If the details of the formatting of the web page will not change often then a little bit of list and string hacking might work
h[s_] := If[DigitQ[StringTake[s, 1]] || StringTake[s, 1] == "-", ToExpression[s], s];
g[s_] := h[StringReplace[s, {"%" -> "", "," -> "."}]];
f[v_] := Join[{First[v]}, Map[g, Rest[v]]];
data3 = Map[f, Rest[data[[3]]]]
which gives this
{{1981, 6.82, 6.15, 5.1, 5.04, 7.39, 3.97, 6.69, 6.28, 5.29, 4.57, 5.88, 4.37},
{1982, 8.32, 6.14, 6.36, 5.18, 6.46, 7.38, 6.54, 5.94, 5.61, 4.59, 4.59, 6.04},
{1983, 10.45, 6.06, 8.54, 7.8, 5.57, 7.87, 12.06, 8.32, 9.48, 10.9, 6.96, 7.44},
{1984, 10.14, 9.1, 8.98, 9.23, 9.88, 8.14, 12.62, 7.32, 11.39, 10.87, 10.35, 10.35},
{1985, 14.61, 8.95, 12.78, 8.8, 6.76, 7.71, 9.27, 12.1, 11.98, 9.6, 11.12, 13.36},
{1986, 16.23, 14.36, -0.11, 0.78, 1.4, 1.27, 1.71, 3.55, 1.72, 1.9, 5.45, 11.65},
{1987, 13.21, 12.64, 16.37, 19.1, 21.45, 19.71, 9.21, 4.87, 7.78, 11.22, 15.08, 14.15},
{1988, 18.89, 15.7, 17.6, 19.29, 17.42, 22., 21.91, 21.59, 27.45, 25.62, 27.94, 28.7},
{1989, 37.49, 16.78, 6.82, 8.33, 17.92, 28.65, 27.74, 33.71, 37.56, 39.77, 47.82, 51.5},
{1990, 67.55, 75.73, 82.39, 15.52, 7.59, 11.75, 12.92, 12.88, 14.41, 14.36, 16.81, 18.44},
{1991, 20.75, 20.72, 11.92, 4.99, 7.43, 11.19, 12.41, 15.63, 15.63, 20.23, 25.21, 23.71},
{1992, 25.94, 24.32, 21.4, 19.93, 24.86, 20.21, 21.83, 22.14, 24.63, 25.24, 22.49, 25.24},
{1993, 30.35, 24.98, 27.26, 27.75, 27.69, 30.07, 30.72, 32.96, 35.69, 33.92, 35.56, 36.84},
{1994, 41.31, 40.27, 42.75, 42.68, 44.03, 47.43, 6.84, 1.86, 1.53, 2.62, 2.81, 1.71},
{1995, 1.7, 1.02, 1.55, 2.43, 2.67, 2.26, 2.36, 0.99, 0.99, 1.41, 1.47, 1.56},
{1996, 1.34, 1.03, 0.35, 1.26, 1.22, 1.19, 1.11, 0.44, 0.15, 0.3, 0.32, 0.47},
{1997, 1.18, 0.5, 0.51, 0.88, 0.41, 0.54, 0.22, -0.02, 0.06, 0.23, 0.17, 0.43},
{1998, 0.71, 0.46, 0.34, 0.24, 0.5, 0.02, -0.12, -0.51, -0.22, 0.02, -0.12, 0.33},
{1999, 0.7, 1.05, 1.1, 0.56, 0.3, 0.19, 1.09, 0.56, 0.31, 1.19, 0.95, 0.6},
{2000, 0.62, 0.13, 0.22, 0.42, 0.01, 0.23, 1.61, 1.31, 0.23, 0.14, 0.32, 0.59},
{2001, 0.57, 0.46, 0.38, 0.58, 0.41, 0.52, 1.33, 0.7, 0.28, 0.83, 0.71, 0.65},
{2002, 0.52, 0.36, 0.6, 0.8, 0.21, 0.42, 1.19, 0.65, 0.72, 1.31, 3.02, 2.1},
{2003, 2.25, 1.57, 1.23, 0.97, 0.61, -0.15, 0.2, 0.34, 0.78, 0.29, 0.34, 0.52},
{2004, 0.76, 0.61, 0.47, 0.37, 0.51, 0.71, 0.91, 0.69, 0.33, 0.44, 0.69, 0.86},
{2005, 0.58, 0.59, 0.61, 0.87, 0.49, -0.02, 0.25, 0.17, 0.35, 0.75, 0.55, 0.36},
{2006, 0.59, 0.41, 0.43, 0.21, 0.1, -0.21, 0.19, 0.05, 0.21, 0.33, 0.31, 0.48},
{2007, 0.44, 0.44, 0.37, 0.25, 0.28, 0.28, 0.24, 0.47, 0.18, 0.3, 0.38, 0.74},
{2008, 0.54, 0.49, 0.48, 0.55, 0.79, 0.74, 0.53, 0.28, 0.26, 0.45, 0.36, 0.28},
{2009, 0.48, 0.55, 0.2, 0.48, 0.47, 0.36, 0.24, 0.15, 0.24, 0.28, 0.41, 0.37},
{2010, 0.75, 0.78, 0.52, 0.57, 0.43, 0., 0.01, 0.04, 0.45, 0.75, 0.83, 0.63},
{2011, 0.83, 0.8, 0.79, 0.77, 0.47, 0.15, 0.16, 0.37, 0.53, 0.43, 0.52, 0.5},
{2012, 0.56, 0.45, 0.21, 0.64, 0.36, 0.08, 0.43, 0.41, 0.57, 0.59, 0.6, 0.79},
{2013, 0.86, 0.6, 0.47, 0.55, 0.37, 0.26, 0.03, 0.24, 0.35, 0.57, 0.54, 0.92},
{2014, 0.55, 0.69, 0.92, 0.67, 0.46, 0.4, 0.01, 0.25, 0.57, 0.42, 0.51, 0.78},
{2015, 1.24, 1.22, 1.32, 0.71, 0.74, 0.79, 0.62, 0.22, 0.54, 0.82, 1.01, 0.96},
{2016, 1.27, 0.9, 0.43, 0.61, 0.78, 0.35, 0.52, 0.44, 0.08, 0.26, 0.18, 0.3},
{2017, 0.38, 0.33, 0.25, 0.14, 0.31, -0.23, "---", "---", "---", "---", "---", "---"}}
and which might be acceptable to export as an Excel file.
But list and string hacking raw data seems like it often comes back to bite you later.
Might it be possible to do this in reverse? Can you write one short expression in Excel which would import the web page onto a sheet, transparently deal with localization of currency and export the desired range of cells of that sheet to an xlsx file which you can then import into Mathematica?
If not then many many many years ago there were fairly inexpensive software products which would try to do a reasonable job of extracting data from html pages and putting that into a useful Excel sheet. These were smarter than just pasting the contents. Excel gives a much broader market and you might find something that would be less "brittle" and likely to fail with the smallest change to the web page format than the simple bits which I wrote.