Thus, while it is inevitable that individual states would experience higher volatility than the national average, it is interesting that the coefficient of variation of monsoon rainfall, which was only 8 per cent at the national level during 1986 to 1999, had ranged from 10 to 40 per cent in the major states. During 1986 to 1999, rainfall variability as measured by its coefficient of variation was highest in Haryana (40 per cent), Gujarat (35.4 per cent), Rajasthan (31.7 per cent) and Punjab (30.9 per cent). But since Punjab and Haryana have almost wholly irrigated agriculture, the effect of the large rainfall variability is muted on production, leaving Gujarat and Rajasthan as the states where monsoon variability has had most effect on agricultural production. This is, of course, nothing new since this region has historically had the highest rainfall variability and had developed responses, including cropping patterns, to cope with this. What is, however, noticeable about this period is that certain other states which traditionally have had high rainfall variability, such as Karnataka, Madhya Pradesh, Maharastra and Orissa, all had low coefficients of rainfall variation during 1986-99 at 10.4, 16.2, 13.9 and 16 per cent respectively. Among the other states, Andhra Pradesh (22.5 per cent), Bihar (18 per cent), and Tamil Nadu (23 per cent) have had relatively high rainfall variation, while Assam (15.2 per cent), Kerala (16.6 per cent), Uttar Pradesh (14.1 per cent), and West Bengal (12.3 per cent) have had relatively low variation.
 
What also matters for the effect of rainfall on crop production at the all-India level is the degree to which rainfall variations across states are correlated with each other, since if the correlation is positive and high the variations at state level are likely to reinforce each other while they could cancel each other if the correlation is negative. As the charts show, there is in most cases a positive correlation between the state-specific and the all-India indices. This positive correlation is high (ie correlation coefficient greater than 75 per cent) in the case of Gujarat, Haryana, Karnataka, Maharashtra, Punjab and Uttar Pradesh. but is low (ie correlation coefficient less than 30 per cent) in the case of Assam, Bihar, Kerala, Orissa, Tamil Nadu. For Andhra Pradesh, Madhya Pradesh and Rajasthan, the correlation between the state-specific and all-India indices lie between 40 and 50 per cent, also indicating relatively low correlation. And, in the case of West Bengal the correlation is in fact negative. Thus, it may be said that there are two basic reasons why the all-India rainfall has been more stable since 1988: first, the rainfall pattern in the East and the South (excluding Karnataka) has had very low correlation with the rest of the country, and, second, certain traditionally high variation states (ie Karnataka, Madhya Pradesh, Maharashtra and Orissa have had relatively low rainfall variations. In addition, it must be noted that there was not a single year since 1988 when rainfall was less than 80 per cent of normal in Andhra Pradesh, Karnataka, Madhya Pradesh, Maharashtra, Punjab and West Bengal; and in Bihar, Rajasthan and Uttar Pradesh rainfall has not fallen below 80 per cent of normal since 1992.