As they introduced longer lags in the price variable , the contribution of their demand coefficient steadily increased .sx From a technical point of view , using the method of least squares estimation , we find the principal factor at work to be that the introduction of a lag in the price variable systematically reduces its coefficient , while the other factors remain relatively stable .sx This effect is increased in the formulation of ( 4.1 ) where a moving average of price changes is taken .sx In all cases , the overall correlation remains virtually unchanged .sx Empirically , we have been unable to determine a unique lag between wage and price changes , and we have therefore had to rest content with the lag of six months built into the model from 6a priori reasoning .sx We are disposed to conclude from our estimates that demand has been an important factor at the bargaining table .sx At the level of aggregation at which we are working , it is the general state of the demand for labour that is relevant .sx It has been pointed out by H. A. Turner that in 1952-53 recorded unemployment in the cotton industry was as high as 30 per cent .sx of the industry's manpower , whilst the 'unions not only presented to the employers a demand for a general wage increase but persisted to the point of partial success' .sx The point here is that the unions' hand was strengthened by the existence of alternative employment .sx The kind of effect that we are considering is that of the influence of general unemployment .sx A 200,000 increase in the level of unemployment will doubtless have a greater effect on the change in wage rates if it is spread over all industries equally , rather than concentrated in one industry alone .sx The situation in the cotton industry can then hardly be cited as evidence against the influence of demand .sx The political variable , F;t ; , represents the influence of cost-inflation .sx It is sometimes argued that the trade union leader's job is to obtain higher wages for his members , and it matters little how he does this .sx He may rationalize his demand for higher money wages in terms of the cost of living , the level of profits or increases in productivity .sx In the absence of such factors he may push for increases in money wages on principle , the strength of the push depending in part on the extent to which the government of the day can create an atmosphere of restraint .sx In the model , the coefficient of F;t ; indicates that in periods when the Conservative government has been in power , unions have been pushing harder to the extent of some three index points per year .sx This is a statistically firm coefficient which may be taken as evidence of the increased importance of autonomous trade union pressure over the second half of our sample period .sx It is extremely difficult on the basis of the evidence , in the form of our estimates , to maintain that over the whole sample period changes in the wage rate index have been 'cost' rather than 'demand' induced .sx This result can be rationalized by arguing that to some extent the strength of the 'cost-push' at the bargaining table is governed by the demand for labour .sx The two complement each other .sx Without the existence of 'excess demand' for labour , the 'cost-push' might not go very far ; without the 'cost-push' in a situation of 'excess demand' , workers might be unable to exploit their favourable position .sx To this extent , changes in wage rates are dependent on both 'cost-push' and the level of 'excess demand' .sx Several writers have drawn attention to changes in the spread between wage earnings and wage rates , as a criterion to distinguish demand from cost inflation .sx The limited information estimate of equation ( 4.2 ) attributes greater significance to hours worked than to productivity , though both can be considered significant .sx The residual terms in this computation were , however , highly autocorrelated .sx The estimate from transformed data has no significant serial correlation in the residuals but gives a different estimate for the relative importance of productivity and hours worked in accounting for the spread .sx The hours worked variable dominates the relation and , if anything , our estimates in ( 4.2 ) are over-generous in attributing fluctuations in the spread to fluctuations in productivity .sx Our results are not inconsistent with the hypothesis that the spread is largely influenced by the level of demand .sx If this is correct , fluctuations in the spread could be regarded as an indicator of changes in the level of demand .sx Hours worked constitute a very sensitive indicator of the level of demand , although absolute changes in the hours worked index are small .sx In one sense hours worked contribute to the spread in a purely accounting manner as do , other things being equal , increases in output per man-hour for piece-workers .sx We are unable , however , to separate out the relative importance of these two influences .sx No doubt , overtime , bonus payments , premium rates , and changes in the length of the 'official' working week , have all been important .sx On the demand side , it is often argued that a high level of demand has led to payments above the 'official' rates to bid labour away from some firms into others .sx Our results are consistent with either hypothesis alone or a combination of both .sx It is interesting to compare our results with those obtained by the Swedish economists , Bent Hansen and " sta Rehn , in their study of the Swedish labour market .sx They start from the assumption that wage rates are fixed institutionally and that the influence of economic forces is reflected in the spread between earnings and wage rates , which they describe as the 'wage drift' .sx In our model we have put forward the hypothesis that changes in wage rates are influenced by changes in the cost of living , by the demand for labour , and by the political climate .sx The procedure followed by Hansen and Rehn was to take a sample of annual data , 1947-54 , for eight main groupings within Swedish manufacturing industry .sx Briefly , their findings suggested that the main influence determining the 'wage drift' in Sweden over these years has