White noise periodogram coming out lower than I expected

I am comparing MATLAB's periodogram.m with my own code for periodogram PSD estimation. I've created 50 realizations of a purely white noise signal with a length of N=512. I take the periodogram of each (rectangular window), average the periodograms together, then plot the average. The results are nearly identical, except with my own code I get a flat line right around 0, but when I use periodogram.m, I get a flat line at -5. Shouldn't it be at 0? What am I missing here?

The following is the code I use:

% Calculate PSD of 50 Realizations and plot the average PSD Estimate.
% Signal Length 512
% MY CODE
clear
rng default             % Default random numbers
n_len = 512;            % Size of the signal
T = 1/(2*n_len-1);      
df = 0:2*T:1;        % Normalized Frequency Vector
% Generate 50 realizations of the signal
nt = randn(50,n_len);
for zz = 1:50   
  % Perform Autocorrelation
  for kk = 1:n_len
      for nn = 1:n_len
          if nn+kk-1 > n_len
              yy(nn) = 0;
          else
              yy(nn) = nt(zz,nn+kk-1)*conj(nt(zz,nn));
          end
      end
      r(kk) = (1/(n_len))*sum(yy);
  end
  % Flip r and combine with original
  r = [flipdim(r,2) r(2:end)];
  % FFT of autocorrelation 
  R(zz,:) = fftshift(abs(fft(r)));   
  clear r yy kk zz
end
% Calculate Ensemble Average
for zz = n_len:size(R,2)
  avR(zz) = mean(R(:,zz));
end
% MATLAB'S periodogram.m
for zz = 1:50
  X = nt(zz,:);
  [pxx(zz,:)] = periodogram(X,[],1024);
end
% Ensemble Average
for zz = 1:size(pxx,2)
  avpxx(zz) = mean(pxx(:,zz));
end
% Plot for Comparison
figure(1);
plot(df,10*log10(avR(n_len:end)), df, 10*log10(avpxx(1:end-1)))
axis([0 1 -20, 10])