1820 and 1850, which confirms their assumption that 1820 clearly was the turning point.  However, Flinn believes that the sharpness of the turning point is slightly blunted by what looks like a short-run hiccup between 1810 and 1819. Flinn blames the inaccuracy of their conclusion on their limitation of their wage data at this period to the years 1805, 1810, 1815 and 1819.  Flinn has not challenged Lindert and Williamson’s main conclusions; he has challenged their turning point. He states, by his calculations, that the real-wage turning point was nearly a decade earlier, and that most real wage gains had been achieved before 1825.

Crafts discovers further problems within Lindert and Williamson’s analysis.  He argues that there are still significant data problems that make real wage measurements inaccurate. Lindert and Williamson’s cost-of-living index is not satisfactory and its deficiencies lead to serious bias in their results. In particular, it seems possible that real wage growth between the years 1821 and 1851 was about 1 per cent lower than they estimated.

Crafts however, concentrates on, for him, rather more serious problems to be found with house-rents, cereals and especially clothing.  Crafts argues that there is a better source for 1801-1851 house rents than results from Lord Stafford’s cottages. ‘Deane and Cole estimated house rental income in current prices, and Feinstein gives the value of the housing stock in constant prices.’ According to Crafts, the two sources together can be used together to make an estimate of a national rent index for benchmark years.  

Lindert and Williamson use Tucker’s index to analyse the cost of clothing prior to 1819 and thereafter the export price of cotton goods.  This new way to deal with the cost of clothing  is unacceptable for Crafts.  The assumption that cotton clothing was not consumed before 1819 is clearly inaccurate and the use of cottons to represent clothing after 1819 is seriously misleading.

With regard to cereal expenditures, Lindert and Williamson again have distorted figures.  Their index has weights of 12.6 percent for bread, 31.2 percent for flour, and nothing for oatmeal or potatoes.  However, cereal purchases showed great regional variations, which is worth noting, as prices of cereals did not appear close throughout 1780-1850.  For example, northern workers purchased more flour than bread, whereas southern workers bought more bread and little flour.  Therefore Lindert and Williamson’s mean figures clearly are inaccurate by disregarding these regional variations.

Finally, Crafts highlights an earlier point by Flinn regarding real wage growth.  In 1820-1850, wages as a share of national income and the share of consumption in national output were reasonably constant (wages were 45.6% of national income in 1821 and 44.9% in 1801).  This allows one to assume therefore, that real wage growth should logically be similar to consumption growth.  It is believed that by 1851, the economy was back to a ‘normal’ level of unemployment.  The addition to real wage growth allowing for this would probably raise estimates to approximately 1.2 percent per year - close to the rate of consumption growth.  Crafts find it hard to account for real wages growth exceeding output per head growth by as much as Lindert and Williamson’s estimates imply.  ‘This rude comparison between national output growth and real wages growth then supports the earlier conclusion:  Lindert and Williamson’s procedure lends to an overestimate of real wage growth for 1820-1850.’ 

In analysing a number of the limitations of the use of index numbers by Lindert and Williamson, it is clear of the small but yet significant mistakes one can make with index data.  Wrong approaches to the data leads to disorted and many times inaccurate assumptions.  However as initially expressed, Flinn and Crafts have not criticised the main conclusions by Lindert and Williamson and clearly still believe that index data offers a valuable method for researching the past, when used wisely.

BIBLIOGRAPHY:

History by numbers – pat hudson (2000) ox uni press

Index numbers in theory and practice – r.g.d allen (1986) macmillan press

The index number problem and its solution – g.stuvel (1989) mac press

A British Food Puzzle, 1770-1850

Author(s):             Clark, Gregory; Huberman, Michael; Lindert, Peter H.

Source:              The Economic History Review, New Series, Vol. 48, No. 2.

                     (May, 1995), pp. 215-237.

Why Was British Growth So Slow During the Industrial

                     Revolution?

Author(s):             Williamson, Jeffrey G.

Source:              The Journal of Economic History, Vol. 44, No. 3. (Sep.,

                     1984), pp. 687-712.

Unequal English Wealth since 1670

Author(s):             Lindert, Peter H.

Source:              The Journal of Political Economy, Vol. 94, No. 6. (Dec.,

                     1986), pp. 1127-1162.

 

---