[Excerpt] Dunbabin takes issue with us on the question of whether and by how much the agricultural workers' movement led by Joseph Arch raised agricultural wages in the early 1870s. A number of authors have speculated on what the wage effect might have been, but these speculations have rarely been based on more than a cursory examination of agricultural wage data and they have not been able to separate the effects of unionism from other factors influencing wages. Our article was the first to narrow down the range of estimates by making the best use of available data for the whole country. We used two different approaches, cross-section and time series, and in each case we were careful to use alterative specifications to ensure the results were reasonably robust. Our results suggested that unionism raised agricultural wages between 1870 and 1875 by around 3 to 6 per cent on average, a modest but perceptible effect.
In our analysis we relied on statistics for wage rates and union membership which we believe are the best available, including the union membership estimates offered some years ago by Dunbabin. Dunbabin now argues that we have put too much faith in these statistics, including his own, and that our estimates are therefore subject to 'an appreciable margin of error.' It was never our intention to produce a precise and definitive figure for the union wage effect, but rather to identify a plausible range of estimates. Dunbabin does not argue that our estimates are too high or too low and offers no alterative estimate or method of estimating. The key issue therefore is whether the data and models we use are sufficiently well grounded to support the inferences we draw from them or whether, in the absence of any alterative, we should retreat from attempting to make any inference at all. Naturally, we believe the former.
The only way to satisfy Dunbabin's misgivings is to see how far our results are vulnerable to the specific comments he makes. If more data of better quality were available we presume that he would be satisfied since he makes no criticism of the model itself—only of the reliability of the data we use to estimate it. His own statements imply that he believes that agricultural wages were determined by essentially the same forces which are embodied in our model. Our approach then is to see how far our results change if we modify certain variables along the lines he suggests. Most of Dunbabin's criticisms apply to the data we used in our cross-sectional estimates and hence we shall concentrate on these and refer to the time series only in passing.