Long records of sea level show decadal and multi-decadal oscillations of synchronous and asynchronous phases, which cannot be detected in short-term records. Without incorporating these oscillations, it is impossible to make useful assessments of present global accelerations and reliable predictions of future changes of sea level. Furthermore, it is well known that local sea-level changes occur also because of local factors such as subsidence due to groundwater or oil extraction, or tectonic movements that may be either up or down. Limited data from limited areas of study are, therefore, unsuitable for making predictions about the whole world sea level. Yet, people continue to make such predictions, often on an alarming scale. Here, we use one example to illustrate the problems associated with trying to make sea-level predictions based on a short record (25 years) in a limited region. Linear and parabolic fittings of monthly average mean sea levels (MSL) of global as well as different local (United States Atlantic Coast, United States Pacific Coast) data sets of long tide gauge records. It is clear from the analyses of the tide gauges of the “NOAA-120”, “US 39”, “PSMSL-162”, “Mitrovica-23”, “Holgate-9”, and “California-8” data sets and the United States Pacific and Atlantic coasts that the sea level has been oscillating about the same almost perfectly linear trend line all over the 20th century and the first 17 years of this century. It is of paramount importance to discuss the proper way to assess the present acceleration of sea levels. This can not be done by focusing on the short-term upward oscillations in selected locations. The information from the tide gauges of the United States does not support any claim of rapidly changing ice mass in Greenland and Antarctica. The data only suggest the sea levels have been oscillating about the same trend line during the last century and this century.