We investigate the temporal dynamics of correlations between sentiment indices worldwide. Employing the tools of Random Matrix Theory (RMT) and Principal Component Analysis (PCA), our paper aims to extract latent information embedded in the interactions between economic and business sentiment indices around the world. We find that, 1. the dynamics of the sentiment indices across countries can be well explained by the evolution of a single factor (the "market mode"), 2. during most periods, some groups of countries exhibit sentiment dynamics less associated with (or divergent from) the market mode, while 3. during the financial crisis, no country or group of countries has been able to escape the market mode, which accounts for almost all movements in the indices. We argue that strong "global" information signals, like the collapse of the US housing market in 2007, can lead to a homogenisation of the expectation structure around the world, as such information can provide a coordination signal for a global phase of low confidence.