In West Gojam, the coefficients for both local and high grade crossbreds were higher for the urban system than for the rural and peri-urban systems.The analysis model correctly classified 50.6% and 38.7% of the 180 sampled farms in each of West Shoa and West Gojam zones, respectively, into rural, periurbanand urban dairy farming systems. The percentage of farms that were correctly classified into their original rural, peri-urban and urban systems were respectively49.2%, 52.5% and 50.0% in West Shoa and 43.2%, 25.0% and 50.0%in West Gojam.The highest misclassification was observed in the peri-urban system where 51.7% of the farms were classified into the rural system in West Gojam, and 45.7% of the farms were classified into urban system in West Shoa. The improvement in accuracy of classification of the farms using cross-validation method, where classification was done by classifying each farm by the discriminant functions derived from all farms other than the farm to be cross-validated, was not significant.

When cross-validation was employed,0.0% to 9.9% of the originally classified farmers were reclassified into a different system . One of the major distinctions between smallholder dairy farming systems in the highlands of Ethiopia is the type of breeds and genotypes kept by farmers. The genetic structure described in the current study is in general agreement with previous results in Ethiopia in that crossbred herds are more predominant in the urban and peri-urban systems than in the rural system  and elsewhere in Africa where use of AI breeding service and herd genetic composition were found to be significant discriminating factors among peri-urban farmers in Mali. The urban/peri-urban systems have been commonly and grossly characterized as crossbred-based systems in Ethiopia , crossbred cows accounting for up to 97.5% of the herds . Such characterization could represent the urban/peri-urban systems in and around big federal and regional cities and major milk sheds. However, our results which were based on rigorous statistical analysis indicated that the characterization of the systems, particularly peri-urban/urban systems, needs qualification.

It was found in the current study that merely38% – 50% of urban/peri-urban farms conformed to the characteristics of these systems, namely breeding for high/medium grade crossbreds. This result is inagreement with a previous report for urban systems in regional towns where the proportion of crossbred cows accounted for only 26% – 40% of the total herd . The current results thus call for the qualification of the classification and/or the characterization of the smallholder dairy systems in and around the big cities and regional towns to inform research and development policies, strategies and interventions.A major distinction between the three systems observed in this study was that rural farmers are more likely to keep low grade crosses with exotic blood levels of around 25% and less likely to keep high grade crossbred cows with exotic blood levels of 75% and above than urban/peri-urban farmers. This is in agreement with and observations that medium to high grade crossbred cows are kept in urban systems in Ethiopia. The exotic inheritance level in village cattle herds could be due to farmers’ conscious preferences for the various exotic blood levels or lack of awareness and/or access to breeding services.

Farmers’highest preference ranking in rural and peri-urban/urban areas respectively werefor <50% and 50% – 75% exotic blood levels in West Shoa and 50% – 75%exotic blood level was preferred elsewhere in the highlands of Ethiopia The choice of breeds and genotypes of crossbreds for smallholder systems in Ethiopianeed to consider the dairy value chain. Where the value chain is not functioning well, such as in the rural areas, where access to inputs and services and profitable markets for fresh milk is limited, the indigenous and low grade crossbred cows could be the best choice. However, it could be more likely that the low grade crosses with 25% exotic blood in rural areas in the current study are a result of lack of access to AI breeding services where farmers are forced to backcross 50%exotic crossbred cows to local bulls. This argument is supported by the higher number of local bulls in rural areas compared to periurban and urban areas in West Shoa where low grade crosses are higher in rural areas, whereas in West Gojam where low grade crosses are higher in Peri-urban and urban areas, the average number of local bulls were 0.12, 0.28and 0.23 in rural, peri-urban and urban systems.

A similar pattern has been reported elsewhere in Ethiopia; for instance most of the farmers in Lay Armacho, Debark and Gondar Zuria districts did not keep the initial proportion of exotic genes in their herd constant Backcrossing of first crosses has been found to significantly affect milk production performance where daily milk yield was reduced from 6.2 in 50% first crosses to 5.6 liters inF2 backcrosses . Maintaining the desired exotic blood level in village herds has proved difficult resulting in genetically admixed village population in Ethiopia due to inefficient AI system and absence of controlled breeding in village herds due to communal grazing systems, particularly in rural areas.