The millions of bits of cellphone data tracked by analytics firm Unacast show the economic impact of the coronavirus spreading across the US like a deep freeze — long-distance travel was hit early on, but eventually overall retail foot traffic slowed to a crawl, too.

As the US started shutting down to fight the coronavirus outbreak, Unacast data shows, one thing became clear: White House messaging matters.

When President Trump declared a national emergency on March 13, there was an initial rush to grocery stores. Then visits to retail outlets overall collapsed by half in the following days, as even Americans who questioned the severity of the crisis appeared to accept it.

Since then, retail visits have remained deeply depressed, according to a new foot traffic scorecard here that breaks down the data by state and industry.

The New York firm tracks anonymized cellphone data from millions of people who opted to share their location through different apps, and cross-references it with physical “points of interest” including department stores, airports and other consumer-facing locations.

According to Unacast’s analysis, even states late to issue stay-at-home rules, such as Florida, have now seen large foot traffic declines, with only a few places, including South Dakota and Vermont, remaining close to their 2019 averages.

More noteworthy, said Unacast Chief Executive Thomas Walle, will be what the stats reveal about when recovery starts — particularly if, as appears might be the case, some states control the virus earlier than others and begin to loosen restrictions sooner.

“People’s behavior is changing drastically and the No. 1 data point to understand what is happening is mobility,” Walle said in an interview. “Social distancing and shelter in place is changing mobility patterns overnight.”

The ongoing retail freeze is no surprise in the midst of the pandemic, and has gone on long enough to start showing up in conventional economic data. US retail sales in March fell by a record 8.7 percent as social distancing rules and mandated shutdowns expanded with the health crisis.

But the lag in most government data, collected through surveys or administrative records and released typically a month or more after the fact, has left policymakers partially blind to the depth and implications of the fast-moving epidemic.

Unacast’s retail traffic data and a popular social distancing scorecard that the company released earlier are part of a flurry of recent efforts to use real-time, technology-driven data to understand the economic impacts of the crisis and the health response.

Economists have teamed with companies like Unacast and Homebase, which provides time management software for small businesses, to aggregate data and show with more frequency the fast decline, for example, in hours worked at restaurants.

Research released this week used cellphone data compiled by SafeGraph to study what researchers concluded is a partisan divide in the crisis response, with areas more supportive of Trump in the 2016 election less strict about social distancing.

That could all matter to the pace of the recovery, as economists look for signs things have stabilized and officials map out rules for a return to business.

Once that happens, cellphones may start moving again.

“Shutting down a store goes pretty quickly,” Walle said. “To reopen, the full machinery needs to be revamped — stocking up, bringing people in. What’s going to be important now is to look at the recovery when it is going to happen, where it is going, where do you direct your goods at what time, in what order.”