been 'excess demand' .sx They tested the further influence of 'excess profits' and the hypothesis that increases in productivity have contributed substantially to the 'wage drift' .sx Neither was found to be significant .sx The relations estimated were between the rates of change of the 'wage drift' , the level of 'excess profit' , the level of 'excess demand' and the rate of change in productivity .sx It may be pointed out that in our model productivity makes a significant contribution to the explanation of the spread between earnings and wage rates , when all variables are expressed as levels , but ceases to be a significant factor in our least squares computation in which variables are subjected to a first-difference transformation .sx It would appear , therefore , that our findings are not inconsistent with those of Hansen and Rehn .sx However , it must be re-emphasized that we have included hours worked in our computation , which are an indicator of the direct influence of demand on the spread , but also of other influences , and that our earnings variable is of average weekly earnings and not of hourly earnings .sx Hansen and Rehn , on the other hand , construct an index of 'excess demand' for labour , in some cases by taking the difference between unfilled vacancies and numbers unemployed .sx Where numbers unemployed were not available for a particular industry , vacancies alone were used .sx It is not clear that the growth of the spread between earnings and wage rates in the UK over the period of our sample can be plausibly explained in 'cost' terms .sx If it is argued that such a gap is automatically opened by the rise in piece-workers' earnings as productivity increases , or by changes in the amount of overtime worked , such changes may themselves be traced back to the existence of a high level of demand .sx Equation ( 4.3 ) illustrates the close relation between hours worked and the level of industrial production , which itself reflects the level of demand .sx Passing through this chain of causation , it would be plausible to expect a high empirical correlation between changes in the 'wage drift' and the level of 'excess demand' for goods and services .sx Under the assumptions implicit in the model , this relationship merely constitutes a derived relation rather than a basic structural equation .sx To estimate the determination of the 'wage drift' in this form would involve obscuring the underlying chain of relationship .sx It is sometimes held that the changes in the 'wage drift' are not governed by the level of 'excess demand' , since this would imply some bidding up of payments to workers over the 'official rates' .sx It is then contended that just as manufacturers have not bid up prices directly in response to 'excess demand' , so they have not bid up wage payments .sx The 'mark-up' equation ( 4.4 ) suggests that earnings have been roughly twice as important at the margin as import prices in determining the general level of consumer prices over the sample period .sx The coefficient of the import price index represents the influence of 'cost-push' to the UK economy .sx The level of earnings , on the other hand , may represent both the influence of 'cost-push' and that of demand , for it is through earnings that demand affects the general consumer price level in our system .sx The Implications of the Model .sx Our particular model of the inflationary process brings out points that have been raised by different writers and attempts to follow through some interrelated patterns of behaviour in the sphere of wage and price determination .sx The model illustrates the influence of both cost and demand elements .sx It is not unique , as judged by its agreement with observed data , and it contains flaws ; nevertheless it appears to be reasonable , and the difficulties that it encounters are inherent in the nature of our basic economic information .sx Our statistical analysis covers the post-war period as a whole .sx As such , it gives a set of average relationships which do not rule out dispersion .sx In the Korean war period , for example , the rise in import prices would appear to have made a much greater contribution to the rise in the general consumer price level than earnings , although , on average over the period as a whole , earnings appear to have been the more important factor .sx This fact limits the usefulness of the model in enabling us to comment on the debate on the character of inflation over such a short period as say 1956-57 .sx Our results seem to show that for the period of the sample as a whole it is not possible to assert categorically that we have had either demand or cost inflation .sx The model attributes significance to both cost and demand elements .sx Even Professor Robbins , a firm protagonist of the importance of the influence of demand over the period , is prepared to concede that for the latter half of 1957 and the first half of 1958 , the rise in final prices may have been largely 'cost-induced' , as an overshoot from the period of 'excess demand' .sx Consider , however , the period 1956-57 when price and wage changes were substantial and over which much controversy has raged .sx In 1955 and 1956 unemployment had fallen considerably from the relatively high level of 1952 .sx If we accept our equation ( 4.1 ) as a basic structural relation , then we are virtually committed to accepting the view that the level of 'excess demand' for labour had a significant effect on wage rate changes in that period .sx It may however be argued that ( 4.1 ) places overmuch weight on the influence of average unemployment .sx In the base year our average of registered unemployment , in numbers , was around 350,000 .sx If the price level were stationary for sufficiently long so that the influence of that variable in ( 4.1 ) were to become zero , then average unemployment in terms of numbers registered unemployed would have to reach only 500,000 before the four-quarter change in the wage rate index would become negative ( assuming [FORMULA] ) .sx Many would , however , find this conclusion implausible .sx They would no doubt argue that ( 4.1 ) can hardly be considered reversible .sx Wage rates go up but they do not come down .sx