Knowledge of livestock population dynamics is important to better understand functional attributes and development potential of pastoral production systems. With a focus on the Borana system of semi-arid Ethiopia for 1980–97, the main objectives of this research were to: (1) Characterize cattle population trends; (2) determine associations of rainfall and stocking rate with change in cattle numbers; and (3) estimate economic losses from cattle mortality. We predicted that the regional cattle population trend would consist of a “boom and bust” cycle with long periods of gradual herd growth punctuated by drought-induced losses. We expected that cattle losses would occur when high stocking rates were combined with large rainfall deficits. Such observations would refute the idea that cattle numbers were erratic and purely controlled by rainfall variation, as predicted by non-equilibrium theory. Cattle dynamics were quantified using herd histories from interviews of 56 households living in 4 sites. Data were aggregated to portray regional cattle population trends and quantify economic losses. Regression analysis was employed to assess associations of rainfall variation and stocking rate with cattle dynamics using 2 approaches: (1) Regional using aggregate herd data, empirical rainfall records, and calculated estimates for stocking rates; and (2) local using site-specific herd data along with recall of rainfall and stocking rate dynamics. Overall, results confirmed that cattle numbers followed a boom and bust cycle. Average cattle holdings dropped from 92 to 58 head/household between 1980 and 1997, respectively. Droughts in 1983–5 and 1991–3 resulted in the deaths of 37 to 42% of all cattle, respectively, up to 15-times higher than net sales. Over 17 years our target population of 7,000 households lost 700,000 cattle with a capital asset loss valued at USD 45 million. Statistical results were more difficult to interpret. Our regional approach indicated neither rainfall nor stocking rate were significantly associated with cattle mortality. We felt this interpretation was erroneous, however, due to a probable—but unmeasured—decline in key grazing resources that lowered carrying capacity, increased herd instability between successive droughts, and undermined relationships among model parameters. Our local approach was somewhat clearer in that results indicated cattle losses were significantly and consistently associated with rainfall deficits, and occasionally associated with high stocking rates that varied by site.