You can use the history function to get historical prices and to compare prices between bars. To get yesterday's close for a particular stock you can do:

history(2, '1d', 'close_price')  

This will return a datapanel with yesterday's close price and today's current price. Note that you'll need to run your algo in minute mode when you use the history function.

Or in daily mode:

def initialize(context):  
    context.stocks = sid(2)

def handle_data(context, data):  
    closes = get_closes(data, context)  
    if closes is None:  
        return  
    # Yeterday's close  
    print closes[context.stocks][0]  

@batch_transform(window_length=2, refresh_period=0)  
def get_closes(data, context):  
    closes = data['close_price']  
    return closes  

P.

Or you can do something like this where you initialize a dictionary containing the sid as the key with the prior closing price, for example, as the value.
Before finishing the main loop, you set the prior_close to the current close to update the dictionary.

Instead of using this structure for storing just a single value, it can be used to store information per security for items such as a list, whole class structure, or nested dictionary.

def initialize(context):  
    context.portfolio_ = [sid(8554), sid(19920)]  # SPY, QQQ  
    context.prior_close = dict(zip(context.portfolio_,  
                                   [None] * len(context.portfolio_)))  
def handle_data(context, data):  
    for stock in context.portfolio_:  
        close = data[stock].close_price  
        # for example...  
        last = context.prior_close[stock]  
        daily_return = close / float(last) - 1 if last is not None \  
                        else None  
        log.info('stock: {0}, close: {1}, last: {2}, daily_return: {3}'\  
                 .format(stock, close, last, daily_return))  
        context.prior_close[stock] = close  

Here's a bit more detail on how to use history and pandas/numpy to extract the prices:

def initialize(context):  
    context.stocks = [sid(8554),sid(33652)]  
def handle_data(context, data):  
    close_prices = history(2, '1d', 'close_price')  
    for stock in context.stocks:  
        print stock.symbol  
        print close_prices[stock].values[0]  

Thank you all for these wonderful answers. Your help is greatly appreciated!