Goldsworthy is not a classicist or strictly speaking a political historian, but rather a military historian (with classical training), and statements like this (although as I recall it's not stated quite so starkly) are where it shows. Goldsworthy's repeating an idea that first really gained prominence with Ronald Syme in the 1930s and 40s as a reaction against Mommsen's anachronistic juxtaposition of contemporary politics with the Roman world, a methodology that unfortunately is still very much alive in popular history and the public imagination. There was a certain amount of comparison in Syme between his world and the Roman world (the comparisons between Hitler and Augustus in The Roman Revolution (1939) are obvious and intended to be striking), but Syme reacted strongly against Mommsen's belief that the socio-political structure of the Roman world was essentially identical to Mommsen's own late nineteenth century Germany. Instead, Syme argued, the senatorial nobility that dominated Republican politics had little interest in actual political issues or the well-being of the state and were concerned mostly with their own private reputations and means. Essentially what Sallust says in the Bellum Catilinum when he accuses people of squandering the fruits of empire on extravagant houses and stuff and encouraging insidious vices in the younger generation. This reading was heavily influenced by Matthias Gelzer, who published in 1969 The Roman Nobility, which is really a collection of two works from 1912 and 1915. Gelzer examined much more closely than Mommsen had the composition of the senatorial nobility and the way in which it behaved internally and externally, as well as how one entered into it. His findings were that the senatorial class, and particularly the nobiles, descended from consulars, played an elaborate game of personal rivalries in which they employed the apparatus of state and popular sovereignty, mainly through control of large bodies of clients that could be mustered, traded, or withheld, all in the interests of personal ambition. A sort of Realpolitik, but one restricted to the most narrow segment of Roman society and one in which the lack of ideology or engagement with "reality" extended down to the citizen body, which formally had great powers but in actuality was controlled pretty much exclusively by their patrons. This idea attracted a lot of attention, and still does at some levels, particularly after Syme argued from it that the end of the Republic was inevitable since the highly competitive nature of the senatorial class prevented them from uniting consistently against any single threat but simultaneously encouraged individuals to rise above the rest of the nobility.

In this respect Goldsworthy's popular history is more up-to-date than most, in that it's not a regurgitation of Mommsen. However, the main authors of this school of thought, which relied heavily on prosopography, wrote rather a long time ago. Syme, Gelzer, Scullard, and Badian were mostly active from the 1930s to the early 1980s or so, and Munzer (probably the greatest classical prosopographer before Badian) died in a Nazi concentration camp in 1942. Much has changed since then. In particular, Fergus Millar's 1984 article "The Political Character of the Classical Roman Republic, 200–151 B.C." was groundbreaking in its contention that in fact clientage was not in fact particularly politically important, that Republican politicians did not simply play at political office with each other, that in fact there were real political issues at stake during forensic contests, and that--perhaps most crucially (and controversially) today--the citizen voting assemblies had real legal and ideological powers, as Polybius insists they did, and were not composed of mobilized bodies of clients voting mindlessly for their own patrons and whomever their patrons supported. Millar's work, which has increasingly gone further and further off into a "democratic" model of the Republic that pretty much nobody but he agrees with in its most radical form, opened the doors entirely to a completely new way of looking at Republican politics, one that Goldsworthy is not really up to speed on. One of Millar's most interesting realizations, which has garnered a lot of attention in the forty years since, is that contiones, the public, sort-of-formal-sort-of-not addresses of magistrates to the crowds in the forum, do not make sense if clients are controlling the assembly. Either the voting electorate in fact had to be won over by speakers or Republican politicians consistently wasted their time on an activity that they knew had no purpose and that was not actually required by any office or candidacy. From this broadly two schools of thought have emerged, probably best exemplified by Mouritsen and Morstein-Marx. Pretty much everybody has discarded the model that clientage was the silver bullet that explains the Republic, since there's not very much evidence for the use of clients as a controlling measure by the senatorial class. Mouritsen's argument has essentially been that although the powers of the voting assemblies and their composition were in theory quite vast, in reality they could be closely controlled by the use of (for example) the auspices and that in any case Roman voting procedures and spaces were so restricted that only a tiny fraction of the Italian citizenry could participate in any given vote. Mouritsen further argued that the forensic crowds that made up the audiences of contiones were predominantly senatorial and equestrian nobles themselves, and that the urban plebs had little interest in or economic ability for participating in politics. Which is not unlike Syme's model except that it discards clients as a relevant factor. However, Yakobson has pointed out that the first class of the Centuriate Assembly was in fact highly diverse and almost guaranteed to be split at the vote, meaning that in practice elections probably frequently descended to the lower property classes--the sympathies of the urban plebs would therefore be extremely important at elections, since candidates could not count on a united first class. Morstein-Marx has further countered that we in fact rarely find the sort of "counter contiones" that Mouritsen says would have existed, in which rival politicians gathered their own bodies of supporters from the Roman gentry and basically argued against each other in enclosed, controlled spaces that did not overlap. Instead, Morstein-Marx found in Mass Oratory that contiones on any particular issue tend to be dominated a single side, and that the opposition is essentially denied the ability to enter the discourse as an equal. This, along with other considerations like his findings that rumor and communication from the lower levels of society exerted influence on Roman politicians of the late Republic, has led Morstein-Marx to argue for a model of Republican politics in which ideology and real political issues were at stake, and in which Republican politicians actively interacted with the citizen body.

That's where the scholarship is pointed these days. Mouritsen's model appears broadly similar to the old one (what North called the "frozen waste" theory of Republican politics) in that it posits fairly tight control over the voting assemblies by the senatorial class and argues that only lip service was paid to the idea of the citizenry's electoral powers, but even Mouritsen's model works according to a very different system, discarding entirely clientage and the neat circles of senatorial factions constructed by prosopography. The model Goldsworthy presents, while not wholly worthless, is these days pretty archaic, and had already been under heavy fire for a good twenty years or so when his book on Caesar was published.